EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: rhb on June 23, 2020, 02:20:58 pm
-
EDIT: I have abandoned this project. The level of noise produced by people ignorant of DSP who insist upon pursuing fallacious arguments ad nauseam makes continuing pointless.
Have Fun!
Reg
-------------------------------------------------------------------------------------------------------------------------------------------------
I've been kicking the idea of doing this around for quite a while and have decided to go ahead. This will be a long running thread in which I apply a wide variety of tests to DSOs, primarily entry level models.
I have acquired a rather substantial set of lab gear so I can perform a lot of rather severe tests. My goal in this is to embarrass the OEMs into improving their products. So I am going to focus on all the things that are wrong of which there are a great many in *all* the DSOs I've tested including a Keysight MSOX3104T and a Rohde & Schwarz RTM3104 both of which are $20K MSRP instruments.
I have for signal sources:
<40 ps 1 MHz and 10 MHz square wave generators from Leo Bodnar
10 MHz 100 ps pulse generator from Leo Bodnar
<20 ps rise time 200 kHz pulses from Tektronix 11801 calibrator and SD-24 sampling head
dual output frequency GPSDO from Leo Bodnar
Keysight 120 MHz AWG with <1 ps jitter spec
For comparisons of signal sources:
Tektronix 7104 1 GHz analog scope
Tektronix 485 350 MHz analog scope
Tektronix 11801 sampling with 12 GHz SD-22 and 20 GHz SD-24 and SD-26 sampling heads
LeCroy DDA-125 1.5 GHz DSO
Lecroy DDA-120 1 GHz DSO
The current list of "contestants":
Rigol DS1102E
Rigol DS1202Z-E
Owon XDS-2102A
Instek MSO-2204EA
I've not decided on which model Siglent to get. For cost reasons I'm leaning towards the SDS-1202X-E as it is directly comparable to the DS1202Z-E.
Have Fun!
Reg
-
Should be interesting! How about the Siglent SDS2000X Plus? Seems to get a lot of interest here. I don't own one but do have 5 other scopes ... always have been a sucker for scopes :o
-
Sounds like fun! I like it that you have a couple of older scopes in there for comparison, for cases where the "state of the art" is moving backwards...
-
I'd throw in an RF generator as well as a test source. At least it will allow to determine the actual bandwidth. A pulse generator only shows step behaviour.
-
I bought the 485 and 7104 precisely because the Keysight and R&S scopes were so bad.
This really aimed at helping first time scope buyers understand what they are getting for what is for most a major purchase. So if it's my money, I'm going with the entry level scopes. I'll include other stuff if I get a unit on a 30-60 day loan. And then only after I have worked out the test program.
Around 1990 I paid Tucker Electronics in Dallas over $300 for a 20 year old Dumont 1062 dual channel 60 MHz scope. About 5 days after the 30 day warranty expired, the horizontal sweep quit working. I was quite devastated as I'd saved a long time for that. I eventually fixed it. But the trauma was palpable and locating the broken solder joint took every evening for a week plus a good bit of the weekend.
@nctnico I've got an 8648C, however, if you know your Fourier transform theory you can determine BW from the step response *very* accurately. Oliver Heaviside pioneered using the step response for frequency domain analysis. The major fault with *all* the DSOs I've tested is attempting to get too much BW for the sample rate and using very steep anti-alias filters. So the step response rings like mad.
Have Fun!
Reg
-
You are right about the step response. I Pavlov-ed into rise time measurements.
-
The SDS-1202XE is a nice fit to this contest list (SDS2000X Plus is already in a different class among others due to FFT), although it makes me wonder what's the point of this test?
Entry level DSOs are not about BW or step response right?
Nevertheless if that's the focus I don't see how rolloff will be characterized with the listed equipments.
This kind of DSOs are frequently used to check non repetitive signals (like not so fast buses, transients) and for that they are pretty OK.
With modern higher BW DSOs you get not just BW but a bunch of other tools, like decent FFT (Siglent is pretty good in this; but this is not the only math) or plenty of protocol decoding options plus there are fancy triggers (to name a few: runt, pattern, zone; not all of them are in entry level though). Another story is measurements (with histograms but this is also not entry level) and cursors (on math like FFT).
-
.................
I've not decided on which model Siglent to get. For cost reasons I'm leaning towards the SDS-1202X-E as it is directly comparable to the DS1202Z-E.
Fair enough, although for a fair comparison covering many models their respective sampling rate specs should be considered.
-
when does this war begin? I just have to take my first entry level oscilloscope; my choice will be between siglent1202 and rigol1202, I don't know the other two brands
^-^
-
when does this war begin? I just have to take my first entry level oscilloscope; my choice will be between siglent1202 and rigol1202, I don't know the other two brands
^-^
I don't yet have a Siglent. The DS1202Z-E came yesterday. I've had the others for quite a while. The Owon is *not* usable as a DSO, but it will acquire 20 Mpts of 12 bit data in a single shot. So unless you need that don't consider it. The Instek GDS-2000E is quite good and as an all around scope is the best I've tried so far. But the GDS-2072E is twice the price of the SDS1202X-E or DS1202Z-E.
So I'll get started with some basic stuff this evening if all goes well.
Have Fun!
Reg
-
.................
I've not decided on which model Siglent to get. For cost reasons I'm leaning towards the SDS-1202X-E as it is directly comparable to the DS1202Z-E.
Fair enough, although for a fair comparison covering many models their respective sampling rate specs should be considered.
Not only am I working for free, I'm paying for the scopes. So I'm not about to do a comprehensive review of everything even if I got loaners for all of them.
This is to help people like CharlotteSwiss.
Reg
-
Can you make up your mind if you are trying to help CharlotteSwiss or embarrass the OEMs?
-
Fair enough, although for a fair comparison covering many models their respective sampling rate specs should be considered.
tautech, this should be a comparison irrespective of what they have under the hood.
::)
Reg has already stated (maybe in another thread) step response performance is one of the reasons for venturing down this road yet at fast timebases for fast edges sufficient sampling rate comes into play.
Single shot captures will need interpolation to show meaningful results and the more samples it has to work with will influence reproduction.
What's under the hood will matter !
-
Dont you guys understand this gibberish talks make no sense to the beginners you are "trying to help". You will need to first educate them on what that technical terms all mean.
-
Dont you guys understand this gibberish talks make no sense to the beginners you are "trying to help". You will need to first educate them on what that technical terms all mean.
Found the thread. I agree. Step response is the last thing a beginner will worry about. For most measurements I do it doesn't even matter.
-
Accompanying videos would be cool
-
I am a beginner squared in the field of oscilloscopes; I certainly don't look for the little things of you experts from the oscilloscope, but ease of use and software stability (no bugs).
but fate seems already (almost) sealed for me: reg he ruled out for me Owon, rigol1102 it seems old of hardware, rigol 1202 they say out there that it's based on z old platform, instek is prohibited price... on the square he remains alone siglent 1202 ^-^
thanks for wars
Charlotte
-
but fate seems already (almost) sealed for me: reg he ruled out for me Owon, rigol1102 it seems old of hardware, rigol 1202 they say out there that it's based on z old platform, instek is prohibited price... on the square he remains alone siglent 1202 ^-^
I think Charlotte is the one teaching many here... ::)
Edit: I removed all my previous BS to not clutter Reg's thread.
-
In many respects, the DS1102E looks better to me than DS1202Z-E. So don't jump to conclusions. The point of this is to apply exactly the same tests to a series of instruments and show what they do well and what they do badly. My observation of Keysight and Rohde & Schwarz new $20k instruments is they are buggy as all hell. I'm told that the Tektronix 3 series crashed during a demo for a friend. It seems that for $20k you get the privilege of being a beta tester for the top tier OEMs.
The DS1202Z-E has a bigger screen, but it is permanently cluttered by menus, so the trace display area is about the same as the DS1102E.
Sorry, but I am *not* doing videos. Not no, but hell no!
Step response *is* the last thing that a novice will think about. It's also the single best test of the performance of the AFE and the DSP design. With 30+ years of DSP experience in the oil industry, I can look at the step response and draw the AFE filter profile on a cocktail napkin. TANSTAFL You cannot have a DSO BW which is more than 50-70% of Nyquist and have the traditional analog Tek <3% overshoot.
If you need to test signal integrity for things like ringing on a data line, how do you do that if the step response has a 7% overshoot such as the Keysight MSOX3104T I bought and returned? I suspect that the MSOX3054T is actually a decent scope as it *should* be able to provide a decent step response, though the FFT is still crap.
I shall explain very thoroughly the tests and what they mean. However, this is an evening entertainment, not a job. I have other things I have to do. So it will not be quick. And I still need to get an SDS-1202X-E or similar.
Have Fun!
Reg
-
In many respects, the DS1102E looks better to me than DS1202Z-E. So don't jump to conclusions. The point of this is to apply exactly the same tests to a series of instruments and show what they do well and what they do badly.
And I still need to get an SDS-1202X-E or similar.
in fact, old scilloscope does not necessarily mean that it is scarcer; hope you also get siglent1202x-e ;)
-
I have sent requests to my contacts at Siglent, Rigol, Owon and Good Will (Instek) informing them of this effort and asking that they facilitate responses to the issues that are raised.
I also stated that I shall be reporting on how responsive OEMs are.
Have Fun!
Reg
-
Around 1990 I paid Tucker Electronics in Dallas over $300 for a 20 year old Dumont 1062 dual channel 60 MHz scope. About 5 days after the 30 day warranty expired, the horizontal sweep quit working. I was quite devastated as I'd saved a long time for that. I eventually fixed it. But the trauma was palpable and locating the broken solder joint took every evening for a week plus a good bit of the weekend.
Ah, Tucker, I remember drooling over scopes in their catalog around that time. I was not even a teenager yet so they were all WAY beyond what I could afford, though I did pick up a Tek 531A for $40 at a garage sale somewhere in that era.
-
I was 36 and had spent 24 years in school. So most of the stuff was way beyond what I could afford also. And Tucker was not cheap. I did a lot better at First Saturday before they kicked out the RF guys and made it just computers.
Reg
-
For some reason the theme "Junkyard Wars" comes to mind. :-DD
-
but fate seems already (almost) sealed for me: reg he ruled out for me Owon, rigol1102 it seems old of hardware, rigol 1202 they say out there that it's based on z old platform, instek is prohibited price... on the square he remains alone siglent 1202 ^-^
Instek can be had real cheap right now:
https://www.tequipment.net/Instek/GDS-1054B/Digital-Oscilloscopes/?search=true (https://www.tequipment.net/Instek/GDS-1054B/Digital-Oscilloscopes/?search=true)
-
but fate seems already (almost) sealed for me: reg he ruled out for me Owon, rigol1102 it seems old of hardware, rigol 1202 they say out there that it's based on z old platform, instek is prohibited price... on the square he remains alone siglent 1202 ^-^
I think Charlotte is the one teaching many here... ::)
Only if you completely ignore the prices. It's obvious that you should get something better if you spend more money.
-
In many respects, the DS1102E looks better to me than DS1202Z-E.
It has a lot less features but it displays wiggly lines just fine. You can get the DS1052E even cheaper and hack the bandwidth to 100MHz.
-
Step response *is* the last thing that a novice will think about. It's also the single best test of the performance of the AFE and the DSP design. With 30+ years of DSP experience in the oil industry, I can look at the step response and draw the AFE filter profile on a cocktail napkin. TANSTAFL You cannot have a DSO BW which is more than 50-70% of Nyquist and have the traditional analog Tek <3% overshoot.
Step response is the first thing I think about because oscilloscopes are primarily time domain instruments.
Something else I would test for now is variation of step response and bandwidth with signal level as this is a problem with the Rigol DS1000Z series which suffers from overload problems; this is why some users report 200+ MHz bandwidth while others report bandwidth commensurate with the specifications.
-
Something else I would test for now is variation of step response and bandwidth with signal level as this is a problem with the Rigol DS1000Z series which suffers from overload problems
All 'scopes suffer that if the signal goes wildly offscreen, the DS1054Z is no worse.
Or maybe you're referring to abuse of the 'fine' vertical control - switch the vertical response to a range which is too low and crank the fine control to try and bring it back up again. I don't really see that as a problem despite it being used endlessly as a stick to beat Rigols with.
-
Instek can be had real cheap right now:
https://www.tequipment.net/Instek/GDS-1054B/Digital-Oscilloscopes/?search=true (https://www.tequipment.net/Instek/GDS-1054B/Digital-Oscilloscopes/?search=true)
for Americans yes, but for those who live in Europe, shipping and taxes are a drain :-\
-
A brief tutorial on the relationship between filter passband in the frequency domain and time domain response.
We consider 3 cases, full amplitude to Nyquist, a corner frequency at 80% of Nyquist and a corner frequency of 50% of Nyquist. The respective cases are shown here:
[attachimg=1]
We next look at the time domain impulse response of this "ideal" low pass filter. It's rather less than ideal if you want to discriminate multiple impulses.
[attachimg=2]
A 500 MSa/s DSO has a Nyquist at 250 MHz. So if we set the corner frequency to 80% of Nyquist this is what we get:
[attachimg=3]
If we are conservative and choose a corner frequency of 50% of Nyquist we get this:
[attachimg=4]
The undershoot at the tail end of the impulse can be reduced by changing the shape of the ramp. I chose to stick to the most basic demonstration to start with. When I get to evaluating particular scopes I'll use the step response to show the actual filter passband.
the Octave and gnuplot files are attached.
If you want to get a detailed understanding of this I recommend:
The Fourier Transform and Its Application
Ronald Bracewell
I have the 2nd ed and it is my most frequently used math book because it has a pictorial dictionary of transform pairs in the back.
Have Fun!
Reg
Dave's server is confused and is not letting me post PNG files. I had posted them when I saw I'd forgotten to change the X axis label. It's been throwing a hissy ever since.
-
Does multiple sampling of the same (repetitive) signal get around some of the limitations of the Nyquist sample rate? E.g. if the impulse was made to repeat...
-
These are the plots for my previous post
[attachimg=1]
[attachimg=2]
[attachimg=3]
[attachimg=4]
-
Here are the Octave and gnuplot files
The point of this 3 piece :-( post is TANSTAFL. The time domain and frequency domain are bound together by the Fourier transform.
I can easily get a nice clean step with <1% overshoot. But I can't claim a -3 dB corner at 80% or more of Nyquist.
Nyquist & Shannon showed what was required to accurately represent a *band limited* signal with *regular* sampling without aliasing.
Sampling scopes collect a single sample per waveform and dither the time of the sample to form a sweep. The Tek 11801 does this at 20 femtosecond intervals.
It is possible to to use random sampling and avoid aliasing entirely, but that is a *very* different topic called "compressive sensing". So far no one here has taken me up on a discussion of that publicly, though one person has in a PM revealed he understands the topic to some degree.
Have Fun!
Reg
-
Sampling scopes collect a single sample per waveform and dither the time of the sample to form a sweep.
You mean older scopes do this. Modern scopes that can be used also for non-repetitive signals (can be important even for hobbyist to check various buses and transients) just take samples at regular intervals dictated by the sampling rate.
-
The acquisition mode makes a difference here, right?
I.e. the behaviour you describe assumes "Normal" acquisition mode?
-
Something else I would test for now is variation of step response and bandwidth with signal level as this is a problem with the Rigol DS1000Z series which suffers from overload problems
All 'scopes suffer that if the signal goes wildly offscreen, the DS1054Z is no worse.
Not all oscilloscopes suffer signal distortion from overload; sampling oscilloscopes by the nature of how their sampling works are immune. Some oscilloscopes implement series, shunt, or feedback clamping to reduce or eliminate overload recovery time.
The DS1000Z series suffers from overload problems even for signals which are completely within its dynamic range and this explains the great variations in reported bandwidth but months ago I pointed out other evidence of overload caused by insufficient full power bandwidth.
Or maybe you're referring to abuse of the 'fine' vertical control - switch the vertical response to a range which is too low and crank the fine control to try and bring it back up again. I don't really see that as a problem despite it being used endlessly as a stick to beat Rigols with.
Whatever the detailed cause, better oscilloscopes do not suffer from it. The shape of a signal within the input range should never change with amplitude or position. Good oscilloscope manufacturers figured this out in the 1950s.
Sampling scopes collect a single sample per waveform and dither the time of the sample to form a sweep.
You mean older scopes do this. Modern scopes that can be used also for non-repetitive signals (can be important even for hobbyist to check various buses and transients) just take samples at regular intervals dictated by the sampling rate.
He means *sampling* oscilloscopes which sample before amplification. DSO stands for digital storage oscilloscope which whether modern or old, operates in a completely different way.
It is possible to to use random sampling and avoid aliasing entirely, but that is a *very* different topic called "compressive sensing". So far no one here has taken me up on a discussion of that publicly, though one person has in a PM revealed he understands the topic to some degree.
I am aware of it but do not find it that interesting because HP was the only one to implement it in DSOs that I know of:
https://www.keysight.com/upload/cmc_upload/All/exp66.pdf (https://www.keysight.com/upload/cmc_upload/All/exp66.pdf)
It does not avoid aliasing entirely but instead converts the tones produced by aliasing into noise.
-
I mean what I said. A "sampling scope" is a very different instrument from a DSO. Do a search on the term.
Here's an aliasing example. I've generated a 10 Hz sine wave sampled at 100 samples/second and a 60 Hz sine wave sampled at the same sample rate.
[attachimg=1]
[attachimg=2]
[attachimg=3]
The Nyquist is 50 Hz, so the aliasing makes it appear that the 60 Hz signal is at 40 Hz rather than 60 Hz.
[attachimg=4]
Here's what the actual samples look like with straight line interpolation between sample points.
[attachimg=5]
Edit: This is completely borked. These are not the figures I posted for this.
-
On my scope, you can choose acquisition modes that "oversample" the signal (e.g. averaging, high resolution, peak detect etc.) and I was wondering how that affects Nyquist (if at all!).
-
The 60Hz signal is 10Hz over the 50Hz Nyquist limit and folded back to 40Hz. I think I've seen already couple of times like that before. Is there anything more to be seen here?
-
On my scope, you can choose acquisition modes that "oversample" the signal (e.g. averaging, high resolution, peak detect etc.) and I was wondering how that affects Nyquist (if at all!).
Averaging happens after the aliasing produced by decimation so has no useful effect on aliasing. High resolution mode and peak detection happen before decimation so prevent aliasing up to (half) the maximum sample rate of the digitizer; high resolution does this by limiting bandwidth with a real time FIR filter and peak detection does this by returning a limited histogram of the original signal.
I consider peak detection invaluable but it is less necessary on DSOs with long record lengths where the maximum sample rate can be sustained at slower time/div settings. My ancient Tektronix 2230 with a 1k record length and peak detection can pick up a 100 nanosecond GPS pulse per second signal at any sweep speed which would require at least a 10 megasample record length (or segmented memory) otherwise. Peak detection might be considered a very limited form of DPO (digital phosphor oscilloscope) operation where DPO mode returns a complete 3D histogram of the signal.
-
Many DSO can still use equivalent time sampling. This is similar to the sampling scopes in many respects. It was especially popular with the early DSOs from the 1990s, but many new one still have it as an option.
Besides using it for a substitute for a higher sampling rate, it may also be used as averaging to get high amplitude resolution / less noise.
It can make a difference to the Nyquist limit, though depending on the way it is implemented (e.g. if the is a fixed ADC clock) it may be help in all cases.
The modern entry level DSOs tend to be 1-2 GSPS and some 100 MHz BW, so not that much limited by the sampling rate. It however gets worse if all channels are used and the sampling rate usually cut in half.
Besides the BW limit, there may also be a slew rate limit in some cases and thus amplitude dependent BW.
-
Good article on equivalent time vs real time sampling here: https://www.edn.com/random-repetitive-sampling-take-2/ (https://www.edn.com/random-repetitive-sampling-take-2/)
-
Or maybe you're referring to abuse of the 'fine' vertical control - switch the vertical response to a range which is too low and crank the fine control to try and bring it back up again. I don't really see that as a problem despite it being used endlessly as a stick to beat Rigols with.
Whatever the detailed cause, better oscilloscopes do not suffer from it. The shape of a signal within the input range should never change with amplitude or position. Good oscilloscope manufacturers figured this out in the 1950s.
"Better" oscilloscopes cost more then $350 and they'll still have design compromises somewhere.
The DS1054Z is beneath your sensibilities? Fine, don't use one... but understand that many people still find it a very useful tool.
-
"Better" oscilloscopes cost more then $350 and they'll still have design compromises somewhere.
My observation with the Keysight MSOX3104T and Rohde & Schwarz RTM3104 is the biggest compromise is software quality in addition to the same cheating on specs as Rigol, etc.
Reg
-
... in addition to the same cheating on specs as Rigol, etc.
The "cheating" is the industry standard.
(and some of the very few that don't do it are Rigols, eg. MSO5000 series)
-
Here's a rigorous explanation of the issue with the way the industry specifies DSOs:
[attachimg=1]
Note that whereas you only need a 7 pole filter to adequately suppress aliasing using an Fc of 50% of Nyquist, if you raise that to 80% you need a 22 pole filter which is pretty much impossible to build on a production line. So what they do is they let the signal alias and they apply a digital filter to kill the region in which the aliasing appears.
BTW I misstated the suppression at Nyquist earlier. It's -6*(#bits-1).
I'll add the minimum phase time domain impulse response later. The server problem this morning beat me up pretty badly trying to post the previous plots. I spent over an hour trying to upload the figures.
Have Fun!
Reg
-
This is the step response of my Tek 485, easily the best portable analog scope ever built. This one was refurbished and given a rough cal. The vertical attenuator section is *not* well adjusted. The procedure takes a lot of time and had the person I bought it from done that to my standards he'd have been lucky to make minimum wage. So no criticism, I just want people to understand that a well tuned 485 is a good bit better. I'll eventually get around to adjusting this one, but it takes most of a day because everything interacts with everything else. So you iterate through all the settings.
The scope is set at 1 ns/div
[attach=1]
The dotted lines are 5 divisions apart, so the solid graticules are the 10% and 90% points and it easily meets the 0.35/BW rise time with minimal overshoot. In this case none, but at other attenuator settings there is some, hence my assessment that it needs to be given a full cal treatment.
This is a "good" scope from the days when Tektronix still set the standard for scopes and set a very high bar. Not the best scope Tek made, but the best portable scope they ever made.
Have Fun!
Reg
-
Many DSO can still use equivalent time sampling. This is similar to the sampling scopes in many respects. It was especially popular with the early DSOs from the 1990s, but many new one still have it as an option.
Besides using it for a substitute for a higher sampling rate, it may also be used as averaging to get high amplitude resolution / less noise.
Some DSOs which use digital triggering support equivalent time sampling but any aliasing if present corrupts the digital trigger so I am not sure they should be counted. If aliasing is present, then the sin(x)/x reconstruction has multiple solutions so the trigger position is indeterminate.
Besides the BW limit, there may also be a slew rate limit in some cases and thus amplitude dependent BW.
I am sure diagnostics and measurement is great fun when the results depend on the volt/div setting. Where have I seen that before? Oh, yes, it was Rigol.
-
Note that whereas you only need a 7 pole filter to adequately suppress aliasing using an Fc of 50% of Nyquist, if you raise that to 80% you need a 22 pole filter which is pretty much impossible to build on a production line. So what they do is they let the signal alias and they apply a digital filter to kill the region in which the aliasing appears.
When I need to take a good look at a signal I simply turn off the channels I don't need and that raises the sample rate.
...maybe this is a big advantage of the 4-channel, 100MHz 'scopes over the 2-channel, 200Mhz 'scopes. They both have the same sample rate but one has a much bigger margin between bandwidth and Nyquist.
-
The DS1202Z-E does quite nicely on the step response in single channel mode, but in dual channel with the corner at 80% of Nyquist it's rather strange. I've not yet analyzed the error in dual channel mode. I blew almost 2 hours this morning battling the server problem Dave was having trying to post a few plots before the server went completely belly up.
In single channel mode, the Nyquist is at 500 MHz which gives a generous margin and a good step response. A DS1054Z should be much better as a 50 MHz scope in 4 channel mode.
I may try to get a minimum phase impulse response generated for the Butterworth filter examples tomorrow. A lot depends upon how much time I spend cleaning up my shop so I can build my new lab.
As I have a 100 ps pulse generator, I'm going to take advantage of the simplicity of using that to measure the DSO AFEs.
Properly they should be switching anti-alias filters when they change the sample rate. We shall see. As I stated at the beginning, I have some large hammers to hit things with. And know how to use them.
Have Fun!
Reg
-
Note that whereas you only need a 7 pole filter to adequately suppress aliasing using an Fc of 50% of Nyquist, if you raise that to 80% you need a 22 pole filter which is pretty much impossible to build on a production line. So what they do is they let the signal alias and they apply a digital filter to kill the region in which the aliasing appears.
When I need to take a good look at a signal I simply turn off the channels I don't need and that raises the sample rate.
...maybe this is a big advantage of the 4-channel, 100MHz 'scopes over the 2-channel, 200Mhz 'scopes. They both have the same sample rate but one has a much bigger margin between bandwidth and Nyquist.
Depends on structure but yeah usually... on pretty much all budget / mid 4 channel scopes its 1 adc rail per 2 ports so getting full specd bandwidth is almost always using 1+3,1+4,2+3,2+4
-
When I need to take a good look at a signal I simply turn off the channels I don't need and that raises the sample rate.
That is not exactly a recommendation for DSOs which suffer from that problem. In the past DSOs which shared digitizers between channels supported equivalent time sampling so the result was longer display generation time.
For a long time Tektronix had distinct "real time" DSOs which dedicated one digitizer to each channel instead of sharing them so the problem was recognized not long after DSOs became generally available.
...maybe this is a big advantage of the 4-channel, 100MHz 'scopes over the 2-channel, 200Mhz 'scopes. They both have the same sample rate but one has a much bigger margin between bandwidth and Nyquist.
On a 4-channel DSO using one of the other channels as a trigger input also compromises performance if digital triggering is used which is almost always the case now. A modern 2-channel DSO with an extra trigger channel might not suffer from that.
-
I'd like an explanation of why digital triggering on a live channel is a problem other than bad design or running out of FPGA fabric.
I can see many ways someone could bungle it, but no reason a skillful implementation would have problems.
Reg
-
I'd like an explanation of why digital triggering on a live channel is a problem other than bad design or running out of FPGA fabric.
I can see many ways someone could bungle it, but no reason a skillful implementation would have problems.
If the digitizers are shared between channels like we are discussing, then using a channel as a trigger source without displaying it has the same effect as displaying it lowering the sample rate of the other channels.
-
I'd like an explanation of why digital triggering on a live channel is a problem other than bad design or running out of FPGA fabric.
I can see many ways someone could bungle it, but no reason a skillful implementation would have problems.
If the digitizers are shared between channels like we are discussing, then using a channel as a trigger source without displaying it has the same effect as displaying it lowering the sample rate of the other channels.
Interestingly my GW Instek scope doesn't have this limitation.
-
If the scope does not get the lower sampling rate from using a trigger channel, it may not have digital trigger, but still old style analog trigger. This can be an advantage with equivalent time sampling.
One could do a proper implementation of the digital trigger for equivalent time sampling, but it takes extra effort to remove the aliasing part from the trigger channel. This would be more like having a good AA filter for the trigger and removing (set to higher BW) it for the signal path for equivalent time sampling. So it may need 2 ADCs if the trigger is from the signal.
I don't think one would find and need 7 th order AA filters in the usual scopes, at least not the classical butterworth type (there are usually a few parasitic limitations involved anyway). Usually the amplitude to higher frequency goes down and some low level aliasing is accepted. So the AA filter is not suppressing down to the full ADC resolution. I would not expect much visible difference if only good for 5 bits.
Switching the speed between 1 or 2 channels per ADC (or the other way around) would need switching the AA filter, but the steps further down when the sampling rate is reduced because of limited memory / low horizontal speed usually use digital averaging, so that no extra filter is needed. With some of the cheap scopes the sampling rate is no longer that limiting so no need to use 50% or even 80% of the Nyquist limit. Using a relatively large fraction of the Nyquist limit is more like a thing of the mid class scopes at some 500 MHz BW. The limited ADC speed was a big thing in the early days (1990s).
-
When I need to take a good look at a signal I simply turn off the channels I don't need and that raises the sample rate.
That is not exactly a recommendation for DSOs which suffer from that problem.
It's something that doesn't happen very often. If the "solution" is to pay triple price for something else then it's simply not worth it, I can use the money more productively elsewhere.
Some people still sleep at night despite things like this. Get over it.
Edit:
It's obviously quite severe on a LeCroy DDA-125 with a 1.5 GHz BW and a 1 GHz Nyquist in 4 channel mode.
Lecroys and Keysights need to turn off channels, too. How do you feel about that?
-
When I need to take a good look at a signal I simply turn off the channels I don't need and that raises the sample rate.
That is not exactly a recommendation for DSOs which suffer from that problem.
It's something that doesn't happen very often. If the "solution" is to pay triple price for something else then it's simply not worth it, I can use the money more productively elsewhere.
Not just that but many high-end DSOs (>1GHz >4Gs/s) from Lecroy and HP/Agilent/Keysight offer to trade-off samplerate versus number of channels available. In some cases (Agilent 54845A for example) even the bandwidth is limited when using a lower samplerate.
A much more interesting question would be: for what kind of measurements does the step response matter and for which ones it doesn't. And if it matters, can the situation be improved by adding an external filter to get a more gradual roll-off?
-
Another trick when having many channels to spare is to put them in parallel, to look at the same signal, and thus eliminate some noise from the analog stages. This is good especially for single events, when averaging can not be used.
-
I cannot see *any* reason that aliasing is acceptable on anything other than a sampling scope where it is defacto unavoidable if you set the instrument up so it is aliased. However, that proved useful as a means of testing the linearity of the time base in my 11801. The manual suggested that factory calibration data had been lost when the NVRAM failed, but a lengthy investigation showed that there were no errors. I deliberately aliased a ramp from my Keysight 33622A which has a spec of <1 ps jitter.
To the extent that it is present in commercial DSOs is to be determined. It's obviously quite severe on a LeCroy DDA-125 with a 1.5 GHz BW and a 1 GHz Nyquist in 4 channel mode.
Have Fun!
Reg
-
I cannot see *any* reason that aliasing is acceptable
It's not excusable on an expensive 'scope where there's no reason to skimp on the filters in the analog front end.
-
I cannot see *any* reason that aliasing is acceptable
It's not excusable on an expensive 'scope where there's no reason to skimp on the filters in the analog front end.
Sometimes, in photography, photographers will forgo the optical anti-alias filter in front of the digitizing sensor in order to get higher level of detail, on the understanding that it can lead to aliasing in some circumstances (moiré patterns if there are fine regular repeating patterns in the image - but in practice most images do not have details that are repetitive "sine waves" so for most images moiré is not a problem, so you can enjoy the higher level of detail).
Is there an equivalent way to think about an electronic signal - i.e. where you may expect most of the high frequency content to be noisy rather than distinct sine waves, so you are actually better off overall with a less aggressive AA filter?
-
Is there an equivalent way to think about an electronic signal - i.e. where you may expect most of the high frequency content to be noisy rather than distinct sine waves, so you are actually better off overall with a less aggressive AA filter?
Definitely.
eg. The owner of this thread is having to use special pulse generators and cables to generate frequencies high enough to reveal these "problems".
Most oscilloscope users will never see them because they'll be pre-filtered by the capacitance in ordinary 'scope probes and cables.
-
I think you hit the nail on the head many times with your post.
I don't think one would find and need 7 th order AA filters in the usual scopes, at least not the classical butterworth type (there are usually a few parasitic limitations involved anyway). Usually the amplitude to higher frequency goes down and some low level aliasing is accepted. So the AA filter is not suppressing down to the full ADC resolution. I would not expect much visible difference if only good for 5 bits.
The probes supplied with the entry level scopes contribute a lot to the BW limitation, that is in real life the trouble is more to get enough BW than that aliasing has noticeable effect to signal shape. Especially that harmonics have smaller and smaller amplitude both by nature and due to BW limitation. There is one place though where aliasing do show up quite remarkable and that's the FFT output. This is due to that some of the scopes have excellent dynamic range (approx. 100dB) after the FFT.
Normally it's more an issue to get the probing to such level that the output is not a degraded sine with fast signals. While it's very easy to provoke scopes via hanging a purpose made circuit to the BNC input that outputs a low fundamental but fast rise time signal probably nothing could be further away from the typical real life use than this.
There is though one user error that can get quite much overshoot/ringing: Impedance misalignments and reflections. These scopes typically have only high impedance input.
Switching the speed between 1 or 2 channels per ADC (or the other way around) would need switching the AA filter, but the steps further down when the sampling rate is reduced because of limited memory / low horizontal speed usually use digital averaging, so that no extra filter is needed. With some of the cheap scopes the sampling rate is no longer that limiting so no need to use 50% or even 80% of the Nyquist limit. Using a relatively large fraction of the Nyquist limit is more like a thing of the mid class scopes at some 500 MHz BW. The limited ADC speed was a big thing in the early days (1990s).
Although I've brought up the idea in one of these threads, I pretty heavily suspect that few scopes (especially in entry level) implement filter switching when ADC has to be divided among multiple channels and as a consequence sampling rate is reduced per channel.
There are many hard to answer but interesting questions that you've brought up. Most VGAs have separate output for triggering where a lower rate ADC could be used. I don't remember any comparison or even advantage/disadvantage summary of this or the main channel FPGA based digital triggering.
For the reduced sample rate due to acquisition memory limitation two (multi) stage properly filtered decimation could indeed provide excellent result but does it fit to FPGAs of entry level scopes?
-
Sometimes, in photography, photographers will forgo the optical anti-alias filter in front of the digitizing sensor in order to get higher level of detail, on the understanding that it can lead to aliasing in some circumstances (moiré patterns if there are fine regular repeating patterns in the image - but in practice most images do not have details that are repetitive "sine waves" so for most images moiré is not a problem, so you can enjoy the higher level of detail).
Most lens are very far from that resolution (and hard to focus on that level) so that the sensor could cause moiré, so the "analogue front end" filters enough that no AA filter is needed. Excellent example and finally a non-car analogy.
-
Is there an equivalent way to think about an electronic signal - i.e. where you may expect most of the high frequency content to be noisy rather than distinct sine waves, so you are actually better off overall with a less aggressive AA filter?
Definitely.
eg. The owner of this thread is having to use special pulse generators and cables to generate frequencies high enough to reveal these "problems".
Most oscilloscope users will never see them because they'll be pre-filtered by the capacitance in ordinary 'scope probes and cables.
Most users aren't sufficiently sophisticated in their probing techniques and would not recognize the issues for what they are if they saw them.
Fast scopes such as the 7104 and 11801 are 50 ohm *only*.
I'm somewhat bemused that no one seems to have taken any note of the figures I posted yesterday. And are offering up conjectural arguments instead of demonstrating their assertions with Octave or Matlab.
I'd very much welcome a demonstration of the difference between a 2 pole & 7 pole, causal anti-alias filters with Fc at 50% & 80% of Nyquist when fed a square wave.
I *know* the answer from spending 30+ years in seismic research and programming where waveform fidelity is paramount. So consider it a challenge. If you assert something doesn't matter, demonstrate it mathematically. There are plenty of programs to simulate a scope probe response and Octave can handle most of the math. Though I did discover that the Butterworth filter stuff in the signal package for Octave is duff. Not quite sure what it actually does as it doesn't resemble the definition of a Butterworth filter in "An Introduction to Digital Signal Processing" by John H. Karl which I highly recommend. I've got over 6 ft of books on DSP and it is by far the one I reference most often.
Have Fun!
Reg
-
Is there an equivalent way to think about an electronic signal - i.e. where you may expect most of the high frequency content to be noisy rather than distinct sine waves, so you are actually better off overall with a less aggressive AA filter?
Definitely.
eg. The owner of this thread is having to use special pulse generators and cables to generate frequencies high enough to reveal these "problems".
Most oscilloscope users will never see them because they'll be pre-filtered by the capacitance in ordinary 'scope probes and cables.
Not just that. A step (from an infinite time -1 to an infinite time +1) contains a very wide frequency spectrum. Somewhat repetitive signals OTOH consist of harmonics of their fundamental frequency. In order to have problems with aliasing and/or sharp roll-off from the anti-aliasing filter the harmonics have to be in the frequency band where the anti-aliasing filter has an effect on the signal which is large enough to make it visible.
-
The first several generations of quality digital cameras all had relatively strong anti-alias (AA) filters in front of their sensors to remove any chance of aliasing. It turned out, in the long run, that as sensors got fine grained enough (high enough spatial sample rate), photographers preferred the extra sharpness of not having an AA filter at all... so the problems caused by aliasing (which are real, and can and do happen) were not considered "worth the price" of reduced performance in other situations (and, in any case, most problems can be mitigated later in the workflow).
The Nikon D800E was the first camera model (to my knowledge) and allowed photographers to make the trade-off of having no AA filter versus the normal Nikon D800 that kept the traditional AA filter, and was otherwise exactly the same in every other respect.
If the same kind of thing happens when looking at an oscilloscope, we should be careful that we don't end up jumping to conclusions based on e.g. taking a whole bunch of pictures where the unfiltered D800E performs worse than the filtered D800, and then concluding that the D800E sucks... even though the E model takes sharper pictures in 99% of real world situations and you would only choose the filtered D800 if you were taking product shots of patterned fabrics or meshes, or just liked the slightly softer look better!
-
If the digitizers are shared between channels like we are discussing, then using a channel as a trigger source without displaying it has the same effect as displaying it lowering the sample rate of the other channels.
Interestingly my GW Instek scope doesn't have this limitation.
It is not a universal problem but I think it is more common in 4-channel DSOs. 2-channel DSOs with a separate external trigger input may make other arrangements so use of the external trigger does not affect operation of the vertical channels. For instance if the triggering is all done during decimation, then no acquisition memory is required for the external trigger channel. But in some DSOs, the digital triggering has an additional trigger qualification step after acquisition so acquisition memory is still required.
If the scope does not get the lower sampling rate from using a trigger channel, it may not have digital trigger, but still old style analog trigger. This can be an advantage with equivalent time sampling.
How the external trigger operates might be known by checking which trigger modes are available. If there is no difference between the vertical inputs and external trigger input, then the external trigger is digital. But I remember some past DSOs where the external trigger only allowed level triggering indicating that it was analog.
-
Fast scopes such as the 7104 and 11801 are 50 ohm *only*.
Sure, but their usage case is completely different than "manually poking around electronic circuits looking for signals".
-
Not just that. A step (from an infinite time -1 to an infinite time +1) contains a very wide frequency spectrum. Somewhat repetitive signals OTOH consist of harmonics of their fundamental frequency. In order to have problems with aliasing and/or sharp roll-off from the anti-aliasing filter the harmonics have to be in the frequency band where the anti-aliasing filter has an effect on the signal which is large enough to make it visible.
A fast (e.g 100 ps) pulse is an almost flat spectrum over a large BW. This is why the seismic industry records the impulse responses of the recording system input amplifiers and filters. It verifies that all the channels are working properly which is a *big* deal when you're being paid $10-20 million to acquire the survey. The client oil company has a person on board, the observer, whose sole job is to make sure that all the many pages of specifications about positioning, sea state and other factors are adhered to. There are contractual allowances for how many guns and receivers may be inoperable.
If you want an accurate waveform, there are certain constraints about filter roll off and stop band at Nyquist which *must* be met. Of course, if you don't care, then it hardly matters, does it?
I use my scopes for a lot more than just manually poking around a circuit. For anything that needs significant BW you have to provide test points in the board design. Otherwise you'll never get a valid signal sample. U.FLs that you can jumper in with a blob of solder are very good for that.
If you provide a numerical example demonstrating your assertions above, I'll produce a counter example disproving them. Otherwise I'm not going to bother until I need to demonstrate why certain DSOs behave the way they do.
Have Fun!
Reg
BTW "Infinite +/- 1" does not exist. ;)
-
Fast scopes such as the 7104 and 11801 are 50 ohm *only*.
Sure, but their usage case is completely different than "manually poking around electronic circuits looking for signals".
The 7104 with its MCP CRT is particularly useful for "manually poking around electronic circuits looking for signals" which works fine with low-z dividing probes, but the 7104 supports high impedance probes also up to 200 or 350 MHz and of course faster active probes were available. Later Tektronix released slower but still fast oscilloscopes with MCP CRTs for exactly that application.
Consider what application the 7104 was specifically designed for and how the Soviets solved the exact same problem. It had nothing to do with electronics.
-
Got a link for the Soviet solution?
-
BTW "Infinite +/- 1" does not exist. ;)
Sure it exists. It's infinite.
-
Not just that. A step (from an infinite time -1 to an infinite time +1) contains a very wide frequency spectrum. Somewhat repetitive signals OTOH consist of harmonics of their fundamental frequency. In order to have problems with aliasing and/or sharp roll-off from the anti-aliasing filter the harmonics have to be in the frequency band where the anti-aliasing filter has an effect on the signal which is large enough to make it visible.
A fast (e.g 100 ps) pulse is an almost flat spectrum over a large BW. This is why the seismic industry records the impulse responses of the recording system input amplifiers and filters. It verifies that all the channels are working properly which is a *big* deal when you're being paid $10-20 million to acquire the survey. The client oil company has a person on board, the observer, whose sole job is to make sure that all the many pages of specifications about positioning, sea state and other factors are adhered to. There are contractual allowances for how many guns and receivers may be inoperable.
If you want an accurate waveform, there are certain constraints about filter roll off and stop band at Nyquist which *must* be met. Of course, if you don't care, then it hardly matters, does it?
Again: It all depends on whether the waveform has frequency content in the area around the Nyquist frequency. If you look at a 400MHz signal using a 1GHz scope like the RTM3004 (using 2.5Gs/s) then the 2nd harmonic is well within the pass-band of the anti aliasing filter and the 3rd harmonic is well beyond. It is just as easy to argue that having a sharp roll-off in the anti-aliasing filter gives you the widest bandwidth for the lowest samplerate. Having a more gradual roll-off (with less bandwidth) also distorts a signal by cutting out the high frequencies. Sure, you won't get ringing but you'll also lose bandwidth. It is a typical case of picking your poison depending on the application. There is no free lunch here. On many oscilloscopes you have several bandwidth limiters; you can use these to create a more gradual roll-off.
40MHz signal with steep edges from an RF synthesizer with full bandwidth (and aliasing artefacts):
[attachimg=1 width=1000]
Bandwidth limited to 500MHz to get a more gradual roll-off:
[attachimg=2 width=1000]
-
Don't you think it is sort of cheating to say we only give you an accurate waveform under certain conditions? And leave the user to figure out when their waveform reflects reality?
-
BTW "Infinite +/- 1" does not exist. ;)
Sure it exists. It's infinite.
It violates the definition of "infinite".
-
Got a link for the Soviet solution?
I managed to find it:
https://groups.io/g/TekScopes/message/79995
With this providing some additional context:
https://groups.io/g/TekScopes/message/89431
The US Government required instruments to fit inside a 19" rack while the Soviet solution was to make their scan converter CRTs 6 meters long to reduce deflection requirements.
-
BTW "Infinite +/- 1" does not exist. ;)
Sure it exists. It's infinite.
It violates the definition of "infinite".
No. You removed the words time which turns the statement into nonsense.
-
It violates the definition of "infinite".
Not true. If you go that route then even "infinite" doesn't exist. (good read (https://www.theory-of-reciprocity.com/infinity.html))
-
Don't you think it is sort of cheating to say we only give you an accurate waveform under certain conditions? And leave the user to figure out when their waveform reflects reality?
It seems a 100% valid question to ask what a particular scope actually does when pushed to the limit - no different from wanting to find the limits of a camera or lens by testing with a resolution chart.
Where things become a judgement call, is whether something is "good enough" for what you are trying to do. Even in the seismic world, there must come a point where enough is enough and you use the equipment you have, understanding its limitations?
-
Don't you think it is sort of cheating to say we only give you an accurate waveform under certain conditions? And leave the user to figure out when their waveform reflects reality?
Read my signature 8)
You have to define 'accurate waveform' first. If you are going after a step response or square wave then the image on the screen will never be accurate simply because both types of signals need infinite bandwidth to reproduce. Read my example very carefully again. On the RTM3004 (just an example) you can cater to both situations where you want a gradual roll-of OR the most bandwidth. Show a step response without ringing by enabling the bandwidth limit OR look at a 400MHz square wave in a way it still looks (somewhat) like a square wave instead of a sine wave using full bandwidth. Which case is cheating?
-
SilverSolder: Yes, but aliasing is *not* acceptable and at 24 bit acquisition challenging. I presume that they are actually substantially oversampling and using digital filters as pure analog filters would be quite impossible.
Compromises are operational. The air guns fail and receiver and recording channels fail. Air guns being the biggest issue. I've never been on a boat, but I presume that they can retrieve an individual air gun from the array, service it and replace it without stopping. The financial incentive for being able to do that is huge as you can't stop the boat. So you have to reshoot the line. At somewhere in the neighborhood of a half million a day for the boat and crew turning the boat with 6-12 25 km streamers towed behind it around twice to reshoot a segment is *very* expensive and it comes out of the acquisition contractor's pocket. The boat is only making a few knots and has to make very wide turns to avoid tangling the streamers.
The location of the receivers towed behind the boat at the time of each shot is specified as a few meters. If the contractor doesn't achieve that they are obliged to reshoot until they do or not get paid. With billion dollar decisions hanging on the results, they are not playing. The only DSP that gets more serious is war.
The entire issue is under what conditions is discretely sampled data *accurately* representative of the analog reality? Seismic was recorded on 2" analog tape with 21 tracks before digital became possible. And *no one* was recording digital data for actual exploration work until it was as good as the analog data.
The more I learn about the DSP skills of the typical EE the more agog I am. And *not* in a good way. There are certainly EEs who can match the seismic community, but I've come to conclude they all work in the defense industry on highly classified projects. And most of the rest wouldn't recognize a clue if they fell over it.
nctnico: I've read your signature, many times. I now understand it much better. It's very simple. What would an analog scope of adequate BW for the task display? . That's what a proper DSO should show. I bought my 485 and 7104 precisely because I came to realize how bad DSOs were. That made me want an arbitrar I could trust. As there is *no* physical system that can *produce* an infinite BW step you are citing a red herring.
Have you done a full cal on an analog scope in the 465, 475, 485 class? A key requirement is a fast step generator. I tried to build one which it turns out was pretty good, but not good enough. Neither I nor the EE who helped me test it at work recognized the impedance matching issue so it seemed to ring excessively and I concluded I had failed. It actually does about 1.5 ns rise time on a 5 V step quite nicely. Ultimately I bought a Tek 106 <1 ns step generator to calibrate my 465.
This is intended to be about evaluating DSOs, not "DSP 101".
Have Fun!
Reg
-
40MHz signal with steep edges from an RF synthesizer with full bandwidth (and aliasing artefacts):
It looks like just sinx to me, improving with 2 samples on the fast the edge instead of 1 or none.
-
Not too far off the mark... But I would exchange spying for war and the kind that looks down ;)
-
Don't you think it is sort of cheating to say we only give you an accurate waveform under certain conditions? And leave the user to figure out when their waveform reflects reality?
Yes, but what's the alternative?
-
40MHz signal with steep edges from an RF synthesizer with full bandwidth (and aliasing artefacts):
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010072;image)
This image works better for me. I can look at that and immediately see a bandwidth limited square wave.
I have no idea what the hell's going on in this one (and your description makes no sense):
Bandwidth limited to 500MHz to get a more gradual roll-off:
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010076;image)
-
@Fungus: The second screen shot is taken with the 500MHz bandwidth limit on. This mimics the behaviour of an oscilloscope with a more gradual roll-off (like analog oscilloscopes and the older equivalent time sampling oscilloscopes have).
-
nctnico: I've read your signature, many times. I now understand it much better. It's very simple. What would an analog scope of adequate BW for the task display? . That's what a proper DSO should show.
No because this would mean that you'll need a way higher samplerate to achieve the same bandwidth. 5 to 10 times more if you want to allow for a Gaussian roll-off. However for a lot of measurements this doesn't make sense so with a factor of 2.5 and a steeper anti-aliasing filter you can get way more bandwidth from the same sampling system. And if you want a more gradual roll-off you enable the bandwidth limiter (at the expense of bandwidth) which will make a DSO behave more like an analog oscilloscope where it comes to the frequency response. IMHO the real problem is that you expect the frequency response from a DSO to be the same compared to an analog scope for a given bandwidth. That just isn't the case. BTW several of the higher end oscilloscopes do some signal processing to improve the frequency response. There is a video from HP/Agilent/Keysight floating around somewhere on how this is done.
This is intended to be about evaluating DSOs, not "DSP 101".
Unfortunately DSO and understanding of DSP (or better put: how signals are sampled) can not be seperated. To really understand the limits of a DSO you need some understanding of signal sampling theory.
-
This image works better for me. I can look at that and immediately see a bandwidth limited square wave.
But then, the question is: was that signal generated like that by the generator, with small wiggles before each edge, or that wiggle before each edge is a DSP artifact introduced by the oscilloscope?
On an analog oscilloscope, there will be no doubt that the signal wiggles right before each edge. On a digital oscilloscope, we don't know. Might be a perfectly square wave with digital artifacts (because of the oscilloscope's DSP), or it might be because the signal really wiggles before each edge.
On the second screen capture, the same problem: before each edge, there is some small wiggle.
- Is that a DSP artifact?
- Is that how the signal really is from the generator?
- Is that because of some reflections?
Most probably it's the first, oscilloscope's DSP artifacts, but we can not say for sure from the screen capture alone, unless we already know the signal. But if we already know for sure the signal's shape, then why would we even bother looking at it. :)
-
@Fungus: The second screen shot is taken with the 500MHz bandwidth limit on.
This is what confuses me. The first one looks more like I'd expect a "bandwidth limited" signal to look like and the second one is more like a gradual rolloff.
-
@Fungus: The second screen shot is taken with the 500MHz bandwidth limit on.
This is what confuses me. The first one looks more like I'd expect a "bandwidth limited" signal to look like and the second one is more like a gradual rolloff.
The first picture shows ringing which is not in the original signal. The bottom picture looks much more like the original signal. As far as I understand rhb, this is exactly the issue he is addressing.
-
But then, the question is: was that signal generated like that by the generator, with small wiggles before each edge, or that wiggle before each edge is a DSP artifact introduced by the oscilloscope?
Search for Gibbs phenomenon.
Even the top of the affordable (?) scopes like MSO8204 cannot show a correct square (provided that the signal has the spectral content for this) because it's theoretically not possible (due to the BW required).
-
Scopes with Gaussian response have gradual rolloff from very low frequencies, so amplitude accuracy will be impacted. -3dB is 30% percent amplitude error. So for a Gaussian response scope, 1GHz scope will have good amplitude accuracy up to cca 300MHz or about.
Scopes with brickwall response will have very good amplitude accuracy, up to cca 90% of it's stated bandwidth.
Also, scopes with Gaussian response will (by virtue of it's response) slow down any edges that are faster than it can handle (by filtering out all frequency components higher than some frequency).
It will behave same as if you took 1 GHz brickwall response scope and put in 500-600 MHz low pass filter in front of it. Simply as that. It won't show any artefacts (like ringing) because you filtered out parts of input signal that would make problems. Basically, it shows nice pulse from Leo's 30 ps pulser, because it took 30 ps rise time pulse, and filtered it (slowed it down, converted it ) to a pulse with 500ns edge, that is not faster that what is scope's rise time.
And that is exactly what Nico demonstrated (thanks for that, you saved me time of making filter, I had same idea, but not filter at hand).
Basically, whenever you look at the pulse that is seriously faster than than rise time of your scope, you get overshoot. You can fight it by deliberately making analog part of scope slower, oversampling by much larger factor (not very practical or cheap at 1GHz and up), or just accepting the fact that it will overshoot if you try to drive it faster than spec.
Actually, I can argue that after learning of how it all works, I like brick wall response better.
Here is why:
1. Good frequency amplitude accuracy up to 90 % of full bandwidth. I can accurately assess 868 MHz amplitude with a 1 GHz scope.
2. Good rise time measurement accuracy. It will overshoot if you drive it fast, close to the edge of spec , but it will give more accurate rise time measurements right there up to the limits of specification.
3. Overshoot is there only if edge is faster than scope's rise time. Which is actually useful information. It tells me edge is faster than what it can handle, by overshooting..
So overshoot is actually useful. It shows you you are trying to look at the signal that is too fast for your scope, instead of happily hiding anything it doesn't like... You fight it by getting faster scope, not one that hides it better...
-
@Fungus: The second screen shot is taken with the 500MHz bandwidth limit on.
This is what confuses me. The first one looks more like I'd expect a "bandwidth limited" signal to look like and the second one is more like a gradual rolloff.
The first picture shows ringing which is not in the original signal. The bottom picture looks much more like the original signal. As far as I understand rhb, this is exactly the issue he is addressing.
But how do you know what the original signal looks like? You only know what some other oscilloscope showed you!
-
A newbie question:
So, we can definitely make a good comparison between 2 scopes just by looking at a (for example, Bodnar's) pulse response, right? (in terms of signal fidelity)
Or, with only that, we'll be missing important factors?
-
A newbie question:
So, we can definitely make a good comparison between 2 scopes just by looking at a (for example, Bodnar's) pulse response, right? (in terms of signal fidelity)
Or, with only that, we'll be missing important factors?
Yes, it's a definite maybe :-\
-
A newbie question:
So, we can definitely make a good comparison between 2 scopes just by looking at a (for example, Bodnar's) pulse response, right? (in terms of signal fidelity)
Or, with only that, we'll be missing important factors?
Yes, it's a definite maybe :-\
Step response characterization will only show step response. You should test 100 MHz scope with a pulse that has 1-3 ns risetime and you will get nice results, for a 1 GHz 400ps scope 150-300 ps would show you nice response...
As I said Gaussian response scopes do exactly that, they convert very fast speed signal to slow enough it can handle it...
To compare brickwall response scopes, you need to do proper frequency sweep to get frequency flatness. Or use calibrated noise source..
While fun and quick and dirty way to check if you "upgraded" your scope from base 100 MHz to something else, pulse response is not very relatable to usual scope work we all do. Unless you're making pulsers and you need to check them.. In which case you get scope with adequate bandwidth and calibrated and well defined pulse response for that particular purpose.
Most of the time, most complicated part is how to get signal to scope input in a form that even cursory resembles what the signal looks like before you connect some probing crap to it...
On clocks and such, you need to verify rise time and how clean is the edge, somewhere around receiver sampling points. If there is some over/undershoot is mostly of no concern to how well system works.... You look at the inside of the eye, not outside...
-
[...]
So overshoot is actually useful. It shows you you are trying to look at the signal that is too fast for your scope, instead of happily hiding anything it doesn't like... You fight it by getting faster scope, not one that hides it better...
Is that really a safe assumption? - i.e. can the overshoot actually be present in the signal you are looking at... so you actually don't really know what it causing it on your display?
-
While fun and quick and dirty way to check if you "upgraded" your scope from base 100 MHz to something else, pulse response is not very relatable to usual scope work we all do. Unless you're making pulsers and you need to check them.. In which case you get scope with adequate bandwidth and calibrated and well defined pulse response for that particular purpose.
My intent was not having an absolute characterization of an equipment BUT only doing a comparison between 2 equipments.
Why wouldn't the pulse response be useful in comparing 2 brickwall response scopes? Because of the greater probability that they would look the same? But, in the BW of interest, for me, that would mean that both had the same fidelity. Correct?
-
A fast square wave is only one instrument needed to do a full evaluation of a scope. However, if you know what to look for it will tell you quite a lot.
I'm using the following:
Bodnar GPSDO
Keysight 33622A 120 MHz AWG <1 ps jitter
Bodnar <40 ps rise time square wave and 100 ps pulse generators
HP 8648C
I'll also be using Octave and waveform files for a lot of tests.
I recommend reading the performance check and calibration sections of the Tektronix 465 service manual to get an idea of what is involved. It's beautiful reading and in addition the theory of operation discusses how every circuit in the scope works. So you'll not only learn how to evaluate the performance, you'll learn how a good analog scope works.
I made a photo yesterday of the 100 ps pulse as shown by my 485 to send to a friend so I've added it just because it's a nice image of a minimum phase impulse response.
Have Fun!
Reg
-
These are the plots for my previous post
(Attachment Link)
(Attachment Link)
(Attachment Link)
(Attachment Link)
There is an unalterable relationship between the shape of the passband and the time domain response. And a step response will tell you which scope has a better passband. I generated impulse responses because it was easier than generating a minimum phase step response and they show the same thing.
The "ideal" low pass filter has rather less than an ideal time domain response. TANSTAFL
Reg
-
While fun and quick and dirty way to check if you "upgraded" your scope from base 100 MHz to something else, pulse response is not very relatable to usual scope work we all do. Unless you're making pulsers and you need to check them.. In which case you get scope with adequate bandwidth and calibrated and well defined pulse response for that particular purpose.
My intent was not having an absolute characterization of an equipment BUT only doing a comparison between 2 equipments.
Why wouldn't the pulse response be useful in comparing 2 brickwall response scopes? Because of the greater probability that they would look the same? But, in the BW of interest, for me, that would mean that both had the same fidelity. Correct?
Absolutely you could relatively compare. As long as there is no simplistic expectation like some (not you) that rise time directly correlates with bandwidth using some fixed simple formula. And expecting that perfect pulse response will give "perfect" frequency response. It will give frequency rolloff that is perfect for looking at pulses and will have nice response for mathematical analysis.
If you want bandwidth flatness, brickwall is better..
https://www.euramet.org/Media/docs/Publications/calguides/EURAMET_cg-7__v_1.0_Calibration_of_Oscilloscopes.pdf (https://www.euramet.org/Media/docs/Publications/calguides/EURAMET_cg-7__v_1.0_Calibration_of_Oscilloscopes.pdf)
https://www.ece.ubc.ca/~robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf (https://www.ece.ubc.ca/~robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf)
https://m.eet.com/media/1119902/c0882paged.pdf (https://m.eet.com/media/1119902/c0882paged.pdf)
-
A fast square wave is only one instrument needed to do a full evaluation of a scope. However, if you know what to look for it will tell you quite a lot.
I'm using the following:
Bodnar GPSDO
Keysight 33622A 120 MHz AWG <1 ps jitter
Bodnar <40 ps rise time square wave and 100 ps pulse generators
HP 8648C
I'll also be using Octave and waveform files for a lot of tests.
I recommend reading the performance check and calibration sections of the Tektronix 465 service manual to get an idea of what is involved. It's beautiful reading and in addition the theory of operation discusses how every circuit in the scope works. So you'll not only learn how to evaluate the performance, you'll learn how a good analog scope works.
I made a photo yesterday of the 100 ps pulse as shown by my 485 to send to a friend so I've added it just because it's a nice image of a minimum phase impulse response.
Have Fun!
Reg
Is the timebase 100ps per division in the picture?
-
[...]
The entire issue is under what conditions is discretely sampled data *accurately* representative of the analog reality? Seismic was recorded on 2" analog tape with 21 tracks before digital became possible. And *no one* was recording digital data for actual exploration work until it was as good as the analog data.
[...]
Very interesting stuff about what you do in the seismic space, I know very little about it but it sounds fascinating.
Regarding accuracy/precision, I guess nobody likes going backwards - which was often a problem with early digital stuff, like the first CD players which were not really as good as the best analog vinyl tech at the time.
Is there still any question today, that digital methods can be as good or better than anything analog (given money is no object) for seismic purposes?
-
Seismic went all digital as soon as they could build ADCs. You cannot do the exotic processing which is the norm today in an analog system. Imagine 3D ultrasound except that it's 10-15 TB of data with 3x velocity contrasts. That takes a 10-20,000 core cluster 7-10 days per iteration. After each iteration the velocity model has to be updated.
My PhD supervisor's claim to fame was doing deconvolution to remove water column reverberation using a magnetic drum with movable heads and adjustable gain on analog data around 1959. It had been done digitally by hand by Enders Robsinson a classmate of my supervisor's around 1952 by hand digitizing the paper sections and then computing the Wiener inverse aka prediction error aka predictive deconvolution filter and applying it with a desk calculator.
The publication of "The Extrapolation, interpolation and Smoothing of Stationary Time Series" by Norbert Wiener in 1949 really got the oil industries attention and a consortium, the Geophysical analysis Group, was organized at MIT working under Wiener. Two of the 7 students, Enders Robinson and Sven Treitel, wrote the papers that established digital signal processing. These were republished for many years by Seismograph Service Company as "The Robinson and Treitel Reader" for may years. They eventually merged it into a single monograph, Geophysical Signal Analysis", in 1980 just before I started in the oil industry. I learned DSP from that and a book by Kanasewich. "Time Sequence Analysis in Geophysics".
My MS was in igneous aka "hard rock" petrology. I'd never even seen a seismic section before I arrived at Amoco in 1982. They had promised lavish training, but I got only a single 2 week course after 18 months on the job. I taught myself DSP from those two books evenings and weekends. In addition to those two books I had two books on reflection seismology. So I cycled through the stack of 4 books, each time learning a little more that had been incomprehensible the previous pass.
Sadly, the EE community doesn't seem to recognize the importance of seismic exploration as a driving force second only to the military in the development of ADCs.
Here's a bit of background:
https://wiki.seg.org/wiki/Ralph_Harris (https://wiki.seg.org/wiki/Ralph_Harris)
http://gsinet.us/Grapevine.html (http://gsinet.us/Grapevine.html)
TI was a subsidiary of GSI formed to make recording equipment.
Have fun!
Reg
-
Thanks for the background @Reg, a whole fascinating area of applied science that isn't as well known as it ought to be.
-
rhb you stil can't refrain from throwing names, book titles and from praising your own ingenuity, can you :palm:
-
rhb you stil can't refrain from throwing names, book titles and from praising your own ingenuity, can you :palm:
Eh.. better reading than some of the other junk posts that crop up.. i thought it was a decent history post anyway
rhb.. i was wondering if you were oil or not when you said seismic awhile ago.. guess that answers that lol
-
Well I've been fooling with the DS1202Z-E.
I setup my Keysight 33622A for a 10 MHz, 20 mVpp square wave.
First, the spectrum as reported by an HP 8560A
[attach=1]
Next the best I have been able to get from the DS1202Z-E. I really can't figure out what they are doing wrong. I'm in discussion with the Rigol support guy who supplied me with a step response.
Edit: This was a consequence of my not realizing there was another menu of settings. This display is limited to the screen data and uses a boxcar window. I'll post a better result later.
[attach=2]
For comparison the FFT from the Instek MSO-2204EA:
[attach=3]
And the SA app from the Instek MDO-2000E
[attach=4]
The SA app has a much better UI than the FFT function, but as you can see there are significant artifacts. Unfortunately, as this is an unauthorized hack it is unlikely that Good Will will address the issue.
Have Fun!
Reg
-
At present I cannot save more than 12,000 samples to a CSV file. If I try to save a longer file it shows a progress bar for a while and then says "Saving failed"
[attach=1]
[attach=2]
Reg
-
On an analog oscilloscope, there will be no doubt that the signal wiggles right before each edge.
That is not completely true. Frequency and phase compensation of the delay line used in an analog oscilloscope to show the leading edge produces pre-shoot which will become visible if the input edge and oscilloscope is fast enough. The 300 to 400 MHz Tektronix 2465 series suffers from this, but the effect is very minor. Oddly enough, the Tektronix 300 MHz 2440 DSO suffers from it also for some weird reason.
On a digital oscilloscope, we don't know. Might be a perfectly square wave with digital artifacts (because of the oscilloscope's DSP), or it might be because the signal really wiggles before each edge.
I really hate that with DSOs opperating at an insufficient sample rate. It is not a problem when equivalent time sampling is used because the sample rate is so much higher.
Scopes with Gaussian response have gradual rolloff from very low frequencies, so amplitude accuracy will be impacted. -3dB is 30% percent amplitude error. So for a Gaussian response scope, 1GHz scope will have good amplitude accuracy up to cca 300MHz or about.
Something to keep in mind is that a Gaussian response is *not* the same as a Bessel response. Compare the shape of a fast edge on an analog oscilloscope at maximum bandwidth and with the bandwidth limiter; they are not remotely the same. The Gaussian response looks like a slew rate limited edge while the Bessel response should show something like an exponential curve. Most bandwidth limits have a single pole response but some use LC sections for a 2 pole response and I know of one example which was 4 poles which is unfortunate because they complicate noise calculations.
It will behave same as if you took 1 GHz brickwall response scope and put in 500-600 MHz low pass filter in front of it.
So it is *not* the same as that produced bv a typical Bessel low pass filter.
So overshoot is actually useful. It shows you you are trying to look at the signal that is too fast for your scope, instead of happily hiding anything it doesn't like... You fight it by getting faster scope, not one that hides it better...
Or the preshoot and/or overshoot was always there, which is why I have reference level pulse generators to test oscilloscopes.
-
Is the timebase 100ps per division in the picture?
100ps/div would be indeed cool, but given that this is supposed to be the impulse response of a Tektronix 485 (-> 350 MHz), I'm afraid your guess is not realistic and rather an order of magnitude too low.
-
The first picture shows ringing which is not in the original signal.
You must admit, though, that the ringing is eventually introduced by the attempt to reconstruct the waveform between the sampled points with sinc interpolation, which was turned on although the prerequisite requirement for this reconstruction method is not fulfilled.
I guess, displaying this signal in points mode would not show any indications of ringing, and linear interpolation would likely show a more pleasing shape, too.
[ Of course nobody can deny that sampling loses (an infinte amount of) information. Any information about the waveform shape between the sampled points is lost. Consequently, an exact reconstruction of the original signal without any a priori knowledge about the signal is impossible. Knowing that the signal fed into the ADC is band-limited to < fs/2 is one kind of useful a priori knowledge, but other kinds of knowledge which could enable reconstruction are imaginable too. ]
-
The first picture shows ringing which is not in the original signal.
You must admit, though, that the ringing is eventually introduced by the attempt to reconstruct the waveform between the sampled points with sinc interpolation, which was turned on although the prerequisite requirement for this reconstruction method is not fulfilled.
I guess, displaying this signal in points mode would not show any indications of ringing, and linear interpolation would likely show a more pleasing shape, too.
Not really. There is some ringing in the signal and in linear/dot mode you just get a large smeared area. Even with the extra ringing sin x/x interpolation gives the most accurate representation given the circumstances.
-
Is the timebase 100ps per division in the picture?
100ps/div would be indeed cool, but given that this is supposed to be the impulse response of a Tektronix 485 (-> 350 MHz), I'm afraid your guess is not realistic and rather an order of magnitude too low.
So the 100ps pulse has been "smeared out" to become 10x wider?
-
100ps time base on Rigol 1202 humm................ let me think :wtf:
-
It is probably a challenge for that particular model! :D
-
Not really. There is some ringing in the signal and in linear/dot mode you just get a large smeared area. Even with the extra ringing sin x/x interpolation gives the most accurate representation given the circumstances.
OK, so the square wave was already degraded before entering the ADC - e.g. by the frontend.
So the 100ps pulse has been "smeared out" to become 10x wider?
This is to be expected. Since 100ps is pretty narrow in relation to the bandwitdh, it can be considered an approximation of a dirac impule. So the smeared pulse you see is close to the scope's impulse response.
-
Is the timebase 100ps per division in the picture?
100ps/div would be indeed cool, but given that this is supposed to be the impulse response of a Tektronix 485 (-> 350 MHz), I'm afraid your guess is not realistic and rather an order of magnitude too low.
So the 100ps pulse has been "smeared out" to become 10x wider?
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010560;image)
Yes, smearing out a sharp pulse is expected to happen in a low pass causal filter, and if we consider those 100ps to be practically zero time, what we see in the image is the impulse response of the system (filter, oscilloscope, whatever), h(t).
https://www.youtube.com/watch?v=-XnL5zyNfSU (https://www.youtube.com/watch?v=-XnL5zyNfSU)
For an intuitive approach, you can think of the low pass filter like a system with a "speed limit" in energy transfer speed (the system is not allowed high frequency, or high speed transfers, only slow exchanges). Now, if you suddenly drop a lump of energy (that 100ps pulse) in such a filter, the energy exchange "speed limit" will make that lump of energy to stay there for a while, until the circuit somehow manages to deal with it, therefore the smearing in time of the initially very narrow pulse.
Note the causal word in the name "causal filter". A causal filter will never produce wiggles in front of a signal, while a DSP filter might produce such artifacts. As an example, Fourier transform is not causal. If you use Fourier to implement a filter, such a filter will produce some fake (artifacts) output before you even start to apply your signal. That's not how physics works. Other way to look at it is to say one can not do a Fourier transform live (in real time).
In the next image, you see before each edge some fake wiggling.
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010072;image)
That is because of non-causal DSP. Analog oscilloscopes never show that kind of artifacts, because they are made with physical analog components. Physical world is always causal.
Back to the impulse response, h(t), of that analogue oscilloscope
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010560;image)
that wiggling waveform completely characterizes the time response of a system. Starting from it, one can deduce the time response (the waveform) for any shape of a given input signal (or else said, deduce what distorted waveshape we will see on that oscilloscope's display when we apply an arbitrary waveform input).
-
It is probably a challenge for that particular model! :D
:-BROKE
The Wavepro 4Gz model has 20ps HTB (analogue mode) and that's just a touch up the ladder from that Rigol :-DD
The Rigol MSO8000 has 200ps max HTB in analogue mode
-
[...]
Back to the impulse response, h(t), of that analogue oscilloscope
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010560;image)
that wiggling waveform completely characterizes the time response of a system. Starting from it, one can deduce the time response (the waveform) for any shape of a given input signal (or else said, deduce what distorted waveshape we will see on that oscilloscope's display when we apply an arbitrary waveform input).
Thank you for a very educational and concise post!
Presumably the delta function that is put into the causal system contains a certain defined amount of energy - is there some "conservation of energy" going on so you can for example say that the area under the input pulse curve is always equal to the area under the smeared output, or something like that?
-
That pulse is 100 ps wide, so most of the energy is lost because the scope has only 350 MHz BW. The time base is 1 ns/div which is as fast as the old girl can go. However, in an ideal lossless case with a Dirac function which has unit area, the area of the minimum phase BW limited impulse response is proportional to the area of the transfer function.
All the energy is being forced into sine waves which start at T0.
The requirement of signal to be causal is that the real and imaginary parts in the frequency domain of a pure real function in the time domain be the Hilbert transform of each other.
I just found a fast one shot pulse generator I built 40 years ago out of a 7400 NAND gate. The pulse width is governed by the propagation delay of the gates. I couldn't measure it then, but I can now.
Have Fun!
Reg
-
Note the causal word in the name "causal filter". A causal filter will never produce wiggles in front of a signal, while a DSP filter might produce such artifacts. As an example, Fourier transform is not causal. If you use Fourier to implement a filter, such a filter will produce some fake (artifacts) output before you even start to apply your signal. That's not how physics works. Other way to look at it is to say one can not do a Fourier transform live (in real time).
In the next image, you see before each edge some fake wiggling.
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010072;image)
That is because of non-causal DSP. Analog oscilloscopes never show that kind of artifacts, because they are made with physical analog components. Physical world is always causal.
Back to the impulse response, h(t), of that analogue oscilloscope
(https://www.eevblog.com/forum/index.php?action=dlattach;topic=244895.0;attach=1010560;image)
that wiggling waveform completely characterizes the time response of a system. Starting from it, one can deduce the time response (the waveform) for any shape of a given input signal (or else said, deduce what distorted waveshape we will see on that oscilloscope's display when we apply an arbitrary waveform input).
Actually you can apply a causal filter in the frequency domain, but you have to multiply by a complex series in which the imaginary and real parts are Hilbert transform pairs. That's how I created the examples I posted for boxcar, 80% and 50% of Nyquist passband impulse responses. Commonly described in some circles as the Kramers-Kronig relation or condition.
The issue is that if the convolution is being done with an FIR, you have to have more multipliers for a causal filter than for an zero phase symmetric filter. It's just crap DSP to allow using 1+(N-1)/2 multipliers instead of 1+N multipliers.
Thanks for the nice summary. The noise level has gotten rather high.
Have Fun!
Reg
-
That pulse is 100 ps wide, so most of the energy is lost because the scope has only 350 MHz BW. The time base is 1 ns/div which is as fast as the old girl can go. However, in an ideal lossless case with a Dirac function which has unit area, the area of the minimum phase BW limited impulse response is proportional to the area of the transfer function.
All the energy is being forced into sine waves which start at T0.
The requirement of signal to be causal is that the real and imaginary parts in the frequency domain of a pure real function in the time domain be the Hilbert transform of each other.
I just found a fast one shot pulse generator I built 40 years ago out of a 7400 NAND gate. The pulse width is governed by the propagation delay of the gates. I couldn't measure it then, but I can now.
Have Fun!
Reg
Is the energy actually lost, though? E.g. imagine an ideal filter made of just capacitors and inductors - the energy cannot be dissipated... it kind of has to be present in the output, even if it doesn't look the same as it did going in... ?
Edit: If the filter is an RC filter, I can see that energy can be dissipated in the resistive part of the filter. That dissipation must be frequency dependent (or the filter wouldn't work as a filter). I am beginning to see how the shape of the response to a single pulse can tell you what the filter has to look like. In the case of the ideal LC filter above, the response would presumably be a sine wave at the resonant frequency that goes on forever... with an amplitude proportional to the energy in the original pulse. Wish I was better at math, all this is probably obvious if you understand the underlying equations...
-
That pulse is 100 ps wide, so most of the energy is lost because the scope has only 350 MHz BW. The time base is 1 ns/div which is as fast as the old girl can go. However, in an ideal lossless case with a Dirac function which has unit area, the area of the minimum phase BW limited impulse response is proportional to the area of the transfer function.
All the energy is being forced into sine waves which start at T0.
The requirement of signal to be causal is that the real and imaginary parts in the frequency domain of a pure real function in the time domain be the Hilbert transform of each other.
I just found a fast one shot pulse generator I built 40 years ago out of a 7400 NAND gate. The pulse width is governed by the propagation delay of the gates. I couldn't measure it then, but I can now.
Have Fun!
Reg
Is the energy actually lost, though? E.g. imagine an ideal filter made of just capacitors and inductors - the energy cannot be dissipated... it kind of has to be present in the output, even if it doesn't look the same as it did going in... ?
Edit: If the filter is an RC filter, I can see that energy can be dissipated in the resistive part of the filter. That dissipation must be frequency dependent (or the filter wouldn't work as a filter). I am beginning to see how the shape of the response to a single pulse can tell you what the filter has to look like. In the case of the ideal LC filter above, the response would presumably be a sine wave at the resonant frequency that goes on forever... with an amplitude proportional to the energy in the original pulse. Wish I was better at math, all this is probably obvious if you understand the underlying equations...
Except every inductor and capacitor is actually a combination of all 3.. no such this as the ideal component and thus the energy dissipates eventually if nothing else.. as heat
-
Six pages already? Let me bookmark this.
-
Much of the energy is reflected back into the source because of the impedance mismatch.
I've started work on a 40 m ultra portable QRP transceiver and an immediate concern that has arisen is the harmonics reflected back into the PA. Having never designed a transceiver from scratch I don't know what matters and what doesn't. I've got books on filter design on order, but they seem to be taking a long time to arrive. It may be that a pair of diplexers dumping the unwanted spectrum into a 50 ohm load will perform better.
Have Fun!
Reg
-
Is the energy actually lost, though? E.g. imagine an ideal filter made of just capacitors and inductors - the energy cannot be dissipated... it kind of has to be present in the output, even if it doesn't look the same as it did going in... ?
Edit: If the filter is an RC filter, I can see that energy can be dissipated in the resistive part of the filter. That dissipation must be frequency dependent (or the filter wouldn't work as a filter). I am beginning to see how the shape of the response to a single pulse can tell you what the filter has to look like. In the case of the ideal LC filter above, the response would presumably be a sine wave at the resonant frequency that goes on forever... with an amplitude proportional to the energy in the original pulse. Wish I was better at math, all this is probably obvious if you understand the underlying equations...
That's correct, with the mention that voltage is not the same thing as energy. Also, the area of a voltage in time (Volt*second) is not the same thing as energy (Joule).
But, yes, energy is never lost, can not be destroyed, nor created. That is what the Energy Conservation Law says. Energy can only be moved from one place to another, or transformed from one form to another.
Speaking about energy in an LC circuit, indeed, if you somehow manage to put a "blob" of energy into an ideal LC tank, that blob of energy will start to "slosh", back and forth, between the capacitor and the inductor, forever, as you said.
That sloshing forever comes with another mention. Even if the coil and the capacitor and the connections between them have no resistance (are all ideal, so no rezistive losses), we will still have some losses because the circuit will radiate energy as radio waves. So, to keep the energy in the LC tank forever, we will need to somehow isolate the LC tank from the surrounding space, in order to prevent energy losses in the form of radio waves photons.
Such an isolation will be hard to imagine for a wire as a coil and a pair of plates as a capacitor, let's say we will just use an ideal resonant cavity ended in perfect mirrors that will reflect back all the radio waves photons, so the oscillations will last forever.
But even so, with all those ideal parts and materials, our Universe continuously expands with time. It is said that space itself expands, therefore the mirrors of our LC tank will keep going further and further apart. The further away an object is, the more speed that object will gain because of space expansion, and at some point the speed of the distant object will become faster than light.
To our resonant cavity ended in mirrors, that will mean that at some moment the photons will bounce on one mirror, but in their way to the other mirror, they will never reach the other mirror, because the other mirror is speeding away faster than light. So, our lump of energy will reflect for the last time, and that energy will never come back. The oscillations will end.
Whatever we do, it seems like it won't oscillate forever and ever, not even in the ideal case of no energy dissipation.
Nothing last forever. ;D
-
Oooh! That was a very fine explanation!
I'd been getting frustrated by all the BS noise, but that makes up for it.
Reg
-
So overshoot is actually useful. It shows you you are trying to look at the signal that is too fast for your scope, instead of happily hiding anything it doesn't like... You fight it by getting faster scope, not one that hides it better...
I had initially objected to this, but in fact having thought about it, it makes very good sense. The overshoot tells you if you have significant signal above 50% of Nyquist. If that's the case, you do need a faster scope to get an accurate waveform.
In analog scopes there was no real choice about the matter. The BW had to extend far past the -3 dB point to be able to build it. That's not the case with DSOs, so I'll have to amend my criteria.
A DSO can allow aliasing and then slap a boxcar filter on the signal in the FPGA which eliminates the aliased part of the spectrum. Point taken.
Thanks,
Reg
-
Not really. There is some ringing in the signal and in linear/dot mode you just get a large smeared area. Even with the extra ringing sin x/x interpolation gives the most accurate representation given the circumstances.
Some DSOs do not allow sin(x)/x reconstruction to be disabled and even if they do, sin(x)/x reconstruction is still present for the digital trigger and *that* is what smears the results; the trigger point is changing depending on the relationship between the sample trigger and input edge. DSOs with analog triggers do not suffer from this problem and sampled points align correctly. If they did not, then equivalent time sampling would not work.
-
Such an isolation will be hard to imagine for a wire as a coil and a pair of plates as a capacitor, let's say we will just use an ideal resonant cavity ended in perfect mirrors that will reflect back all the radio waves photons, so the oscillations will last forever.
But even so, with all those ideal parts and materials, our Universe continuously expands with time. It is said that space itself expands, therefore the mirrors of our LC tank will keep going further and further apart. The further away an object is, the more speed that object will gain because of space expansion, and at some point the speed of the distant object will become faster than light.
To our resonant cavity ended in mirrors, that will mean that at some moment the photons will bounce on one mirror, but in their way to the other mirror, they will never reach the other mirror, because the other mirror is speeding away faster than light. So, our lump of energy will reflect for the last time, and that energy will never come back. The oscillations will end.
Ooooh ... lots of problem with the physics in this analogy. No good way to fix it, so best to just discard it.
-
Such an isolation will be hard to imagine for a wire as a coil and a pair of plates as a capacitor, let's say we will just use an ideal resonant cavity ended in perfect mirrors that will reflect back all the radio waves photons, so the oscillations will last forever.
But even so, with all those ideal parts and materials, our Universe continuously expands with time. It is said that space itself expands, therefore the mirrors of our LC tank will keep going further and further apart. The further away an object is, the more speed that object will gain because of space expansion, and at some point the speed of the distant object will become faster than light.
To our resonant cavity ended in mirrors, that will mean that at some moment the photons will bounce on one mirror, but in their way to the other mirror, they will never reach the other mirror, because the other mirror is speeding away faster than light. So, our lump of energy will reflect for the last time, and that energy will never come back. The oscillations will end.
Ooooh ... lots of problem with the physics in this analogy. No good way to fix it, so best to just discard it.
Then perhaps you'd be happy to provide a better analogy?
-
Such an isolation will be hard to imagine for a wire as a coil and a pair of plates as a capacitor, let's say we will just use an ideal resonant cavity ended in perfect mirrors that will reflect back all the radio waves photons, so the oscillations will last forever.
But even so, with all those ideal parts and materials, our Universe continuously expands with time. It is said that space itself expands, therefore the mirrors of our LC tank will keep going further and further apart. The further away an object is, the more speed that object will gain because of space expansion, and at some point the speed of the distant object will become faster than light.
To our resonant cavity ended in mirrors, that will mean that at some moment the photons will bounce on one mirror, but in their way to the other mirror, they will never reach the other mirror, because the other mirror is speeding away faster than light. So, our lump of energy will reflect for the last time, and that energy will never come back. The oscillations will end.
Ooooh ... lots of problem with the physics in this analogy. No good way to fix it, so best to just discard it.
Then perhaps you'd be happy to provide a better analogy?
Probably the non-zero resistance in any real LC circuit, and the losses due to radiation, will have caused the oscillations to die out long before the expansion of the universe gets to a point where it influences our scope tests! :D
-
Some DSOs do not allow sin(x)/x reconstruction to be disabled and even if they do, sin(x)/x reconstruction is still present for the digital trigger and *that* is what smears the results; the trigger point is changing depending on the relationship between the sample trigger and input edge. DSOs with analog triggers do not suffer from this problem and sampled points align correctly. If they did not, then equivalent time sampling would not work.
Trigger is indeed a good point. When I think about it, any sub-sample resolution triggering requires interpolation, (a) in order to determin the trigger point location between two adjacent sample points (in case of a digital trigger - as you say), and (b) in order to time-shift the captured samples by a fraction of the sampling interval. Even an analog trigger still requires (b), since the pre-trigger samples are already captured (with an arbitrary ADC clock pahse) when the trigger fires, and I guess even for post-trigger samples it were hard to enforce a different ADC clock phase momentarily, if I assume a PLL generated clock. On the other hand, if sub-sample resolution triggering is renounced, then the trigger point gets time-quantized, leading to trigger-jitter of +- 1/2 sampling interval (for single-shots, this does not matter of course, as subsequent frames don't neet to line-up on the time axis).
-
I had initially objected to this, but in fact having thought about it, it makes very good sense. The overshoot tells you if you have significant signal above 50% of Nyquist. If that's the case, you do need a faster scope...
IMO this applies not only if the sampling theorem was violated, but also if the ringing was introduced by an AA filter which has a steep roll-off. On the other hand, an AA filter with a low cut-off frequency and a gentle roll-off, which degrades the signal softly, could make you believe that the signal does not contain high-frequency contents, although it does.
-
Not really. There is some ringing in the signal and in linear/dot mode you just get a large smeared area. Even with the extra ringing sin x/x interpolation gives the most accurate representation given the circumstances.
Some DSOs do not allow sin(x)/x reconstruction to be disabled and even if they do, sin(x)/x reconstruction is still present for the digital trigger and *that* is what smears the results; the trigger point is changing depending on the relationship between the sample trigger and input edge. DSOs with analog triggers do not suffer from this problem and sampled points align correctly. If they did not, then equivalent time sampling would not work.
No (and that wasn't the effect I was describing; the trigger is rock solid in that case). Sin x/x only goes slightly wrong at sharp, undersampled edges, not at the zero crossing. Still it is likely the trigger system uses linear interpolation anyway. In fact you can get a far more stable trigger in the digital domain because you don't have the noise from the analog circuitry. The biggest trick however is to use the right threshold levels so the traces don't converge in a single dot at the trigger point (a typical effect on Siglent scopes).
-
No (and that wasn't the effect I was describing; the trigger is rock solid in that case).
How else would you explain the "area" you see in points/lines mode? An area with a non-zero horizontal extent is usually an indication that subsequent frames do not line-up exactly in horizontal direction.
I'm in fact interested how it looks exactly. Could you please post a screen shot of point/line mode as well?
-
No (and that wasn't the effect I was describing; the trigger is rock solid in that case).
How else would you explain the "area" you see in points/lines mode? An area with a non-zero horizontal extent is usually an indication that subsequent frames do not line-up exactly in horizontal direction.
I'm in fact interested how it looks exactly. Could you please post a screen shot of point/line mode as well?
The smearing only happens at the edges in non- sin x/x mode because the linear interpolation will draw lines between the sample points.
-
I had initially objected to this, but in fact having thought about it, it makes very good sense. The overshoot tells you if you have significant signal above 50% of Nyquist. If that's the case, you do need a faster scope...
IMO this applies not only if the sampling theorem was violated, but also if the ringing was introduced by an AA filter which has a steep roll-off. On the other hand, an AA filter with a low cut-off frequency and a gentle roll-off, which degrades the signal softly, could make you believe that the signal does not contain high-frequency contents, although it does.
That is exactly the problem with AA filtering in photography - a soft degradation which solves the problem of aliasing, but introduces a new issue of suppressing high frequency content.
This is a trade-off between being able to see the high frequency, and accepting that you are going to see AA issues in some situations.
Perhaps a good question to ask is - how important are the highest frequencies to getting the right idea of what the signal is doing? In other words, how important is the performance at the limit of what a particular scope, rated at a particular bandwidth, can do?
Are we basically talking about 'benign behaviour during extreme conditions' here?
-
I had initially objected to this, but in fact having thought about it, it makes very good sense. The overshoot tells you if you have significant signal above 50% of Nyquist. If that's the case, you do need a faster scope...
IMO this applies not only if the sampling theorem was violated, but also if the ringing was introduced by an AA filter which has a steep roll-off. On the other hand, an AA filter with a low cut-off frequency and a gentle roll-off, which degrades the signal softly, could make you believe that the signal does not contain high-frequency contents, although it does.
A <40 ps step has spectral contents up over 10 GHz. The ringing you see on a slower scope *is* the AA filter shape. In summary, if you see ringing on a fast step and the rise time is faster than .7/Nyquist you need a faster scope. If the rise time is slower than that you've got a signal integrity issue.
It's an interesting subtlety to DSOs I'd not appreciated properly as it is not an issue that arises in seismic work or with an analog scope.
Have Fun!
Reg
-
In summary, if you see ringing on a fast step and the rise time is faster than .7/Nyquist you need a faster scope. If the rise time is slower than that you've got a signal integrity issue.
That is very concise, and easily applicable in practice. :-+
-
A <40 ps step has spectral contents up over 10 GHz. The ringing you see on a slower scope *is* the AA filter shape. In summary, if you see ringing on a fast step and the rise time is faster than .7/Nyquist you need a faster scope. If the rise time is slower than that you've got a signal integrity issue.
It's an interesting subtlety to DSOs I'd not appreciated properly as it is not an issue that arises in seismic work or with an analog scope.
Have Fun!
Reg
Now THAT is very true , on topic and very well said....
-
Not really. There is some ringing in the signal and in linear/dot mode you just get a large smeared area. Even with the extra ringing sin x/x interpolation gives the most accurate representation given the circumstances.
Some DSOs do not allow sin(x)/x reconstruction to be disabled and even if they do, sin(x)/x reconstruction is still present for the digital trigger and *that* is what smears the results; the trigger point is changing depending on the relationship between the sample trigger and input edge. DSOs with analog triggers do not suffer from this problem and sampled points align correctly. If they did not, then equivalent time sampling would not work.
No (and that wasn't the effect I was describing; the trigger is rock solid in that case). Sin x/x only goes slightly wrong at sharp, undersampled edges, not at the zero crossing. Still it is likely the trigger system uses linear interpolation anyway. In fact you can get a far more stable trigger in the digital domain because you don't have the noise from the analog circuitry. The biggest trick however is to use the right threshold levels so the traces don't converge in a single dot at the trigger point (a typical effect on Siglent scopes).
The trigger point looks rock solid because with digital triggering, the trigger point has zero jitter because by definition that is the point being aligned. It *cannot* have any jitter because it is defined as the trigger point.
Go look at the old advertising videos LeCroy made touting how superior their digital triggering was compared to analog triggering and you can see exactly the problem I am describing. The trigger point they show is exact, and the edge immediately above and below the trigger point is smeared. They look pretty but really showed performance no better than an analog trigger and if aliasing is present, worse.
The other source of smearing is intermodulation between the sample clock and input which produces aliasing products and you might be seeing this, however this also smears the trigger point.
-
Not really. There is some ringing in the signal and in linear/dot mode you just get a large smeared area. Even with the extra ringing sin x/x interpolation gives the most accurate representation given the circumstances.
Some DSOs do not allow sin(x)/x reconstruction to be disabled and even if they do, sin(x)/x reconstruction is still present for the digital trigger and *that* is what smears the results; the trigger point is changing depending on the relationship between the sample trigger and input edge. DSOs with analog triggers do not suffer from this problem and sampled points align correctly. If they did not, then equivalent time sampling would not work.
No (and that wasn't the effect I was describing; the trigger is rock solid in that case). Sin x/x only goes slightly wrong at sharp, undersampled edges, not at the zero crossing. Still it is likely the trigger system uses linear interpolation anyway. In fact you can get a far more stable trigger in the digital domain because you don't have the noise from the analog circuitry. The biggest trick however is to use the right threshold levels so the traces don't converge in a single dot at the trigger point (a typical effect on Siglent scopes).
The trigger point looks rock solid because with digital triggering, the trigger point has zero jitter because by definition that is the point being aligned. It *cannot* have any jitter because it is defined as the trigger point.
Go look at the old advertising videos LeCroy made touting how superior their digital triggering was compared to analog triggering and you can see exactly the problem I am describing. The trigger point they show is exact, and the edge immediately above and below the trigger point is smeared. They look pretty but really showed performance no better than an analog trigger and if aliasing is present, worse.
That is exactly the effect that the Siglent DSOs show nowadays. However it is a matter of using the right thresholds to avoid smearing and thus showing a correct waveform; it is a software problem.
-
That is exactly the effect that the Siglent DSOs show nowadays. However it is a matter of using the right thresholds to avoid smearing and thus showing a correct waveform; it is a software problem.
Ooooh! Let's have endless threads on this!
-
Could we please stick to something remotely relevant to a comparison of intro level DSOs? I rather fear that no one except the very opinionated will ever read the actual test results.
The entire point is to apply *exactly* the same tests to multiple instruments with comparisons to higher end equipment so that a prospective buyer can make informed decisions and OEMs who conspicuously fail are motivated to fix the issues.
Reg
-
Could we please stick to something remotely relevant to a comparison of intro level DSOs? I rather fear that no one except the very opinionated will ever read the actual test results.
The type of tests to be performed probably shouldn't be used as a definitive buyer's guide for people who buy $350 'scopes anyway, so... :-//
-
Let's see what the results say. It is definitely interesting to see if there are significant differences between the different brands/models.
-
After reading most of this thread it seems to me there's whole bunch of this going on......
https://www.youtube.com/watch?v=nMHGOHDDaLo (https://www.youtube.com/watch?v=nMHGOHDDaLo)
Some of you guys take this stuff way too serious. :o :palm: Just think, you could be this poor sod. No aliasssing here. >:D
(https://imagizer.imageshack.com/v2/xq90/924/aYCW2q.jpg) (https://imageshack.com/i/poaYCW2qj)
-
@med6753
Ahahaha, LOL, that commercial is pure gold... Thank you for that...
-
A DSO can allow aliasing and then slap a boxcar filter on the signal in the FPGA which eliminates the aliased part of the spectrum. Point taken.
But one needs to take care not to end up with an overall frequency response like in the middle of picture 1.2 (https://www.eevblog.com/forum/testgear/siglent-sds1204x-e-released-for-domestic-markets-in-china/?action=dlattach;attach=465945;image) from the following message (https://www.eevblog.com/forum/testgear/siglent-sds1204x-e-released-for-domestic-markets-in-china/msg1635386/#msg1635386).
(which is supposed to be an unmodofied SDS1204X-E @500 MSPS)
While there is indeed a notch around fs/2, the attenuation recovers at ~2*fs/3 to only -10dB, and is still only -15dB at fs. I don't know, but I guess the apparent notch around fs/2 might be actually a digital low-pass filter, whose frequency response appears mirrored at fs/2 when considering the frequency band from fs/2...fs.
-
For a long time my only scope was a 5 MHz, recurrent sweep, Heathkit IO-18. It has banana jacks for the input. I've tried multiple times to find a new home for it, but none of them used it, so I took it back. It is still looking for a new home.
It's not a great scope, but it is a *lot* better than no scope.
Reg
-
A DSO can allow aliasing and then slap a boxcar filter on the signal in the FPGA which eliminates the aliased part of the spectrum. Point taken.
But one needs to take care not to end up with an overall frequency response like in the middle of picture 1.2 (https://www.eevblog.com/forum/testgear/siglent-sds1204x-e-released-for-domestic-markets-in-china/?action=dlattach;attach=465945;image) from the following message (https://www.eevblog.com/forum/testgear/siglent-sds1204x-e-released-for-domestic-markets-in-china/msg1635386/#msg1635386).
(which is supposed to be an unmodofied SDS1204X-E @500 MSPS)
While there is indeed a notch around fs/2, the attenuation recovers at ~2*fs/3 to only -10dB, and is still only -15dB at fs. I don't know, but I guess the apparent notch around fs/2 might be actually a digital low-pass filter, whose frequency response appears mirrored at fs/2 when considering the frequency band from fs/2...fs.
I'd like an impulse response from that as a CSV file. I question whether the FFT is valid. If it is, there are real problems. But that's also the point of this thread.
Reg
-
In analog scopes there was no real choice about the matter. The BW had to extend far past the -3 dB point to be able to build it.
I woulnd't say far beyond. I used to own an old HP scope with 300MHz bandwidth and that rolled off pretty quick. The amplifier to drive the CRT had a bootstrap to get a bit more bandwidth but I don't think that worked miracles for the phase. I can't remember the model name but it was a mainframe which had a display section on top and room for 2 modules (vertical amplifier and time base).
-
Could we please stick to something remotely relevant to a comparison of intro level DSOs? I rather fear that no one except the very opinionated will ever read the actual test results.
The type of tests to be performed probably shouldn't be used as a definitive buyer's guide for people who buy $350 'scopes anyway, so... :-//
I agree but in the tests I would use to evaluate low end oscilloscopes, instruments like the Rigol DS1000Z series come up last because of functional, design, and documentation problems What rhb proposed when he started this discussion is at least objective.
When I evaluate modern DSOs in person, one of the first things I use is a reference level pulse generator and precision RF stepped attenuator because if there are aberrations or differences at different volt/div settings, then I know the design is junk and no further testing is required.
-
In analog scopes there was no real choice about the matter. The BW had to extend far past the -3 dB point to be able to build it.
I woulnd't say far beyond. I used to own an old HP scope with 300MHz bandwidth and that rolled off pretty quick. The amplifier to drive the CRT had a bootstrap to get a bit more bandwidth but I don't think that worked miracles for the phase. I can't remember the model name but it was a mainframe which had a display section on top and room for 2 modules (vertical amplifier and time base).
Depends on the overshoot. The lower the cutoff slope the less the overshoot. Which is why that was a major criterion for analog scopes.
Reg
-
in the tests I would use to evaluate low end oscilloscopes, instruments like the Rigol DS1000Z series come up last
Instruments like the Rigol DS1000Z are among the cheapest available so that should surprise nobody.
The point of these tests should be to see if the more expensive instruments are really any better than the Rigol.
-
Here are the Rigol DS1102E and the Owon XDS2102A doing FFTs. I'll need to investigate further, but it appears that the FFT on the DS1202Z-E is worse than the DS1102E.
However, until I can get trace data off the DS102Z-E I don't really have a way to tell what is going on other than to feed a signal from the 8648C which is 10-20% above Nyquist and examine the time domain traces. There are a lot of ways to screw up an FFT spectral analysis and I've seen a lot of them over the years.,
Of note, the Owon cuts off the frequency range at 50% of Nyquist. If you switch to 2 channels or turn on 12 bit the sample rate drops to 500 MSa/s and the FFT changes to only show up to 125 MHz.
There is almost no user control of the FFT parameters on either of theses scopes as is the case with the DS1202Z-E also. The MSO-2204EA allows some, but only the SA app provides reasonable control albeit with serious problems with spurious artifacts.
As with the MSO-2204EA and DS1202Z-E the input is a 20 mVpp 10 MHz square wave from the Keysight 33622A. The low signal level is to avoid damaging the 8560A and the SDRs.
As I've also got an SDRplay RSP2 and V2 & V3 RTL-SDR dongles I'll capture spectra with those to compare against the 8560A and the DSOs.
Have Fun!
Reg
Reg
-
You mean, why the spectrum is not mirrored exactly at fs/2 (or even, why is it displayed beyond fs/2 at all)?
My Hantek seems to calculate the FFT from the interplated/upsampled data when I select a fast timebase where the displayed trace needs to be interpolated. Clearly the upsampled data (und thus their FFT) have a much higher Nyquist frequency (several GHz) than the raw ADC samples (fs/2 = 500MHz), and eventually I get a similarly strange display beyond 500 MHz as you. So what I see is obviously the spectrum of the trace as it appears on the screen, which is not necessarily the signal spectrum. Only with a much slower timebase setting (slowest timebase which still uses 1GSPS) I seem to get what I want/expect.
I'm not sure, of course, if the issue with the Rigol is the same. Does the apparent FFT spectrum change when interpolation is switched between sinc and linear?
Btw, hopefully you can save raw samples. My Hantek saves interpolated samples too (at a higher samling rate than ADC), if a too fast timebase was selected.
-
in the tests I would use to evaluate low end oscilloscopes, instruments like the Rigol DS1000Z series come up last
Instruments like the Rigol DS1000Z are among the cheapest available so that should surprise nobody.
The point of these tests should be to see if the more expensive instruments are really any better than the Rigol.
Even ignoring design flaws, the Rigol DS1000Z series would come up last in my list only because Rigol is deliberately misleading with their documentation and design. Being inexpensive has nothing to do with the manufacturer deliberately deceiving its customers.
-
Just for fun, I fired up some ancient equipment to see how it compares with the newer scopes (i.e. interesting to compare used equipment in the entry level price range).
The scope is an Agilent 54622D, fed 400mV of 10MHz square wave from an HP8012A pulse generator synced to a HP3325A for the frequency.
First, the square wave on its own. Is the overshoot due to the flanks of the pulses getting close to the scope's 3dB point (at 100MHz) ?
[attachimg=1]
Close-up of the leading edge of the square.
[attachimg=2]
Then, there is the FFT function.
How does this scope, with its modest 200M samples per second rate, get as high as 4GSamples per second for its FFT function??
Vertical is 10dBV/division, offset by -54dBV
[attachimg=3]
-
Even ignoring design flaws, the Rigol DS1000Z series would come up last in my list only because Rigol is deliberately misleading with their documentation and design. Being inexpensive has nothing to do with the manufacturer deliberately deceiving its customers.
Would you please document those assertions? I'd like to prove or disprove them and compare the performance metric to other OEM products.
Thanks,
Reg
-
How does this scope, with its modest 200M samples per second rate, get as high as 4GSamples per second for its FFT function??
Vertical is 10dBV/division, offset by -54dBV
Probably the same way the DS1202Z-E gets 50 GSa/s, by reporting something which is incorrect because the UI programmer didn't understand how to do the calculations.
Or not reading the API documentation as happened to me when I was the lead scientific programmer on a project. one of the UI programmers complained that my code was crashing. So I went and ran his example under the debugger. He was passing the sample rate in the wrong units which caused my code to calculate the wrong offset in memory. I went back and told him a few minutes later that he was passing an argument in the wrong units and went back to my work. I don't think I said 5 sentences. I just stated what the issue was and thought the matter settled. No recrimination other than read the function header before calling a routine.
But he complained to the project manager and I got lectured about "not being a team player" or such. She also made me go fishing all day as a "team building" exercise when the project started which I thought a huge waste of time.
It's going to get rather interesting when I start pulling sample data into Octave.
Have Fun!
Reg
-
Or not reading the API documentation as happened to me when I was the lead scientific programmer on a project. one of the UI programmers complained that my code was crashing. So I went and ran his example under the debugger. He was passing the sample rate in the wrong units which caused my code to calculate the wrong offset in memory. I went back and told him a few minutes later that he was passing an argument in the wrong units and went back to my work. I don't think I said 5 sentences. I just stated what the issue was and thought the matter settled. No recrimination other than read the function header before calling a routine.
Offtopic: you should write code which can deal with such conditions without crashing. Especially pointer errors shouldn't be allowed to happen because they can go unnoticed for a long time. If your code throws an error then people won't bother you; they figure out they did something wrong themselves.
-
How does this scope, with its modest 200M samples per second rate, get as high as 4GSamples per second for its FFT function??
Vertical is 10dBV/division, offset by -54dBV
Probably the same way the DS1202Z-E gets 50 GSa/s, by reporting something which is incorrect because the UI programmer didn't understand how to do the calculations.
Or not reading the API documentation as happened to me when I was the lead scientific programmer on a project. one of the UI programmers complained that my code was crashing. So I went and ran his example under the debugger. He was passing the sample rate in the wrong units which caused my code to calculate the wrong offset in memory. I went back and told him a few minutes later that he was passing an argument in the wrong units and went back to my work. I don't think I said 5 sentences. I just stated what the issue was and thought the matter settled. No recrimination other than read the function header before calling a routine.
But he complained to the project manager and I got lectured about "not being a team player" or such. She also made me go fishing all day as a "team building" exercise when the project started which I thought a huge waste of time.
It's going to get rather interesting when I start pulling sample data into Octave.
Have Fun!
Reg
Sometimes, engineers get very matter-of-factly in the interest of clarity, and that gets misinterpreted as aggression and/or rudeness by touchy feely people that prefer to present all matters of fact in ways that reduce the chances of anyone feeling offended by said facts. I think a lot of the political divide can be explained by this mechanism. And I don't think either side is right or wrong about this, there's no harm in being considerate, but on the other hand there is nothing wrong with being concise in a professional environment...
Anyway, back to the thread: Is the answer to the conundrum that the real sample rate for the FFT function is the same as the scope's basic sample rate when it showed the waveform itself - in this case, 200Msamples/sec?
-
Here's another thing I don't understand about FFT on oscilloscopes.
How is the FFT function able to display signals above the Nyquist frequency (100MHz in this case, assuming 200Msamples/sec)?
[attachimg=1]
-
I still guess that it might be the same reason as on mine: The FFT is calulated from the samples in the display buffer which are interpolated and re-samped at a higher rate in order to fill the gaps between the original sampling points.
-
I don't know the details of this instrument, but HPAK introduced sample dithering which phase shifted the clock from sweep to sweep. If you divide the actual sample rate by 20 and shift the phase of the clock by that amount on each sweep, then you get an effective sample rate on a repetitive waveform which is 20 times greater. For spectrum analysis which has traditionally been done by sweeping the LO, that is a very viable approach and will produce the same answer.
I've started work on a spectrum analysis code for DSOs with the following API to be released under a Gnu library license:
Input
-------
resolution BW from a fine grained set of choices
visual BW from a display limited range of choices
window choice for all or most of the windows on the Wikipedia list
number of sweeps to average
anti-alias filter complex frequency response
Output
---------
complex spectrum
display interpolation filter
There will probably be a few changes or enhancements, but the above is the minimum functionality. So with a swept input it will provide vector network analysis to the limit of the DSO and signal source.
For those lacking a high level of DSP background, the display interpolation operator is a function of the minimum phase anti-alias filter, which must also be corrected to zero phase, and the window operator.
It's routine stuff in seismic, so very familiar territory for me. The sole limitation is free time. My hope is that this thread will lead OEMs to use it to replace the crap they currently have.
Have Fun!
Reg
-
I still guess that it might be the same reason as on mine: The FFT is calulated from the samples in the display buffer which are interpolated and re-samped at a higher rate in order to fill the gaps between the original sampling points.
You may be right - the FFT number of points is 2048, and the display is 1024... so by doubling up the number of points and interpolating, the "new Nyquist" would now become 200MHz, which is what we can see in the screen grab.
But... how can there be information in the signal beyond the 100MHz bandwidth of the analog front end, even if you interpolate in the screen buffer you are not really adding signal information to what was there in the first place - right?
Perhaps they are actually doing wizardry along the lines that @Reg suggests above.
-
There are lies, damn lies and marketing brochures.
-
Two tests using the FFT in an HP 54542A oscilloscope. 1 MHz square wave in each case, using an HP 3310A Function Generator and an HP 8011A Pulse Generator.
-
But... how can there be information in the signal beyond the 100MHz bandwidth of the analog front end, even if you interpolate in the screen buffer you are not really adding signal information to what was there in the first place - right?
Remember how upsampling by factor N is usually done. Insert N-1 zeros between the original samles, then apply a digital low pass filter. The zero insertion creates harmonics of the original ADC clock, and IM sidebands at each harmonic. The low-pass filter is supposed to eliminate anything beyond the original Nyqist then. Ideally it's a boxcar filter (in the frequency domain), having a sinc pulse response in the time domain (thus resultig in "sinc interpolation"). The ideal filter would indeed clean-up all generated frequencies beyond the original Nyqist. If there are still frequencies left, then the low-pass filter was not a perfect boxcar. Of course it can't in practice, since sinc has an infinite extent, so at least it needs to be truncated. And maybe some pulse-shaping is applied additionaly. While frequencies in the upsampled FFT spectrum beyond the original ADC's Nyqist does not provide additional information about the original signal, they do contain some information about the interpolation kernel which was used to generate the displayed trace from the raw samples. Not sure, though, whether it is sufficient to reconstruct the interpolation kernel, due to noise and zeros in the spectrum.
-
Even ignoring design flaws, the Rigol DS1000Z series would come up last in my list only because Rigol is deliberately misleading with their documentation and design. Being inexpensive has nothing to do with the manufacturer deliberately deceiving its customers.
Would you please document those assertions? I'd like to prove or disprove them and compare the performance metric to other OEM products.
I have posted about them multiple times in the past. Off the top of my head:
1. Confusing peak detection with envelope detection - the former requires hardware support while the later can be done in software.
2. Confusing dual timebase operation with magnification - if these were the same thing, then how could other DSOs with actual dual timebase support have both? There was also something here about confusing delayed trigger as applied to dual delayed timebase operation with a qualified delayed trigger which is completely unrelated.
3. Ground coupling which is implemented late in the amplifier chain - this makes the instrument seem to have less input noise than it really has. This would not be such a big deal if they did not take advantage of it in their marketing materials touting low noise which was anything but; it was low noise because they were not amplifying the noise. This might be responsible for the inability to null small offsets which users have noticed; the instrument can't see them because ground coupling attenuates them.
4. Lack of sufficient full power bandwidth to support the transition time or specified bandwidth - this is why users report wildly different bandwidths for the same model of instrument. It also shows up as a non-linearity from 2 to 10 nanoseconds because of overload or cutoff after a fast edge at certain time/div settings. I suspect this comes about because the preamplifier stage operates over 10 times the voltage range with the same transistors that other oscilloscopes manage and for good reason. Tektronix for instance used transistors which were 3 times faster with an input range which was 1/20th as large.
-
As you're such a fan of Rigol, :o I hope you didn't miss the Capactor Charging In Steps thread.
https://www.eevblog.com/forum/chat/does-a-capacitor-charges-smooth-or-in-stairs/?action=dlattach;attach=1006164;image (https://www.eevblog.com/forum/chat/does-a-capacitor-charges-smooth-or-in-stairs/?action=dlattach;attach=1006164;image)
https://www.eevblog.com/forum/chat/does-a-capacitor-charges-smooth-or-in-stairs/msg3103617/#msg3103617 (https://www.eevblog.com/forum/chat/does-a-capacitor-charges-smooth-or-in-stairs/msg3103617/#msg3103617)
-
I don't know the details of this instrument, but HPAK introduced sample dithering which phase shifted the clock from sweep to sweep. If you divide the actual sample rate by 20 and shift the phase of the clock by that amount on each sweep, then you get an effective sample rate on a repetitive waveform which is 20 times greater. For spectrum analysis which has traditionally been done by sweeping the LO, that is a very viable approach and will produce the same answer.
[...]
That is very clever. I found an HP Journal article that explains how their random sample dithering works to dramatically reduce aliasing artifacts (attached for reference), when doing equivalent time sampling.
Presumably this scope is still limited by the analog front end, which is why we don't see anything past 200MHz even if the sampling circuitry is capable of it (FFT mode).
It looks to me like the FFT function in this scope is exactly the same as in its bigger brother models that have 1Gsample/second sample rates and more analog bandwidth. So, in the lower range models, the FFT is better than the input circuitry is actually able to supply...
-
Two tests using the FFT in an HP 54542A oscilloscope. 1 MHz square wave in each case, using an HP 3310A Function Generator and an HP 8011A Pulse Generator.
This looks like superb performance.
What is the difference between the first and the second test?
-
I missed the developments, will my sds1202 also be tested, or has the 4 channels been preferred?
:popcorn: :popcorn: :popcorn:
-
I still guess that it might be the same reason as on mine: The FFT is calulated from the samples in the display buffer which are interpolated and re-samped at a higher rate in order to fill the gaps between the original sampling points.
You may be right - the FFT number of points is 2048, and the display is 1024... so by doubling up the number of points and interpolating, the "new Nyquist" would now become 200MHz, which is what we can see in the screen grab.
But... how can there be information in the signal beyond the 100MHz bandwidth of the analog front end, even if you interpolate in the screen buffer you are not really adding signal information to what was there in the first place - right?
Perhaps they are actually doing wizardry along the lines that @Reg suggests above.
If the ADC is triggered on the rising edge of the clock, just flipping the phase of the clock would collect the samples between the samples. That could easily be done with a NAND gate on the clock line. It can't be done by interpolation because you don't have the information.
Reg
-
Two tests using the FFT in an HP 54542A oscilloscope. 1 MHz square wave in each case, using an HP 3310A Function Generator and an HP 8011A Pulse Generator.
This looks like superb performance.
What is the difference between the first and the second test?
The HP 3310A function generator happens to have a square wave setting, while the 8011A is purposefully built as a pulse generator.
Setup was done using the HP 3310A. Lots of tweaking of the FFT setup. Aliasing was very easy to encounter even when careful about the setup. I calculated the FFT coefficients to check against the ideal square wave spectrum. I normalized the calculations at 1 MHz. Compared to the 8011A the 3310A drops off more quickly, so by 25 MHz it's down by 8 dB. The 8011A is off by only 2.5 dB at 25 MHz. However, the 8810A has higher power in the even harmonics.
-
Two tests using the FFT in an HP 54542A oscilloscope. 1 MHz square wave in each case, using an HP 3310A Function Generator and an HP 8011A Pulse Generator.
This looks like superb performance.
What is the difference between the first and the second test?
The HP 3310A function generator happens to have a square wave setting, while the 8011A is purposefully built as a pulse generator.
Setup was done using the HP 3310A. Lots of tweaking of the FFT setup. Aliasing was very easy to encounter even when careful about the setup. I calculated the FFT coefficients to check against the ideal square wave spectrum. I normalized the calculations at 1 MHz. Compared to the 8011A the 3310A drops off more quickly, so by 25 MHz it's down by 8 dB. The 8011A is off by only 2.5 dB at 25 MHz. However, the 8810A has higher power in the even harmonics.
Ah, I misunderstood, I thought you were using the pulse generator on both. - Very cool! 8)
-
To keep things simple, one thing at a time:
1. Confusing peak detection with envelope detection - the former requires hardware support while the later can be done in software.
I'm not sure I understand what you mean by this. One does not have to sample the peak to determine the magnitude or when it occurs.
The input here is the 100 ps 10 MHz repetition pulse. The scope samples are 1000 ps apart. It's quite clear that the band limited pulse is well captured in dot mode and shows a ~ 1 ns rise time.
So why are these supposed to be different and how?
I should like to note that the small dip in most of the displays just prior to the rise of the pulse is real. The pulse actually starts there, but there is phase dispersion and delay in the analog circuitry. I'll demonstrate that later using a 20 GHz SD-26 sampling head which can more accurately display a 100 ps pulse.
The ringing before the step in average mode is the result of using a zero phase sinc interpolator. The correct interpolator is the minimum phase impulse response of the AFE which is precisely what is shown using persistence and dot mode.
Watching the live display, the distribution of the dots varies with time. I suspect it is simply the result of clock jitter rather than a design feature. However, dot mode and persistence let you get a very good representation of the pulse.
Have Fun!
Reg
-
Here's the same setup on the GW Instek MSO-2204EA. Interestingly we get a clear picture of the waveform at 5 ns/div, but not at 2 ns/div
Reg
Edit: Sigh. the server has borked itself and the 5 ns/div Instek image has been replaced by the Owon image in the next post by me.
-
Here's the same setup on the GW Instek MSO-2204EA. Interestingly we get a clear picture of the waveform at 5 ns/div, but not at 2 ns/div
Reg
Look this Rigol trigged edge time position. Quite ok.
Look then this Good Will trigged edge position. Totally out of order. Is this GoodWill model at all with digital trigger engine.. In 5ns/din display scale position is 2.5ns wrong when sampling interval is 1ns. 2ns/div time scale just "game over"...
-
I don't recall the Instek behaving this way before, but it gets a little difficult keeping track of what does what with so many scopes around. I plan to reexamine the Instek result and inquire with Good Will about it.
The shift in the trigger point is because I was triggering off the sampling scope trigger signal produced by the pulser to see if that would help the 2 ns/div display and forgot to switch back to triggering on channel 1 when I made the screen dumps.
Here are the Owon XDS-2102A and Rigol DS1102E with the same setup.
BTW My previous FFT result for the DS1202Z-E was dominated by user error. I'll be redoing that soon.
Have Fun!
Reg
-
Look then this Good Will trigged edge position. Totally out of order. Is this GoodWill model at all with digital trigger engine.. In 5ns/din display scale position is 2.5ns wrong when sampling interval is 1ns. 2ns/div time scale just "game over"...
If you look more closely you'll see the actual trigger point is off-screen. What you see is the next pulse. I don't know what rhb did in the second image. I can't reproduce it on my unit.
-
Look then this Good Will trigged edge position. Totally out of order. Is this GoodWill model at all with digital trigger engine.. In 5ns/din display scale position is 2.5ns wrong when sampling interval is 1ns. 2ns/div time scale just "game over"...
If you look more closely you'll see the actual trigger point is off-screen. What you see is the next pulse. I don't know what rhb did in the second image. I can't reproduce it on my unit.
Reg doesn't know either. Rather reminiscent of my 11801 which from time to time gets into a weird state and starts producing bogus events until I do an "autoset". Fortunately, I've learned to recognize the artifacts.
So here's the 5 ns/div display that got trashed by Dave's server and a 1 ns/div after doing an "autoset" and readjusting the time base. I don't think I was able to get faster than 2 ns/div when I made the other displays.
Have Fun!
Reg
-
Here's an FFT from the DS1202Z-E with sensible parameters. Sample length is 12K samples and I'm using a triangle aka Bartlett window in the time domain which is a sinc(f)**2 in frequency giving a sensible looking display of a fast spike train.
Turning anti-aliasing on or off seems to have no effect. I've gotten a FW update, but have not installed it yet.
I have discovered that if I stop the acquisition I can save a PNG file *much* faster.
Have Fun!
Reg
-
So, if the DS1202Z-E is a 200MHz scope, how is it able to display any frequency content at 500MHz with any accuracy in its FFT function?
Doesn't the analog front end limit the signal before the FFT code even sees it?
-
To keep things simple, one thing at a time:
1. Confusing peak detection with envelope detection - the former requires hardware support while the later can be done in software.
I'm not sure I understand what you mean by this. One does not have to sample the peak to determine the magnitude or when it occurs.
That doesn't refer to the DS1000Z series which supports peak detection, which was why the DS1000Z series was the first of their DSOs to draw my serious consideration. The previous Rigol series, which the DS1000Z replaced, lacked the hardware to support peak detection which was a common feature by then (1) so Rigol confused the issue by conflating envelope detection with peak detection so they could claim to support it.
The input here is the 100 ps 10 MHz repetition pulse. The scope samples are 1000 ps apart. It's quite clear that the band limited pulse is well captured in dot mode and shows a ~ 1 ns rise time.
So why are these supposed to be different and how?
Peak detection applies when the maximum sample rate is not available, usually because of a limited record length. It is also not exclusively for peaks; it also prevents aliasing due to a lower than maximum sample rate.
So use a shorter record length or a slower time/div to see its significance.
(1) Peak detection has been around since the late 1980s.
-
So, if the DS1202Z-E is a 200MHz scope, how is it able to display any frequency content at 500MHz with any accuracy in its FFT function?
Doesn't the analog front end limit the signal before the FFT code even sees it?
That's a *very* good question. Just what *is* going on? I just did an update that is supposed to fix the CSV save bug. We shall see. Once I can get a CSV file I can do an FFT and see what the actual AFE response is. I was able to save a 120 Kpts file, but it's been sitting for about 5 minutes saving a 24 Mpts file.
Have Fun!
Reg
-
Some DSOs do not allow sin(x)/x reconstruction to be disabled and even if they do, sin(x)/x reconstruction is still present for the digital trigger and *that* is what smears the results; the trigger point is changing depending on the relationship between the sample trigger and input edge. DSOs with analog triggers do not suffer from this problem and sampled points align correctly. If they did not, then equivalent time sampling would not work.
Trigger is indeed a good point. When I think about it, any sub-sample resolution triggering requires interpolation, (a) in order to determin the trigger point location between two adjacent sample points (in case of a digital trigger - as you say), and (b) in order to time-shift the captured samples by a fraction of the sampling interval. Even an analog trigger still requires (b), since the pre-trigger samples are already captured (with an arbitrary ADC clock pahse) when the trigger fires, and I guess even for post-trigger samples it were hard to enforce a different ADC clock phase momentarily, if I assume a PLL generated clock. On the other hand, if sub-sample resolution triggering is renounced, then the trigger point gets time-quantized, leading to trigger-jitter of +- 1/2 sampling interval (for single-shots, this does not matter of course, as subsequent frames don't neet to line-up on the time axis).
HP describe how they did that in the HP54100A:
https://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1986-04.pdf (https://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1986-04.pdf)
Page 7 "Interpolator"
-
Peak detection applies when the maximum sample rate is not available, usually because of a limited record length. It is also not exclusively for peaks; it also prevents aliasing due to a lower than maximum sample rate.
So use a shorter record length or a slower time/div to see its significance.
(1) Peak detection has been around since the late 1980s.
So "peak detection" substitutes the peak value during the sample period for the value at the time of the sample clock? If so, that has some very *interesting* effects. None of which seems to me desirable.
If that is a correct interpretation I cannot understand why anyone would want to do it except for lack of comprehension of sampling theory. Aliasing is a consequence of regular sampling, so there is *no* possibility of "peak detection" preventing aliasing. This is dead obvious if you simply draw the time domain multiplication and the corresponding frequency domain convolution.
Seismic data is often acquired at not quite regular spatial sampling and there have been hundreds of professional papers written on how to mitigate the effects of that on the data as it causes a lot of problems in seismic imaging.
Have Fun!
Reg
-
Peak detection applies when the maximum sample rate is not available, usually because of a limited record length. It is also not exclusively for peaks; it also prevents aliasing due to a lower than maximum sample rate.
So use a shorter record length or a slower time/div to see its significance.
(1) Peak detection has been around since the late 1980s.
So "peak detection" substitutes the peak value during the sample period for the value at the time of the sample clock? If so, that has some very *interesting* effects. None of which seems to me desirable.
If that is a correct interpretation I cannot understand why anyone would want to do it except for lack of comprehension of sampling theory. Aliasing is a consequence of regular sampling, so there is *no* possibility of "peak detection" preventing aliasing. This is dead obvious if you simply draw the time domain multiplication and the corresponding frequency domain convolution.
Seismic data is often acquired at not quite regular spatial sampling and there have been hundreds of professional papers written on how to mitigate the effects of that on the data as it causes a lot of problems in seismic imaging.
Have Fun!
Reg
The reason would be to avoid missing a short pulse on a slow sweep.
-
Peak detection applies when the maximum sample rate is not available, usually because of a limited record length. It is also not exclusively for peaks; it also prevents aliasing due to a lower than maximum sample rate.
So use a shorter record length or a slower time/div to see its significance.
(1) Peak detection has been around since the late 1980s.
So "peak detection" substitutes the peak value during the sample period for the value at the time of the sample clock? If so, that has some very *interesting* effects. None of which seems to me desirable.
If that is a correct interpretation I cannot understand why anyone would want to do it except for lack of comprehension of sampling theory. Aliasing is a consequence of regular sampling, so there is *no* possibility of "peak detection" preventing aliasing. [...]
I think a lot of people use an oscilloscope for troubleshooting and finding components of signals which they didn't expect in the first place. In this case an oscilloscope is a tool to gather as much information as possible. A peak detector can capture transients which would otherwise be unnotable.
To me, an oscilloscope, which performs antialiasing or lacks a peak detector (or equivalent features like high samplerate and lots of memory), would be useless.
Andreas
-
Peak detection applies when the maximum sample rate is not available, usually because of a limited record length. It is also not exclusively for peaks; it also prevents aliasing due to a lower than maximum sample rate.
So use a shorter record length or a slower time/div to see its significance.
(1) Peak detection has been around since the late 1980s.
So "peak detection" substitutes the peak value during the sample period for the value at the time of the sample clock? If so, that has some very *interesting* effects. None of which seems to me desirable.
If that is a correct interpretation I cannot understand why anyone would want to do it except for lack of comprehension of sampling theory. Aliasing is a consequence of regular sampling, so there is *no* possibility of "peak detection" preventing aliasing. This is dead obvious if you simply draw the time domain multiplication and the corresponding frequency domain convolution.
Well, you have to understand that an oscilloscope's purpose is to look at a signal and not to acquire a signal for processing later on. Without peak-detect you can get all kinds of weird looking signals / miss parts of a signal when the samplerate is too low. I regard peak detect as an essential feature on a DSO. Try to check a 1PPS signal from a time reference for example. The pulse might be too narrow compared to the samplerate available at time/div settings of 500ms/div.
-
If that is a correct interpretation I cannot understand why anyone would want to do it except for lack of comprehension of sampling theory.
They do it when they don't want to display the curve, they only want to count the number of peaks.
The Rigol DS1000Zs do all their calculations using the "on screen" data. The main CPU doesn't appear to have access to the original sample data, only the 1200 byte decimated version displayed on screen. All the measurements are done from that. This leads to a couple of hacks.
The only exception to this is the "memory" FFT which Rigol added after too many people noticed that the 1200 byte "screen" FFT was useless.
-
With careful attention to the FFT settings to keep the span within range of the actual bandwidth of the scope, an almost credible looking FFT comes out of the 54622d.
It is so sensitive to even minor tweaking that it is hard to know when you can trust what you are seeing...
Presumably the magic of HP's dithering methods is what allows greater than Nyquist performance. Scope sample rate is 200Msamples/sec, so Nyquist is 100MHz, at the center vertical line in the screen - but the FFT clearly performs well beyond that.
Offset -50dB, 20dB/division
[attachimg=1]
-
The Fourier transform of a Dirac functional is a constant from DC to infinity.
The 100 ps pulse I've been using is only 10% of the 1 ns sample interval, but it still shows up.
Mathematically even a 1 ps spike should show up at slower sampling rates. However, I find that a 5 ns impulse at 1 second intervals does not consistently show up unless I set peak detection mode. Why I still do not understand, but clearly it is useful.
I do find that the peak detection mode in the DS1102E works just as well as on the DS1202Z-E and the MSO-2204EA. I've not yet tried the Owon.
So my question for David Hess, is which model Rigol?
This thread is intended to be an antidote to rumor and insinuation as much as a goad to fix issues.
Have Fun!
Reg
-
With careful attention to the FFT settings to keep the span within range of the actual bandwidth of the scope, an almost credible looking FFT comes out of the 54622d.
It is so sensitive to even minor tweaking that it is hard to know when you can trust what you are seeing...
Presumably the magic of HP's dithering methods is what allows greater than Nyquist performance. Scope sample rate is 200Msamples/sec, so Nyquist is 100MHz, at the center vertical line in the screen - but the FFT clearly performs well beyond that.
Offset -50dB, 20dB/division
(Attachment Link)
I have a TDS5000 series DPO at work that will happily show you "signals" in FFT out to 10s of GHz. It's a 1 GHz scope. :-DD Sorry, no pictures will be forthcoming from that.
-
Watching the live display, the distribution of the dots varies with time. I suspect it is simply the result of clock jitter rather than a design feature. However, dot mode and persistence let you get a very good representation of the pulse.
Have Fun!
Reg
This was hashed out here a while ago and I think we concluded that, at least on the DS1054Z, that the 'dots' mode was deceptive in that the dots were not actual sample values, but just dots places where the scope thought they should be.
-
With careful attention to the FFT settings to keep the span within range of the actual bandwidth of the scope, an almost credible looking FFT comes out of the 54622d.
It is so sensitive to even minor tweaking that it is hard to know when you can trust what you are seeing...
Presumably the magic of HP's dithering methods is what allows greater than Nyquist performance. Scope sample rate is 200Msamples/sec, so Nyquist is 100MHz, at the center vertical line in the screen - but the FFT clearly performs well beyond that.
Offset -50dB, 20dB/division
(Attachment Link)
I have a TDS5000 series DPO at work that will happily show you "signals" in FFT out to 10s of GHz. It's a 1 GHz scope. :-DD Sorry, no pictures will be forthcoming from that.
It's as if they only had one piece of code that does FFT, taken from their highest end product, that they use everywhere! :D
Look at this ridiculous display on a 100MHz scope:
[attachimg=1]
-
Look then this Good Will trigged edge position. Totally out of order. Is this GoodWill model at all with digital trigger engine.. In 5ns/din display scale position is 2.5ns wrong when sampling interval is 1ns. 2ns/div time scale just "game over"...
If you look more closely you'll see the actual trigger point is off-screen. What you see is the next pulse. I don't know what rhb did in the second image. I can't reproduce it on my unit.
Yep, my error. When I look images I assume he did even somehow comparative tests when he show Rigol images and then GoodWill. Also Owon was there. But yes, GW trig position is out of screen and my previous comment is bullshit. His images are not at all comparable - for what reasons they are - I do not want even guess.
-
With careful attention to the FFT settings to keep the span within range of the actual bandwidth of the scope, an almost credible looking FFT comes out of the 54622d.
It is so sensitive to even minor tweaking that it is hard to know when you can trust what you are seeing...
Presumably the magic of HP's dithering methods is what allows greater than Nyquist performance. Scope sample rate is 200Msamples/sec, so Nyquist is 100MHz, at the center vertical line in the screen - but the FFT clearly performs well beyond that.
Offset -50dB, 20dB/division
(Attachment Link)
I have a TDS5000 series DPO at work that will happily show you "signals" in FFT out to 10s of GHz. It's a 1 GHz scope. :-DD Sorry, no pictures will be forthcoming from that.
It's as if they only had one piece of code that does FFT, taken from their highest end product, that they use everywhere! :D
Look at this ridiculous display on a 100MHz scope:
(Attachment Link)
Yeah, it does seem that the software people just couldn't be arsed to change the code for lesser models, leading to that. My scope is a decent enough scope, the DPO mode is pretty handy sometimes. We have a far better Keysight MSO-X6000 series in the lab though. But that's a $30k+ scope, which is a little outside the price ranges being considered here.
-
So "peak detection" substitutes the peak value during the sample period for the value at the time of the sample clock? If so, that has some very *interesting* effects. None of which seems to me desirable.
That is not how it is implemented.
During decimation which would normally produce a fractional number of points in the acquisition record by dropping samples, instead the maximum and minimum samples during the duration of two points of the acquisition record are saved which necessarily halves the acquisition record length. So an envelope of the input signal is captured during a single acquisition, within the limits of the maximum sample rate.
CCD based digitizers also did peak detection but in the analog domain and I am not sure how; it sure was not with analog comparators or diode peak detectors. I suspect Asgard technology was involved.
If that is a correct interpretation I cannot understand why anyone would want to do it except for lack of comprehension of sampling theory. Aliasing is a consequence of regular sampling, so there is *no* possibility of "peak detection" preventing aliasing. This is dead obvious if you simply draw the time domain multiplication and the corresponding frequency domain convolution.
What is ultimately captured is the envelope of the input signal during a single acquisition record. Under conditions where aliasing would be present because of decimation, a non-graded envelope is captured. A DPO (digital phosphor) style DSO produces a similar display but captures a histogram instead of an envelope so it may be index graded; DPO operation requires an order of magnitude more acquisition memory.
As far as the practicality of the display, in most cases it looks normal but with all peak-to-peak noise displayed unless you are Tektronix who managed to implement a noise reduction algorithm to produce a display which looks like a sample display but also shows detected peaks. Luckily this was Tektronix of the past and this feature was configurable.
Below is an example of what peak detection allows. The display shows 30+ kHz switching noise on an analog control signal for a switching power supply controller during startup so it was a single shot acquisition. A coaxial connection was required to get a clean signal. If peak detection was not used, then the sample rate would have been 50 kHz instead of 100 MHz and the shaded area of the trace would be entertaining but questionable.
This image was how I recognized that Tektronix had implemented a noise reduction algorithm for use during peak detection. I "knew" it in the sense that I had read about it when I read the manual cover to cover but had not made the connection to what it actually did until examining this photograph later and thinking, "That does not look right. Where is the added noise from peak detection? The readout says peak detection." Since then I have never found their noise reduction algorithm lying, but I still never trust it. Honestly though it has revealed details which would normally be missed. I might trust it if Tektronix had described how it works.
The 100 ps pulse I've been using is only 10% of the 1 ns sample interval, but it still shows up.
Mathematically even a 1 ps spike should show up at slower sampling rates. However, I find that a 5 ns impulse at 1 second intervals does not consistently show up unless I set peak detection mode. Why I still do not understand, but clearly it is useful.
Narrow pulses get spread and attenuated by the limited bandwidth before sampling so you end up with a probably to capture a given pulse width to a specific accuracy. On older DSOs, Tektronix actually listed these probabilities and accuracies but later DSOs had sample rates high enough to essential capture the results of narrow pulse which made it through the input amplifiers.
So my question for David Hess, is which model Rigol?
This thread is intended to be an antidote to rumor and insinuation as much as a goad to fix issues.
It has been a couple years but I can check my notes; it was the DS1000D and DS1000E series. From page 2-56 of the manual:
Peak Detect Aqusition: Peak Detect mode captures the maximum and minimum values of a signal. Finds highest and lowest record points over many acquisitions.
That is not peak detection as it was commonly understood and it contradicts later Rigol documentation for their DSOs which do support peak detection. I went to the manual after informally evaluating a DS1000D.
What I find a little ironic is that the ancient Tektronix 2440 series of DSOs are advertised as having peak detection, and they do, but have no such acquisition mode; the designers were a little too clever and implemented peak detection as envelope mode with the number of acquisitions set to the minimum of 1.
-
[...] the designers were a little too clever and implemented peak detection as envelope mode with the number of acquisitions set to the minimum of 1.
Well, technically... :-DD
-
With careful attention to the FFT settings to keep the span within range of the actual bandwidth of the scope, an almost credible looking FFT comes out of the 54622d.
It is so sensitive to even minor tweaking that it is hard to know when you can trust what you are seeing...
Presumably the magic of HP's dithering methods is what allows greater than Nyquist performance. Scope sample rate is 200Msamples/sec, so Nyquist is 100MHz, at the center vertical line in the screen - but the FFT clearly performs well beyond that.
Offset -50dB, 20dB/division
(Attachment Link)
I have a TDS5000 series DPO at work that will happily show you "signals" in FFT out to 10s of GHz. It's a 1 GHz scope. :-DD Sorry, no pictures will be forthcoming from that.
It's as if they only had one piece of code that does FFT, taken from their highest end product, that they use everywhere! :D
Look at this ridiculous display on a 100MHz scope:
(Attachment Link)
It is not at all any kind of problem because A+ brand manufacturer did it, so B brands can follow and if someone claim it is wrong they can jump behind A brand and say "look how They did, so what is wrong in our products" just like example in history I send some images to manufacturer X how they scope have some (other) things what need repair. Next day I get image from manufacturer X when they did same test with A brand one model and they show same thing more bad and ask...how I can think this name X product is bad because this A brand show more bad... what to do after then. Only think that keep your X name shits and I keep my money in my pocket... and shut off mouth because A brand is always right.
-
With careful attention to the FFT settings to keep the span within range of the actual bandwidth of the scope, an almost credible looking FFT comes out of the 54622d.
It is so sensitive to even minor tweaking that it is hard to know when you can trust what you are seeing...
Presumably the magic of HP's dithering methods is what allows greater than Nyquist performance. Scope sample rate is 200Msamples/sec, so Nyquist is 100MHz, at the center vertical line in the screen - but the FFT clearly performs well beyond that.
Offset -50dB, 20dB/division
(Attachment Link)
I have a TDS5000 series DPO at work that will happily show you "signals" in FFT out to 10s of GHz. It's a 1 GHz scope. :-DD Sorry, no pictures will be forthcoming from that.
It's as if they only had one piece of code that does FFT, taken from their highest end product, that they use everywhere! :D
Look at this ridiculous display on a 100MHz scope:
(Attachment Link)
It is not at all any kind of problem because A+ brand manufacturer did it, so B brands can follow and if someone claim it is wrong they can jump behind A brand and say "look how They did, so what is wrong in our products" just like example in history I send some images to manufacturer X how they scope have some (other) things what need repair. Next day I get image from manufacturer X when they did same test with A brand one model and they show same thing more bad and ask...how I can think this name X product is bad because this A brand show more bad... what to do after then. Only think that keep your X name shits and I keep my money in my pocket... and shut off mouth because A brand is always right.
You are probably right, any improvements have to start in the "good" brands (unless one of the "B" brands sees an opportunity to look better than the competition). I wonder how often people actually use the FFT functions on their scopes, though - if it is not a super popular feature, it might always end up as the lowest priority for the r&d team...
-
I personally would never use a scope FFT if I had a spectrum analyzer available. I see the FFT functionality as more of a quick and dirty way to get frequency domain. When it gets more interesting is something like the Tek MDOs where you can time correlate the spectrum analyzer trace with the time domain. If only they weren't so bloody overpriced. :palm:
-
What I find a little ironic is that the ancient Tektronix 2440 series of DSOs are advertised as having peak detection, and they do, but have no such acquisition mode; the designers were a little too clever and implemented peak detection as envelope mode with the number of acquisitions set to the minimum of 1.
I remember it. Long time ago I have owned and used it lot.
2440 Peak detect works well.
These years Tek engineers did nice job - mostly.
-
[...] the designers were a little too clever and implemented peak detection as envelope mode with the number of acquisitions set to the minimum of 1.
Well, technically... :-DD
Envelope mode over 1 acquisition is a contradiction in terms unless peak detection is also used, which is what they did. For documentation purposes, it would have been nice to have a separate peak detection setting and that is what they did with later DSOs.
I remember it. Long time ago I have owned and used it lot.
2440 Peak detect works well.
These years Tek engineers did nice job - mostly.
Considering how they work, they are amazing.
-
I personally would never use a scope FFT if I had a spectrum analyzer available. I see the FFT functionality as more of a quick and dirty way to get frequency domain. When it gets more interesting is something like the Tek MDOs where you can time correlate the spectrum analyzer trace with the time domain. If only they weren't so bloody overpriced. :palm:
I never appreciated until this thread, that you really have to keep your wits about you when looking at the scope FFT, it is way too easy to get it to display near total nonsense in a plausible looking way! :P
-
I wonder how often people actually use the FFT functions on their scopes, though - if it is not a super popular feature, it might always end up as the lowest priority for the r&d team...
I 'used' (as opposed to played around with) the FFT on my Rigol DS1054Z exactly one time--to verify that a garage door opener was transmitting and determine on what frequency--and it was 315MHz. It worked.
-
This was hashed out here a while ago and I think we concluded that, at least on the DS1054Z, that the 'dots' mode was deceptive in that the dots were not actual sample values, but just dots places where the scope thought they should be.
The dots are not raw sample values? Correct.
That doesn't mean they're random though, they have sin(x)/x applied to them until you turn on more than 2 channels at maximum zoom. At this point it seems to use a special rigol reconstruction filter that works slightly better than sin(x)/x on the outer limits.
Bottom line: Any plans along the lines of "I'll download the sample data and process it myself" don't really work out.
Edit: Well.... they do so long as you're not getting too close to Nyquist.
-
Scopes are not acquisition digitizers, despite sometimes being used like that. That is exactly where LeCroy's business is coming from: people who sample signal to analyze it, not just as an interactive scope.
Scope optimized for digitizing will have different priorities to one optimized for interactive work (comparison LeCroy - Keysight 3000/4000/6000 series).
And it is debatable that a scope can be made (without being unnecessarily expensive) that could do both right.
As far as FFT on a scope goes, that is also always a mixed bag. FFT is never trivial. You need to mindful of bin width (resolution bandwidth in SA speak), windowing used (which one has minimum spectral leakage, which one will have best amplitude accuracy at peak, what is correlation between peak size and bin width, what is correlation of numbers of bins, sample rate and bin width, and windowing and amplitudes......... Also aliasing effects will downsample frequencies above Nyquist to lower frequencies and show them there in a wrong place...
But I keep hearing that people say they don't use FFT on scopes because that is SA job. Well, that is true if you're doing RF. FFT on scope is really useful at lower frequencies. Also, it is real time, simultaneous bandwidth. My Picoscope will show you 200 MHz (or another one 5 MHz) of bandwidth simultaneously from single capture. And with 1 Mbins ( 2 MPoints) it will show it with very high frequency resolution.
So yes it is very useful, but it needs to be done properly and you need to understand what are you looking at.
And also, more than once, I just needed to see if something is sending something, so scope FFT is great for this.
The more these kinds of discussions are dragging out, the more I think many people don't understand some basic things.
On some level scopes are like multimeters ( and any other instrument really) : you have 3.5 digit meters, and 4.5 digit meters, and 5.5 and 6.5 and 7.5 and 8.5.
3.5 and 4.5 digit meters are used 99.5% of time and 99% of people. Because most of the time we need to check something is there and it is roughly in spec.
But 3.5 and 4.5 digit meters look retarded and broken when compared to, say, 7.5 digit meters.
So we should stop using that "shit", right. Wrong.
Most of the people will rarely have need for a scope that has "better signal fidelity" than Rigol DS1054Z. Most people will never have need for anything better than Rigol DS5000 and Siglent SDS200X+, even in professional environments, if money is thigh and will need to make do. We buy better because we can afford it (we all like our toys) and sometimes it makes for better productivity. Sometimes we buy it because our customers expect us to have nice shiny "ping machines"
-
I actually find the scope FFT useful especially when performing power analysis on incoming a/c it gives a respectable indicator as to weather you need to dig further into the harmonic content with the real time SA.
Also when exploring data signal fidelity having an eye diagram/histogram and an FFT really does give you a decent chance of chasing down the exact cause of the issues.
Not to be forgotten basic RF probing again will show whether to fire up the SA or not first, though I do use multi domain investigations more these and find it very useful.
Here is my thoughts, you can spend six figures plus on a scope or under £400 as 2N3055 has mentioned and I also advocate WHAT are you looking to achieve with it?
I could happily live with the Rigol MSO 5000 or the Siglant 2000x + and bit more model :-DD both the scopes have many features and have performance that a few years ago on the big boys were charging five figures + for.
For myself I need to ability to perform really low noise dynamic measurements, I have a desirable volt meter and power analyzer, the Rigol 8000 is almost there in certain areas and not in others. The Lecroy Wavepro is just the ticket it does exactly what it is does, no fuss and it delivers accurate repeatable and RELIABLE results at the low noise floor I require. I can justify the cost as business capital purchase, however in the real world do you really need to have such a device 100% of the time?
Careful selection of test equipment FOR YOUR NEEDS will better serve your purpose and importantly budget
-
Reading you all I come to the conclusion that people throw in more money in their scope buy in the expectation that it covers their lack of knowledge of the fundamentals. Some of you show such a deep knowledge that you make do with "almost" any model.
-
2N3055, all very well and eloquently said. :clap:
FFT on scope is really useful at lower frequencies.
(...)
So yes it is very useful, but it needs to be done properly and you need to understand what are you looking at.
That is absolutely true. My DS4014 has terrible FFT, but it is alright when you choose properly its frequency of interest and bandwidth.
3.5 and 4.5 digit meters are used 99.5% of time and 99% of people. Because most of the time we need to check something is there and it is roughly in spec.
Exactly. A 3.5 digit meter, a DS1202Z or a SDS1202X-E are more than enough for heaps of people. Heck, even a cheaper Owon VDS1022 USB-based could be considered the new 20MHz/2ch staple scope to see wiggly lines.
-
If downsampling is being done in a DSO by throwing away samples it would explain the need to invent a new sampling mode. It's also brain dead stupid DSP.
As for the "dots" stuff, the server borked my last post and sent 2 photos to never never land.
However, the DS1202Z-E, DS1102E and MSO-2204EA do *not* have a sinc(t) applied to them. That should be obvious on inspection of the leading edge.
I'll write more when the server is feeling better. There are quite a few other things that have been said that are not accurate
Reg
-
What I'm seeing in the comments above and my tests is pretty appalling.
The DSP in DSOs is a major fail even from the A list. It's not even good enough to get a passing grade as a DSP 101 homework exercise. The first chapter of any DSP text explains aliasing and why you cannot decimate data by throwing away samples. You *have* to low pass filter the data. That's not hard to do and doesn't require a lot of resources, so I'm agog that DSOs are decimating data by discarding samples. The results of my tests with a 5 ns pulse at 1 s intervals showed that "peak detection" is necessary to offset the absolute bodge of downsampling by decimation.
I'd always wondered what "peak detection" was when I got my DS1102E. BTW it does just as well as the 1202Z-E and MSO-2204EA. I'll post photos when the server is feeling better. Now I know and I am quite appalled at the reason it exists.
Most users don't understand the FFT at all. As an example, window functions for the FFT. Keysight doesn't have a triangle (aka Bartlett) window option, essential to get good spectral resolution, on the MSOX-3104T, a $20K list instrument. And the FFT is so limited as to be almost completely useless.
The chief problem is that the FFT on most DSOs is *just* an FFT with almost no control over the parameters. To be useful it must be configured the same way as a spectrum analyzer. That *is* implemented in the SA app for the Instek MDO series, although it clearly has errors in the implementation. I'll sort those out eventually by analyzing a scope record and comparing my spectrum analysis to what they show. The FFT is just a tool, it is not a solution. And if you you don't have adequate control over the parameters it's not even useful. The main limitation of spectrum analysis on a DSO is the dynamic range. The other problems can be fixed by implementing it correctly.
The notion that you cannot process a sampled series for any reason other than an inability to get it out of the DSO is preposterous. It doesn't matter if it's aliased or what the spectrum is at all. It is sampled data and the mathematics were fully developed by Norbert Wiener in 1940 and published in 1949 as "The Extrapolation, Interpolation and smoothing of Stationary Time Series".
A skilled person can measure microsecond level multiplexer skew in 2 ms sample rate multichannel data with a maximum signal content of 60 Hz.
Have Fun!
Reg
-
Why does the spectrum look different when you change the time base? - I am not well versed in the math of FFT, but intuitively, it seems to me that whether you are looking at two cycles of a wave form, or 20 cycles, the spectral content should be exactly the same - after all, the wave didn't fundamentally change just because you look at more or less cycles of it. But it does make a big difference to the scope! To wit, looking at exactly the same 10MHz signal with two different timebase settings:
Agilent 54622D, timebase: 200ns/div, Scale: 10dBV/div Offset: -50.0dBV
[attachimg=2]
Timebase: 50ns/div
[attachimg=1]
-
Why does the spectrum look different when you change the time base?
The basic shape does not look so diffferent, though. The 2nd one was obvioulsy calculated from a lower number of points (-> lower frequency resolution), and was up-sampled then, in order to obtain the same number of frequency bins as in the first one.
You can basically achieve this by zero-padding the available samples to a larger number, and calculating the FFT from the zero-padded samples then.
EDIT: I presume that you did not select different window functions, did you?
-
Reading you all I come to the conclusion that people throw in more money in their scope buy in the expectation that it covers their lack of knowledge of the fundamentals. Some of you show such a deep knowledge that you make do with "almost" any model.
That's always been my take on it, biggest thing i look in a scope for is time alignment of whatever two or more signals i am looking for / debugging, once you have the raw data you can parse it to however many ends as you want, weather thats inside the same scope or another computer etc
-
That is the affect of the window length. The true spectrum is a series of spikes which is convolved with the window spectrum. So each spike is replaced by a scaled version of the Fourier transform of the window.
There's a pretty good description here of a wide variety of windows:
https://en.wikipedia.org/wiki/Window_function
While many of them are very familiar, there are a number I'd never heard of. Probably the most common window is a raised cosine at each end of a time series and then flat in between. A depressingly common DSP programming error is the failure to apply a taper to the ends of the series and to pad it with zeros to prevent wrap around effects. Another common error is to not remove and then replace a linear trend from the data when doing a forward and inverse transform so you can manipulate the data i the frequency domain.
If the series is not zero padded then the sidelobes of the window function wrap around to both ends. This can be very important if you are chopping a non-stationary time series into approximately stationary segments, applying an FFT, doing some operation such as spectrum equalization, back transforming and splicing the pieces back together.
In settings where all the data has been recorded before processing begins, it is very common to flip from time to frequency and back, often several times as an FFT, multiply and inverse FFT is much faster than a long convolution.
Have Fun!
Reg
-
The 2nd one was obvioulsy calculated from a lower number of points (-> lower frequency resolution),...
The 2nd trace does of course cover a shorter time interval (and thus also a lower number of original ADC samples) than the first one.
-
Why does the spectrum look different when you change the time base?
The basic shape does not look so diffferent, though. The 2nd one was obvioulsy calculated from a lower number of points (-> lower frequency resolution), and was up-sampled then, in order to obtain the same number of frequency bins as in the first one.
You can basically achieve this by zero-padding the available samples to a larger number, and calculating the FFT from the zero-padded samples then.
EDIT: I presume that you did not select different window functions, did you?
Everything was exactly the same in the two, apart from the time base. (Hanning window)
Why does the FFT spectrum become less detailed when you zoom in on the waveform by decreasing the sweep time? I would expect as you zoom out (increase sweep time) that you would LOSE resolution... yes, you have more cycles of the wave on the screen, but each cycle is made of less samples...
-
Most users don't understand the FFT at all. As an example, window functions for the FFT. Keysight doesn't have a triangle (aka Bartlett) window option, essential to get good spectral resolution, on the MSOX-3104T, a $20K list instrument. And the FFT is so limited as to be almost completely useless.
Sure it does have triangular window (User manual , top of page 93)... It's just you returned it before you had time to read manual end to end ^-^.
-
That is the affect of the window length. The true spectrum is a series of spikes which is convolved with the window spectrum. So each spike is replaced by a scaled version of the Fourier transform of the window.
I think I see... so basically, if the scope had used a shorter window length for the faster sweep speed, the FFT would have looked the same in the two cases?
The FFT buffer is always 2048 points and fixed. It looks like the window length is also fixed - there is no obvious way to change it, and it doesn't seem to change automatically with sweep speed.
I guess we are back to some of your earlier comments that scope FFTs don't really give the user enough control.
-
The sample rate sets the Nyquist frequency. The number of samples sets the RBW. The time length and shape of the window set the shape of an isolated peak. The window sidelobes control how adjacent peaks look in between.
I suggest creating a sine wave in Octave, applying various time domain windows, doing an FFT and looking at the spectra. Try sine waves close to DC or Nyquist with a rectangular window with and without zero padding to double the length.
Have Fun!
Reg
BTW What's the date on the MSOX3104T manual?
-
That is the affect of the window length. The true spectrum is a series of spikes which is convolved with the window spectrum. So each spike is replaced by a scaled version of the Fourier transform of the window.
I think I see... so basically, if the scope had used a shorter window length for the faster sweep speed, the FFT would have looked the same in the two cases?
The FFT buffer is always 2048 points and fixed. It looks like the window length is also fixed - there is no obvious way to change it, and it doesn't seem to change automatically with sweep speed.
Using a window which is significantly shorter than the FFT size does basically imply that you discard a large fraction of the available samples (at least if the window has a finite extent like Hanning). Why would you you do that? One good reason is, when fewer samples than the FFT size are available in the first place.
-
That is the affect of the window length. The true spectrum is a series of spikes which is convolved with the window spectrum. So each spike is replaced by a scaled version of the Fourier transform of the window.
I think I see... so basically, if the scope had used a shorter window length for the faster sweep speed, the FFT would have looked the same in the two cases?
The FFT buffer is always 2048 points and fixed. It looks like the window length is also fixed - there is no obvious way to change it, and it doesn't seem to change automatically with sweep speed.
Using a window which is significantly shorter than the FFT size does basically imply that you discard a large fraction of the available samples (at least if the window has a finite extent like Hanning). Why would you you do that? One good reason is, when fewer samples than the FFT size are available in the first place.
Alternatively I still consider the possibility that the FFT was not calculated from the raw ADC samples @100 MSPS, but from the (interpolated) screen buffer samples, so that the reported 1GSPS and 4GSPS do indeed reflect the sampling rate for the FFT calculation.
2048 points @4GSPS leads to frequency bins with a width of ~1.9MHz, while 2048 points @1GSPS gives you a 4x higher resolution of ~0.49MHz/bin. Note, when you select a span of 200 MHz, you don't increase this resolution - the span selection does just a zoom-in into the spectrum, interpolating between the lower-resolution bins.
[ Note, even with this different consideration it still bails down to fewer original ADC samples being included in the FFT, since the majority of the 1GSPS or 4GSPS samples are just interpolated. ]
Does actualy exist any (slower) timebase, where the reported FFT sampling rate and the ADC sampling rate are the same?
-
Fundamentally it's a botched implementation. However it is implemented, the programmer should have recognized that the frequencies are not in the GHz range. That such should appear under the Agilent name is truly sad.
Reg
-
Reg,
there is a lot of misunderstanding (by you) about how scopes are conceptually made, it seems.
The way FFT is usually done on scopes is that scope works in "scope mode", where sampling parameters are governed by scope settings (timebase, input atten etc etc).
To implement FFT in "spectrum mode", you would have to control sampling in a way that is optimal for FFT calculations for a SA use. That includes overlapped FFT, averaging etc, etc. All those things you correctly mention.. One scope that does that is Picoscope, which in Spectrum mode takes over control of sampling completely. But in this mode you cannot have time and frequency domain at the same time. GW instek MDO does something similar, and Tektronix MDOs in their way
All of that is not because Keysight (or Siglent or Rigol or LeCroy) are stupid and don't know how to do it.
It is actually deliberate design decision. Scope does what scope does, and then in math channel you get to massage buffer of sampled data with many functions, including FFT. It is by definition math applied as an afterthought.
FFT on those scopes is not making them mixed domain scopes.
Don't get me wrong, if somebody would make inexpensive mixed domain scope that would work properly in a manner you describe, I would buy it right away. There is a scope that does things as you describe, new Keysight MXR..And some LeCroys do it right too..
But those are very high end...
We have very basic misunderstanding here, it seems. I'm not saying your math is wrong, and you know about DSP more than many of us here. No argue there. And certainly there is nothing wrong with wishing things were better.
I'm sayin your expectations are unrealistics, for a lowest end scopes you can buy. Or even for mid range scopes today. We would need serious hardware upgrades to support independent simultaneous synchronous scope and realtime SA mode to standard scope architecture.
Best regards,
Sinisa
-
That is the affect of the window length. The true spectrum is a series of spikes which is convolved with the window spectrum. So each spike is replaced by a scaled version of the Fourier transform of the window.
I think I see... so basically, if the scope had used a shorter window length for the faster sweep speed, the FFT would have looked the same in the two cases?
The FFT buffer is always 2048 points and fixed. It looks like the window length is also fixed - there is no obvious way to change it, and it doesn't seem to change automatically with sweep speed.
Using a window which is significantly shorter than the FFT size does basically imply that you discard a large fraction of the available samples (at least if the window has a finite extent like Hanning). Why would you you do that? One good reason is, when fewer samples than the FFT size are available in the first place.
Alternatively I still consider the possibility that the FFT was not calculated from the raw ADC samples @100 MSPS, but from the (interpolated) screen buffer samples, so that the reported 1GSPS and 4GSPS do indeed reflect the sampling rate for the FFT calculation.
2048 points @4GSPS leads to frequency bins with a width of ~1.9MHz, while 2048 points @1GSPS gives you a 4x higher resolution of ~0.49MHz/bin. Note, when you select a span of 200 MHz, you don't increase this resolution - the span selection does just a zoom-in into the spectrum, interpolating between the lower-resolution bins.
[ Note, even with this different consideration it still bails down to fewer original ADC samples being included in the FFT, since the majority of the 1GSPS or 4GSPS samples are just interpolated. ]
Does actualy exist any (slower) timebase, where the reported FFT sampling rate and the ADC sampling rate are the same?
Yes, indeed there does. Source is 10MHz square wave from the 8012A pulse generator as before (pulse shape not exactly the same as the settings are analog on this device)
Timebase: 1uS/div
[attachimg=1]
Here, the Span actually matches what the scope is specced for (100MHz BW, 200Msamples/sec) and cannot be made any wider.
As an aside: I actually found a good use for the FFT function here - it makes it very easy to set up a clean pulse train from the pulse generator, tweaking the slopes and timings on the generator to minimize the even harmonics! Before, I was doing it by visual inspection of the waveform, but FFT is far easier and more accurate.
-
What I'm seeing in the comments above and my tests is pretty appalling.
The DSP in DSOs is a major fail even from the A list. It's not even good enough to get a passing grade as a DSP 101 homework exercise. The first chapter of any DSP text explains aliasing and why you cannot decimate data by throwing away samples.
I'm fairly sure the people who design these things know the math.
I suspect the problem is that there's not enough processing power to do a proper decimation when every pixel on screen has dozens of samples in it.
-
What I'm seeing in the comments above and my tests is pretty appalling.
The DSP in DSOs is a major fail even from the A list. It's not even good enough to get a passing grade as a DSP 101 homework exercise. The first chapter of any DSP text explains aliasing and why you cannot decimate data by throwing away samples.
I'm fairly sure the people who design these things know the math.
Not just that, they also know what is useful or not. You can make a DSO show all kinds of fun signals if you push it but oddly enough you never run into these during normal use.
-
An FFT example from my low-cost USB scope, when interpolation is in effect.
The FFT is clearly calculated from the interpolated (upsampled) points.
Signal is 10 MHz square wave (AWG is not that fast), 1GS/s ADC sampling rate,
timebase 20ns/div (-> so interpolation needed for display).
Reported FFT sampling rate is also higher than ADC sampling rate.
Specified FFT size is 1024.
If I select "step" interpolation, then the FFT clearly shows the 1GHz interpolation steps (as they show up in the time domain display), their harmonics at N*1GHz, as well as their IM with the baseband:
[attach=1]
Sin(x)/x interpolation still shows an (unexpected) peak @1GHz, but its harmonics are almost gone. Obviously the scope's sin(x)/x interpolation filter is not a perfect boxcar (in frequency domain).
[attach=2]
Linear interpolation gives the cleanest FFT result for this signal:
[attach=3]
Isn't that great? The scope's FFT assists assessing the scope's interpolation algorithms ;D
-
Sinisa,
I've started working on code to do the SA function properly. I'm still doing the requirements analysis and don't expect to start writing code for several weeks. There are many nuances to getting this right. I have a long standing rule: I never write code until I've run out of excuses for not writing it.
This morning I was examining heterodyning a narrow span and doing the FFT on the lower sideband. In an FPGA heterodyning is basically a single multiplier and and a sine table in block RAM. I'm sure there is some stuff I've overlooked, but that's not a lot of resources and it allows downsampling by summing samples to increase the dynamic range. So probably 2 DSP and 2 RAM blocks in a Zynq. If you want to look at a filter response over a 500 kHz BW with 10 Hz RBW it's a 100 kSa FFT on data sampled at 1 Msa/s. If the ADC is sampling at 1 GSa/s there is 60 dB increase in dynamic range over the 42 dB range of the 8 bit data coming out of the ADC. So now we have a 102 dB dynamic range. The screen can't be updated faster than 60 fps or so. As a consequence there is loads of time to do the processing.
This is not particularly difficult and does not require changing the data acquisition. The UI needs to be sensible, frequency & span, RBW, VBW, etc. Not the "center frequency, Hz/div and very limited selection of sample length imposed by current DSOs which provide no meaningful control over the FFT.
It just has to be done properly. Relative to the code for a 10 TB input imaging job that takes 20,000 cores 7-10 days to complete it's very easy. A code for that is 1-2 years work full time by a top rate scientist/programmer. A first rate SA app for a DSO is a few weeks work at most to do the numerical stuff. I don't do UIs.
Have Fun!
Reg
-
This is not particularly difficult and does not require changing the data acquisition. The UI needs to be sensible, frequency & span, RBW, VBW, etc. Not the "center frequency, Hz/div and very limited selection of sample length imposed by current DSOs which provide no meaningful control over the FFT.
Already you have control over RBW/VBW with the timebase setting used for FFT.
Span is set via Hz/div and CF.
It's just a different way at arriving at the same result as a fully fledged SA.
When I've finished my pulser that hopefully will get into the few 100ps range I'll show some examples.
-
An FFT example from my low-cost USB scope, when interpolation is in effect.
The FFT is clearly calculated from the interpolated (upsampled) points.
Signal is 10 MHz square wave (AWG is not that fast), 1GS/s ADC sampling rate,
timebase 20ns/div (-> so interpolation needed for display).
Reported FFT sampling rate is also higher than ADC sampling rate.
Specified FFT size is 1024.
If I select "step" interpolation, then the FFT clearly shows the 1GHz interpolation steps (as they show up in the time domain display), their harmonics at N*1GHz, as well as their IM with the baseband:
(Attachment Link)
Sin(x)/x interpolation still shows an (unexpected) peak @1GHz, but its harmonics are almost gone. Obviously the scope's sin(x)/x interpolation filter is not a perfect boxcar (in frequency domain).
(Attachment Link)
Linear interpolation gives the cleanest FFT result for this signal:
(Attachment Link)
Isn't that great? The scope's FFT assists assessing the scope's interpolation algorithms ;D
Why do you have the 20 MHz BW filter turned on for this?
A 1024 point FFT with a 1 ns sample interval has a 500 MHz Nyquist and 977 kHz RBW. With a 1024 pixel wide screen *every* sample can be shown. There is no interpolation needed.
Reg
-
Already you have control over RBW/VBW with the timebase setting used for FFT.
Span is set via Hz/div and CF.
It's just a different way at arriving at the same result as a fully fledged SA.
When I've finished my pulser that hopefully will get into the few 100ps range I'll show some examples.
I'm sorry, but that is pure BS. To provide the functionality of a proper SA you need much more control over the FFT. Setting the timebase determines the time display. It should not have any influence on a spectral display. If it does, as is commonly the case, it is *wrong*!.
This is why we see stupid stuff like 50 GSa/S on the DS1202Z-E at 2 ns/div. Programming by idiots.
Reg
-
Why do you have the 20 MHz BW filter turned on for this?
Sorry, the AWG isn't faster :(
A 1024 point FFT with a 1 ns sample interval has a 500 MHz Nyquist and 977 kHz RBW. With a 1024 pixel wide screen *every* sample can be shown. There is no interpolation needed.
Timebase is 20ns/div => ~1000 pixel = 200 ns, plus some headroom for zoom-in after stopping the capture. In the time domain display of the first image (-> step interpolation) you can see that the 1ns samples from the ADC are more than one pixel wide.
I'm pretty confident that the reported 12.5Gs/s FFT sampling rate is indeed the (upsampled) display buffer sampling rate for the selected timebase. Particularly in the first image, the spectrum beyond 1GHz matches the expected spectrum of the step-interpolation as shown in the time-domain trace.
-
[...] This is why we see stupid stuff like 50 GSa/S on the DS1202Z-E at 2 ns/div. [...]
Like having a speedometer in a Trabant that goes to 100,000 miles per hour? :D
-
[...] This is why we see stupid stuff like 50 GSa/S on the DS1202Z-E at 2 ns/div. [...]
Like having a speedometer in a Trabant that goes to 100,000 miles per hour? :D
That would be about right. If one accounts for the Earth's orbital velocity around the Sun.
-
Why do you have the 20 MHz BW filter turned on for this?
Sorry, the AWG isn't faster :(
A 1024 point FFT with a 1 ns sample interval has a 500 MHz Nyquist and 977 kHz RBW. With a 1024 pixel wide screen *every* sample can be shown. There is no interpolation needed.
Timebase is 20ns/div => ~1000 pixel = 200 ns, plus some headroom for zoom-in after stopping the capture. In the time domain display of the first image (-> step interpolation) you can see that the 1ns samples from the ADC are more than one pixel wide.
I'm pretty confident that the reported 12.5Gs/s FFT sampling rate is indeed the (upsampled) display buffer sampling rate for the selected timebase. Particularly in the first image, the spectrum beyond 1GHz matches the expected spectrum of the step-interpolation as shown in the time-domain trace.
Please put a 50 ohm thru termination on it and repost. A BNC tee with a 50 ohm terminator is "good enough". I'll address the rest later. Interpolation does *not* imply greater BW. Common knowledge if you resample via Fourier transform as I usually do.
Reg
-
A side note.
If I'm not wrong at least once or twice a day I'm clearly not trying hard enough. There is *no* shame in being wrong. The only shame is in failing to admit the error.
Have Fun!
Reg
-
Sinisa,
I've started working on code to do the SA function properly. I'm still doing the requirements analysis and don't expect to start writing code for several weeks. There are many nuances to getting this right. I have a long standing rule: I never write code until I've run out of excuses for not writing it.
This morning I was examining heterodyning a narrow span and doing the FFT on the lower sideband. In an FPGA heterodyning is basically a single multiplier and and a sine table in block RAM. I'm sure there is some stuff I've overlooked, but that's not a lot of resources and it allows downsampling by summing samples to increase the dynamic range. So probably 2 DSP and 2 RAM blocks in a Zynq. If you want to look at a filter response over a 500 kHz BW with 10 Hz RBW it's a 100 kSa FFT on data sampled at 1 Msa/s. If the ADC is sampling at 1 GSa/s there is 60 dB increase in dynamic range over the 42 dB range of the 8 bit data coming out of the ADC. So now we have a 102 dB dynamic range. The screen can't be updated faster than 60 fps or so. As a consequence there is loads of time to do the processing.
This is not particularly difficult and does not require changing the data acquisition. The UI needs to be sensible, frequency & span, RBW, VBW, etc. Not the "center frequency, Hz/div and very limited selection of sample length imposed by current DSOs which provide no meaningful control over the FFT.
It just has to be done properly. Relative to the code for a 10 TB input imaging job that takes 20,000 cores 7-10 days to complete it's very easy. A code for that is 1-2 years work full time by a top rate scientist/programmer. A first rate SA app for a DSO is a few weeks work at most to do the numerical stuff. I don't do UIs.
Have Fun!
Reg
Reg,
thanks for discussion.
FFT crunching alone is not a problem. Feeding it, is, and also showing results.
A/D feeding display engine is 5GB/s datapump. Scope deals with it in realtime. Sure there are pauses in there, but even with those, a lot of data gets moved around.
And also, there is a problem of display. That would need combining FFT results, math results, decode results, measurement results.. Each by itself not a problem, but combining and orchestrating them gets complicated too.
And FFT can use data from normal acquisitions. Sometimes. This is how it works now. Problem is that sample memory length/timebase/sampling ratios are governed by scope settings for time domain. So somewhere in the middle of timebase range, where we are still operating at full sample speed, but already have long memory depth, we can use scope's capture. In that case we have data that can be decimated, downsampled etc. And there is enough of it to have your bins nicely filled to length you need.
But what to do when I'm looking at signal at slow timebase (that scope is sampling at 100 Mpoints/s) , and would like to use FFT to see if some spikes I see in there are 868 MHz bursts from telemetry receiver... That would need me to have scope sampling at 2GS/s or more at the same time it samples for my time domain display at 100 MS/s.
Answer is to either have separate A/D converter (Tek MDO series), have A/D always sample at full speed and decimate and distribute data to separate and parallel datapaths, or have scope always have time domain priority (and have FFT have whatever data is available, useful or not), or have scope have FFT priority, and start changing sampling to always cater to FFT needs and not show time domain data if not available.
First two options are expensive. As for second two options, since we are not talking about SA with auxiliary time domain display, but about scopes with auxiliary FFT (frequency domain display), those scopes prioritize sampling to cater to scopes time domain needs primarily and FFT gets what it does. Simple as that.
Do i think that option with having paralel datapaths for scope and FFT is possible? Absolutely, in Zynq definitely. But it is not a classic architecture, and would need to be developed.
Funny thing, Rigol's new chipset is used in both their new scopes and SAs, and it possibly could be used to make mixed signal scope that would be able to do what are you talking here...
Take care
-
Not just that, they also know what is useful or not. You can make a DSO show all kinds of fun signals if you push it but oddly enough you never run into these during normal use.
Yep.
Weird, that...
-
Timebase is 20ns/div => ~1000 pixel = 200 ns, plus some headroom for zoom-in after stopping the capture. In the time domain display of the first image (-> step interpolation) you can see that the 1ns samples from the ADC are more than one pixel wide.
I'm pretty confident that the reported 12.5Gs/s FFT sampling rate is indeed the (upsampled) display buffer sampling rate for the selected timebase. Particularly in the first image, the spectrum beyond 1GHz matches the expected spectrum of the step-interpolation as shown in the time-domain trace.
Please put a 50 ohm thru termination on it and repost. A BNC tee with a 50 ohm terminator is "good enough". I'll address the rest later. Interpolation does *not* imply greater BW. Common knowledge if you resample via Fourier transform as I usually do.
A 50 Ohm feed-through was already in place (but the AWG signal does not have steep edges anyway, so it does not make much difference).
I fully agree, that perfect sinc interpolation is not supposed to increase BW by "producing" new frequencies. And if the sinc up-sampling is done via zero-insertion + boxcar filtering via FFT, then it becomes clear that it cannot, since the freqeuncy band beyond the the original Nyquist is explicitly cleared before ifft.
Other interpolators (e.g. nearest neibhgor, polynomial, etc.) can well introduce "new" frequencies, though, since they approximate just time domain w/o caring about Fourier stuff and frequency domain. A sinc interpolation kernel, on the other hand, is in fact an implicit boxcar filter in the frequency domain. Convolution with a truncated sinc does no longer give a perfect boxcar frequency response either, so zero-insertion + filtering with only a truncated sinc does not necessarily eliminate all frequences which where introduced by the zero insertion.
Btw, with "display buffer" I do not mean 1:1 the screen pixels, but intermediate data prepared for screen display. Mapping to the actual screen pixels still requires an additional resampling step. Eventually there seem to be 3 steps
1) Sampled data (at ADC sampling rate)
2) Interpolated/upsampled data (at a higher sampling rate, which depends on the selected timebase)
3) resampling to actual screen pixels (yet a different sampling rate, depending on current screen-window size, zoom)
FFT (and other math functions) seem to operate on (2), at least on my scope.
-
So, are there any "rules of thumb" for getting the best FFT performance out of a scope, given that it is a secondary function to the time domain?
Is the best performance going to be when the FFT sample rate matches the scope's sample rate?
-
The way a DSO *should* work when showing a time and/or frequency display:
ADC *always* samples at full speed. There is one ADC data stream which is fed to 2 DSP pipelines.
Data for time domain display is downsampled to sample rate appropriate to timebase setting using a folded LP filter
Data for frequency domain display is heterodyned and downsampled appropriate to SA settings using a multiplier and folded LP filter
Time domain display interpolation is by means of a minimum phase interpolator which combines the passband of the anti-alias filter and the downsampling LP filter on the time domain data stream. The coefficients of that filter are optimized to minimize errors. A similar process is applied to interpolating the frequency domain data.
None of the above requires a lot of resources. If it does, it's being done wrong. I suggest reading:
VLSI Digital Signal Processing Systems: Design and Implementation
K.K. Parhi
Wiley 1999
for the details on how to implement the above in hardware with minimum resources. Because of the screen update rate limitation the FFT can be done in the NEON cores of a Zynq.
Claims that you need to sample the input data with a 2nd ADC are the sort of reason that current DSOs are so bad. Not doing it in the manner outlined above is either a failure to perform the correct mathematical operations or a failure to implement them efficiently.
The above is not speculative, It is the product of a considerable amount of work over the last 2 years. I already knew DSP on a general purpose computer. I did not know the details of how to implement it on an FPGA. I shall explain why I would do such in a few months.
Have Fun!
Reg
-
So, are there any "rules of thumb" for getting the best FFT performance out of a scope, given that it is a secondary function to the time domain?
Is the best performance going to be when the FFT sample rate matches the scope's sample rate?
Best I can offer is try all the knobs and make sure you didn't miss part of the FFT menu structure. I *always* find it difficult to get what I want using anything other than the Instek SA app.
And use a test signal for which you know what the spectrum should look like to set it up. I find it extremely annoying and frustrating.
Reg
-
Alternatively I still consider the possibility that the FFT was not calculated from the raw ADC samples @100 MSPS, but from the (interpolated) screen buffer samples, so that the reported 1GSPS and 4GSPS do indeed reflect the sampling rate for the FFT calculation.
2048 points @4GSPS leads to frequency bins with a width of ~1.9MHz, while 2048 points @1GSPS gives you a 4x higher resolution of ~0.49MHz/bin. Note, when you select a span of 200 MHz, you don't increase this resolution - the span selection does just a zoom-in into the spectrum, interpolating between the lower-resolution bins.
[ Note, even with this different consideration it still bails down to fewer original ADC samples being included in the FFT, since the majority of the 1GSPS or 4GSPS samples are just interpolated. ]
Does actualy exist any (slower) timebase, where the reported FFT sampling rate and the ADC sampling rate are the same?
The screen display is just the screen display. That sampling has *nothing* to do with the frequency range of the data being displayed. The DS1202Z-E set at 2 ns/div reports 2.5 GHz/div 50 GSa/s. That's complete nonsense. That's a crap programming job done by someone who didn't bother to look at the results and recognize it was pure nonsense.
Reg
-
I did not know the details of how to implement it on an FPGA.
I think this is the problem. :)
FPGAs have a limited number of gates and multiply-accumulate units need a lot of them. It doesn't matter how simple your algorithm is on paper, if there's not enough gates then it can't be done.
-
Alternatively I still consider the possibility that the FFT was not calculated from the raw ADC samples @100 MSPS, but from the (interpolated) screen buffer samples, so that the reported 1GSPS and 4GSPS do indeed reflect the sampling rate for the FFT calculation.
2048 points @4GSPS leads to frequency bins with a width of ~1.9MHz, while 2048 points @1GSPS gives you a 4x higher resolution of ~0.49MHz/bin. Note, when you select a span of 200 MHz, you don't increase this resolution - the span selection does just a zoom-in into the spectrum, interpolating between the lower-resolution bins.
[ Note, even with this different consideration it still bails down to fewer original ADC samples being included in the FFT, since the majority of the 1GSPS or 4GSPS samples are just interpolated. ]
Does actualy exist any (slower) timebase, where the reported FFT sampling rate and the ADC sampling rate are the same?
The screen display is just the screen display. That sampling has *nothing* to do with the frequency range of the data being displayed. The DS1202Z-E set at 2 ns/div reports 2.5 GHz/div 50 GSa/s. That's complete nonsense. That's a crap programming job done by someone who didn't bother to look at the results and recognize it was pure nonsense.
Reg
It looks like the "complete nonsense" approach was started by the A brands back in the day - and now many/most of the B brands are doing the same thing, trying to look just as good as an A brand, as @rf-loop pointed out earlier in this thread.
Dismissing crazy numbers like a 100GHz FFT span on a 100MHz scope is not "beginner level", as we can see in this thread so far!
If you are going to make a simplified display of complex information to non-expert users (which is really what FFT on an entry level scope is), one of the first rules of GUI design is - never allow the user to enter settings that ends up displaying nonsense results. That is something that could have been done without even altering the underlying hardware or code.
-
Isn't that great? The scope's FFT assists assessing the scope's interpolation algorithms ;D
Maybe they started implementing it to do something like that and, then, decided to make it public (without any further validations). ::)
-
The screen display is just the screen display. That sampling has *nothing* to do with the frequency range of the data being displayed. The DS1202Z-E set at 2 ns/div reports 2.5 GHz/div 50 GSa/s. That's complete nonsense. That's a crap programming job done by someone who didn't bother to look at the results and recognize it was pure nonsense.
Seems to me that that the scope has a notion of a "desired number of points/div" (say 100, which happens to result in a desired sampling rate of 50 Gsa/s then, at 2ns/div), and the captured data are generally converted to this desired rate. If the desired sampling rate is higher than the max. ADC sampling rate, then the data are up-samped/interpolated to the desired rate before further processing, ensuring a particular minimum number of points/div.
If a buffer containing 50 GSa/s samples was passed to the FFT engine as input, why should it not report 50 GSa/s? Likely the FFT engine does not even know about the origin of the data (i.e. real vs. up-sampled), and it also does not care.
I do not deny that up-sampling the data before calulating the FFT is not really helpful. On the contrary, it reduces the frequency resolution if the number of FFT points is fixed.
For other math-functions like e.g. RMS, it can indeed make sense, though, to calculate them from up-sampled/interpolated data instead of using just a few sparse original ADC samples per period for the calculation, which won't be accurate then. Obvioulsy, FFT is classified as "math function", too, so I guess it is fed with the same data as other math functions.
-
The DSP in DSOs is a major fail even from the A list. It's not even good enough to get a passing grade as a DSP 101 homework exercise. The first chapter of any DSP text explains aliasing and why you cannot decimate data by throwing away samples. You *have* to low pass filter the data. That's not hard to do and doesn't require a lot of resources, so I'm agog that DSOs are decimating data by discarding samples. The results of my tests with a 5 ns pulse at 1 s intervals showed that "peak detection" is necessary to offset the absolute bodge of downsampling by decimation.
That illustrates the difference between theory and reality. There are good reasons that digital storage oscilloscopes work the way they do. As pointed out above, if you want a digitizer, than buy a digitizer, or an older DSO from LeCroy who initially based their DSOs on digitizers.
On the practical side, for a majority of time DSOs simply could not do more than the simplest processing during decimation short of a heroic effort and price. Indeed, they could not even store the undecimated acquisition because memory was not long and fast enough. The best which could be done is to discard samples, or the very simplest processing like boxcar averaging (high resolution mode) or peak detection. See below about implementing a filter based on decimation ratio or sweep speed.
Later when real time processing became feasible with custom logic, the DPO (digital phosphor oscilloscope) was invented. It has no need of decimation because it produces a histogram of the input in real time and every sample contributes to the acquisition record at any sweep speed.
The way a DSO *should* work when showing a time and/or frequency display:
ADC *always* samples at full speed. There is one ADC data stream which is fed to 2 DSP pipelines.
Data for time domain display is downsampled to sample rate appropriate to timebase setting using a folded LP filter
Time domain display interpolation is by means of a minimum phase interpolator which combines the passband of the anti-alias filter and the downsampling LP filter on the time domain data stream. The coefficients of that filter are optimized to minimize errors. A similar process is applied to interpolating the frequency domain data.
The problem with what you describe is that it results in a bandwidth which depends on decimated sample rate which depends on sweep speed. That is not the case for analog oscilloscopes where transition time does not depend on sweep speed.
It is also why I keep repeating that DSOs do not implement anti-aliasing filters, at least as commonly understood. If they did, then their bandwidth would vary with the time/div setting. You certainly could implement such now but why? The display will be no different except for two factors:
1. Signal envelopes will be wrong as higher frequencies are attenuated. Of course with aliasing it may not be apparent that a signal even has an envelope like in the example photograph that I posted which shows a fine envelope despite potential aliasing because peak detection was used.
2. The histogram for the signal will be corrupted. All of the characteristics and measurements which depend on the histogram will now vary with sweep speed; won't that be fun! Of course those who are used to Rigol's automatic measurements will not notice a difference because Rigol already does this by making measurements on the display record which has a corrupted histogram from the processing required to create it. This suggests that no uncontrolled processing which alters the histogram should be performed between acquisition and measurement.
To give a concrete example of the above, the DSOs I use can measure RMS and peak-to-peak noise easily and accurately within the limits of their fixed input bandwidth. But if decimation was done as you describe, then noise measurements would vary with sweep speed and be essentially useless. Changing the sweep speed should not alter the input bandwidth.
Another example is transition time would vary with sweep speed. Now you might think from using modern DSOs that this would not be a big issue but how did old DSOs handle it? They could not report the correct transition time if too few samples were taken at an edge and the designers knew this! Instead, they return the questionable measurement and include a warning that the decimated sample rate is insufficient. Many modern DSOs just lie. Those old DSOs also adjusted their returned number of significant digits to account for measurement precision.
-
[...] Changing the sweep speed should not alter the input bandwidth. [...]
Not sure I follow... (happens a lot, sadly)
If the ADC is running at full tilt all the time, why would the bandwidth alter with the sweep speed? Presumably it would have a filter in front of it that is tuned to its sample rate (Nyquist) and that never changes?
Are you saying it would be a bad idea for the scope to apply another, second filter that depends on the maximum frequency that the display is able to show at the selected sweep speed?
-
[...] Changing the sweep speed should not alter the input bandwidth. [...]
Not sure I follow... (happens a lot, sadly)
If the ADC is running at full tilt all the time, why would the bandwidth alter with the sweep speed? Presumably it would have a filter in front of it that is tuned to its sample rate (Nyquist) and that never changes?
Are you saying it would be a bad idea for the scope to apply another, second filter that depends on the maximum frequency that the display is able to show at the selected sweep speed?
I think David Hess is referring to the situation where the samplerate has to drop because the memory is too short to capture a full screen at the maximum samplerate.
I also think David Hess has an interesting point about a DSO not changing it's bandwidth depending on the samplerate because it would make peak-detect not working. OTOH I wonder what good peak-detect does on the FFT operation anyway; do you want to have both enabled at the same time? Also doing FFT on a seperate processing path as rhb suggested would make it impossible to change the part of a signal FFT is performed on (scroll left/right). I think the bottom line is that FFT is a useful tool on a DSO but there are trade-offs and every manufacturer implements it differently (decimated data / non decimated data, full record length / partial record length, etc).
-
[...] Changing the sweep speed should not alter the input bandwidth. [...]
Not sure I follow... (happens a lot, sadly)
If the ADC is running at full tilt all the time, why would the bandwidth alter with the sweep speed? Presumably it would have a filter in front of it that is tuned to its sample rate (Nyquist) and that never changes?
Are you saying it would be a bad idea for the scope to apply another, second filter that depends on the maximum frequency that the display is able to show at the selected sweep speed?
Let's take simple signal Dave likes so much: AM modulated carrier. You take say, 20 MHz carrier, AM modulate it with 100 Hz.
If you were to do decimation by a folded LP filter (as suggested), you would end up with carrier filtered out... simple as that. Scope screen wouldn't show what it should. That is also a reason to have Peak detect mode, so we can detect pulses that are faster than what sampling would show on slower sampling rates. That is not proper sampling for DSP. That is proper sampling to show signal shape on screen. That is what scopes are used for. To look at signal shape on screen. All other math and measurements are just utilities thrown in for better value. Including FFT.
I wanted scope that has FFT done better than other scopes. So I bought two Picoscopes. One is 16Bit low noise 5MHz bandwidth. They are also great for decoding. My other scope is MSOX 3104T. I use it as a scope.
And it's FFT works great, you can set it as on SA, with center frequency and span, or start - stop frequency. You can set averaging, peak, min max mode, search for peaks, measure it with cursors etc etc. But timebase still governs sampling, so sometimes you need to set timebase differently so you get data needed to do FFT at settings you want to set. It works great, actually, except it is limited to only 64K points. Which means you will have limited resolution bandwidth... When you compare it with Picos 2 Mpoints (for 1 M bins) you see the difference.
There is also one thing: Keysight claims they are using Chirp-Z Transform (CZT) in 3000T (and guess more of the lines) instead of vanilla FFT, because it is more flexible..
-
My other scope is MSOX 3104T. I use it as a scope.
And it's FFT works great, you can set it as on SA, with center frequency and span, or start - stop frequency. You can set averaging, peak, min max mode, search for peaks, measure it with cursors etc etc. But timebase still governs sampling, so sometimes you need to set timebase differently so you get data needed to do FFT at settings you want to set. It works great, actually, except it is limited to only 64K points. Which means you will have limited resolution bandwidth... When you compare it with Picos 2 Mpoints (for 1 M bins) you see the difference.
This ^, something Reg called out as BS. ::)
-
My other scope is MSOX 3104T. I use it as a scope.
And it's FFT works great, you can set it as on SA, with center frequency and span, or start - stop frequency. You can set averaging, peak, min max mode, search for peaks, measure it with cursors etc etc. But timebase still governs sampling, so sometimes you need to set timebase differently so you get data needed to do FFT at settings you want to set. It works great, actually, except it is limited to only 64K points. Which means you will have limited resolution bandwidth... When you compare it with Picos 2 Mpoints (for 1 M bins) you see the difference.
This ^, something Reg called out as BS. ::)
Exactly, because he thinks doing realtime DSP on 10 GB/s worth of data is trivial, together with sophisticated triggering, zone trigger, digital phosphor emulation, decoding, segmenting, history, search, measurements, math, parametric filtering. All of those things are not problem by itself. Doing all that at the same time, not so trivial.
I already said, there are high end scopes from Keysight, LeCroy, R&S that are doing it better. But those are out of reach of us mortals.
Also there is some stuff scopes don't do usually. His idea of decimation by LP filtering to gain more resolution at slower time bases sounds nice.. But wait, isn't that Hires mode (or 10 Bit mode on SDS200X+)?? I guess scopes ARE doing it. Except you don't want it all the time. Like David said well, slow timebase doesn't mean low bandwidth.
He does have nice ideas, and he is right about math part. And I would like if Siglent and Rigol would make lower price MDO scope with correlated time/frequency domain. With frequency mask trigger please.. But how realistic is that, really ??
-
My first 2 scopes were analog and those just start drawing from the left side of the screen after a trigger.
Then I got into digital scopes, and they all default to triggering in the center of the screen, which I find quite annoying.
For repetitive signals it's moot of course, but for all one-off signals the left half of the screen just has no useful information at all, because the "interesting" stuff happens after the trigger.
My best analog scope had a delay line (10+m of coax wound around it's tube) and with that I could see just a little bit before the trigger, which is perfect.
I would prefer my digital scope to default the trigger position to one division from the left side of the screen, so you can see a little bit of "before" the trigger, and still use almost the whole screen for the rest of the signal.
You can of course manually move the trigger location horizontally, but that messes up with zooming in and out on a digital scope.
-
He does have nice ideas, and he is right about math part. And I would like if Siglent and Rigol would make lower price MDO scope with correlated time/frequency domain. With frequency mask trigger please.. But how realistic is that, really ??
Technically it is possible but their oscilloscopes would cost the same as buying one from an A-brand.
-
My other scope is MSOX 3104T. I use it as a scope.
And it's FFT works great, you can set it as on SA, with center frequency and span, or start - stop frequency. You can set averaging, peak, min max mode, search for peaks, measure it with cursors etc etc. But timebase still governs sampling, so sometimes you need to set timebase differently so you get data needed to do FFT at settings you want to set. It works great, actually, except it is limited to only 64K points. Which means you will have limited resolution bandwidth... When you compare it with Picos 2 Mpoints (for 1 M bins) you see the difference.
This ^, something Reg called out as BS. ::)
Exactly, because he thinks doing realtime DSP on 10 GB/s worth of data is trivial, together with sophisticated triggering, zone trigger, digital phosphor emulation, decoding, segmenting, history, search, measurements, math, parametric filtering. All of those things are not problem by itself. Doing all that at the same time, not so trivial.
I already said, there are high end scopes from Keysight, LeCroy, R&S that are doing it better. But those are out of reach of us mortals.
Also there is some stuff scopes don't do usually. His idea of decimation by LP filtering to gain more resolution at slower time bases sounds nice.. But wait, isn't that Hires mode (or 10 Bit mode on SDS200X+)?? I guess scopes ARE doing it. Except you don't want it all the time. Like David said well, slow timebase doesn't mean low bandwidth.
:-+
He does have nice ideas, and he is right about math part.
I highly respect Reg and the amazing seismic exploratory work he has done and analysis of TB's of data acquired over weeks however scopes work in realtime or thereabouts and the data stream is managed so to not have results that are slow as pouring molasses to arrive at the display.
Scopes don't have the luxury of providing information sometime after all calculations are completed.
-
He does have nice ideas, and he is right about math part. And I would like if Siglent and Rigol would make lower price MDO scope with correlated time/frequency domain. With frequency mask trigger please.. But how realistic is that, really ??
Technically it is possible but their oscilloscopes would cost the same as buying one from an A-brand.
Exactly.
-
My first 2 scopes were analog and those just start drawing from the left side of the screen after a trigger.
Then I got into digital scopes, and they all default to triggering in the center of the screen, which I find quite annoying.
For repetitive signals it's moot of course, but for all one-off signals the left half of the screen just has no useful information at all, because the "interesting" stuff happens after the trigger.
My best analog scope had a delay line (10+m of coax wound around it's tube) and with that I could see just a little bit before the trigger, which is perfect.
I would prefer my digital scope to default the trigger position to one division from the left side of the screen, so you can see a little bit of "before" the trigger, and still use almost the whole screen for the rest of the signal.
You can of course manually move the trigger location horizontally, but that messes up with zooming in and out on a digital scope.
The 56422D gives you the choice of where to put the trigger location: right, center, left. I would have thought all digital scopes offers that choice?
-
[...] Changing the sweep speed should not alter the input bandwidth. [...]
Not sure I follow... (happens a lot, sadly)
If the ADC is running at full tilt all the time, why would the bandwidth alter with the sweep speed? Presumably it would have a filter in front of it that is tuned to its sample rate (Nyquist) and that never changes?
Are you saying it would be a bad idea for the scope to apply another, second filter that depends on the maximum frequency that the display is able to show at the selected sweep speed?
Let's take simple signal Dave likes so much: AM modulated carrier. You take say, 20 MHz carrier, AM modulate it with 100 Hz.
If you were to do decimation by a folded LP filter (as suggested), you would end up with carrier filtered out... simple as that. Scope screen wouldn't show what it should. That is also a reason to have Peak detect mode, so we can detect pulses that are faster than what sampling would show on slower sampling rates. That is not proper sampling for DSP. That is proper sampling to show signal shape on screen. That is what scopes are used for. To look at signal shape on screen. All other math and measurements are just utilities thrown in for better value. Including FFT.
I wanted scope that has FFT done better than other scopes. So I bought two Picoscopes. One is 16Bit low noise 5MHz bandwidth. They are also great for decoding. My other scope is MSOX 3104T. I use it as a scope.
And it's FFT works great, you can set it as on SA, with center frequency and span, or start - stop frequency. You can set averaging, peak, min max mode, search for peaks, measure it with cursors etc etc. But timebase still governs sampling, so sometimes you need to set timebase differently so you get data needed to do FFT at settings you want to set. It works great, actually, except it is limited to only 64K points. Which means you will have limited resolution bandwidth... When you compare it with Picos 2 Mpoints (for 1 M bins) you see the difference.
There is also one thing: Keysight claims they are using Chirp-Z Transform (CZT) in 3000T (and guess more of the lines) instead of vanilla FFT, because it is more flexible..
I think I see what you are saying. On a traditional analog scope, if you view a high frequency signal at a slow sweep speed, the line just get fatter... but you know there is a signal there, even if you don't know exactly what it is and even though it is not accurately represented (apart from amplitude... so analog scopes do peak detect naturally, in reality!)
So I agree that filtering the signal depending on sweep speed is not desirable in all circumstances. It could be a cool option, though, to eliminate aliasing in some circumstances?
-
Yes you got it. That is exactly the point.
-
A few points, there's too much pure silliness to bother about all of it.
The discrete Fourier transform of a band limited impulse is a constant value. If I remove half of the values from the middle of the transform, so instead of 10,000 at 1 ns I have 5000 and I back transform I will have a band limited impulse at a 2 ns sample rate. It doesn't matter what the new sample rate is. It will always be a spike. I have used this method to do band limited resampling for 35 years including converting sampling between fractional meters and feet and made the results *exact* to the limits of the input BW with *no* phase errors. That last bit was a *very* painful 2 weeks of having my nose rubbed in the definition of the discrete Fourier transform.
If a signal is an impulse at the fastest sampling rate it is an impulse at *all* slower sampling rates and it is *always* there.
If you properly downsample with low pass filtering rather than decimating by throwing away samples you will *always* have a band limited spike. It will *not* disappear. However, it *will* make "peak detection" moot.
I've been counting DSP and logic blocks in FPGAs in the context of a DSO for a couple of years now. If the clock rate of the FPGA is not fast enough to process the data stream in a serial pipeline, one uses a parallel interleaved filter, c.f. Parhi for the details. In a 1 GHz clocked FPGA a 10 GSa/s data stream will need a 10x interleaved parallel filter. Dealing with Vivado/Vitis is more difficult than the actual topology for the implementation.
As it is clear that few understand what a "folded" filter is, it's a single adder-multiplier which accepts data at one clock rate and outputs it at a lower clock rate after performing multiple filter tap operations. It is the hardware equivalent of a loop.
In the case of an analog scope, on a high BW scope at slow sweep speeds, a fast impulse will likely not be visible at all unless the intensity is set very high.
I made a long list of tests to perform today. I'll start doing some of those over the next few days. Jitter is probably next as I'm curious about the apparent sample dithering.
Have Fun!
Reg
-
So, I sat down in front of the scope and tried the 100Hz modulation of a 20MHz signal - nothing like a real example, right?
It seems to me that the digital scope behaves pretty much like an analog scope in this case. The slow sweep speed does not seem to detract from the ability to display an outline of the 20MHz signal (even without Peak Detection enabled).
Obviously, filtering out the 20MHz would destroy the information as displayed here.
(note: normal mode, not peak detect)
[attachimg=1]
-
A few more points:
The purpose of the thread is to document what is wrong with DSOs on the market.
I am not interested in teaching DSP 101.
I am not willing to spend the time to address every misunderstanding of DSP someone has.
I appreciate having someone point out an error I make.
If you are not willing to take the time to demonstrate an alleged error on my part, I shall ignore you unless it is stated such that it is immediately obvious.
Please do not raise issues which are amply addressed by references I provide. If you are not willing to read the references, I'm not willing to explain it to you.
I'm sorry, but I'm not a "nice" guy. I am not willing to indulge adults who still behave like children.
Have Fun!
Reg
-
So, I sat down in front of the scope and tried the 100Hz modulation of a 20MHz signal - nothing like a real example, right?
It seems to me that the digital scope behaves pretty much like an analog scope in this case. The slow sweep speed does not seem to detract from the ability to display an outline of the 20MHz signal (even without Peak Detection enabled).
Obviously, filtering out the 20MHz would destroy the information as displayed here.
(note: normal mode, not peak detect)
(Attachment Link)
Thank you for a very nice example. As you correctly note, it's obvious that you cannot sensibly filter out the carrier and not kill the side bands.
If the signal were down converted to DC - 500 Hz with the carrier at 250 Hz followed by a 500 Hz LP filter it would be very accurately displayed in the frequency domain provided the programmer took the frequency translation into account. A mere change in the carrier frequency would not change the picture unless the ratio of carrier and the modulation were very small. And that would only be in the time domain.
Have Fun!
Reg
-
If you properly downsample with low pass filtering rather than decimating by throwing away samples you will *always* have a band limited spike. It will *not* disappear. However, it *will* make "peak detection" moot.
Won't its amplitude change? And why would it 'disappear' using decimation provided its bandwidth is under Nyquist? EDIT: To be clear, I mean the Nyquist limit of the new, lower sample rate.
-
Won't its amplitude change?
Nothing in the passband will change amplitude.
And why would it 'disappear' using decimation provided its bandwidth is under Nyquist?
When you throw away samples you're doing very bad things in the frequency domain.
-
So, I sat down in front of the scope and tried the 100Hz modulation of a 20MHz signal - nothing like a real example, right?
It seems to me that the digital scope behaves pretty much like an analog scope in this case. The slow sweep speed does not seem to detract from the ability to display an outline of the 20MHz signal (even without Peak Detection enabled).
Obviously, filtering out the 20MHz would destroy the information as displayed here.
(note: normal mode, not peak detect)
(Attachment Link)
Thank you for a very nice example. As you correctly note, it's obvious that you cannot sensibly filter out the carrier and not kill the side bands.
If the signal were down converted to DC - 500 Hz with the carrier at 250 Hz followed by a 500 Hz LP filter it would be very accurately displayed in the frequency domain provided the programmer took the frequency translation into account. A mere change in the carrier frequency would not change the picture unless the ratio of carrier and the modulation were very small. And that would only be in the time domain.
Have Fun!
Reg
I can see how the down-conversion won't damage the frequency domain display, but the time domain looks terrible...
Below, I used a 250Hz carrier modulated by 100Hz to demonstrate the downconverted case.
[attachimg=2]
It doesn't really look good in the time domain. To make it look similar to the 20MHz image, I had to lower the frequency of the modulation and increase the sweep time.
Using my limited math skills -
The first example with the 20MHz carrier could have been sampled at 40MHz for a sweep of 50ms, so would consist of 2,000,000 samples.
The downconverted 250Hz carrier is samled at 500Hz for 4,000 seconds, with a modulation of 0.0015Hz. This also has 2,000,000 samples and looks exactly the same.
It seems to me that to display this waveform, we end up with the same number of data points needing to be plotted even after downconversion?
Example with 250Hz carrier, but 0.5Hz modulation (scope cannot sweep slow enough to show 0.0015Hz)
[attachimg=1]
-
Won't its amplitude change?
Nothing in the passband will change amplitude.
And why would it 'disappear' using decimation provided its bandwidth is under Nyquist?
When you throw away samples you're doing very bad things in the frequency domain.
Some numbers in an example would help. Unless I'm missing something, throwing away samples only does bad things if the remainder is insufficient to meet the Nyquist criterion, plus a margin for the practical limits of reconstruction. And if you use LP filter on a pulse with a bandwidth within Nyquist before decimation but beyond it after, it will be smeared or spread out and have a lower amplitude. This is what peak detection prevents.
-
Won't its amplitude change?
Nothing in the passband will change amplitude.
And why would it 'disappear' using decimation provided its bandwidth is under Nyquist?
When you throw away samples you're doing very bad things in the frequency domain.
Some numbers in an example would help. Unless I'm missing something, throwing away samples only does bad things if the remainder is insufficient to meet the Nyquist criterion, plus a margin for the practical limits of reconstruction. And if you use LP filter on a pulse with a bandwidth within Nyquist before decimation but beyond it after, it will be smeared or spread out and have a lower amplitude. This is what peak detection prevents.
Say we have a sampled waveform in a buffer with 10,000,000 samples that we want to display on our screen, which has a width of 1,000 pixels.
The sampled waveform is a 20MHz carrier, modulated by 100 Hz, sample length 50ms. The sample rate is 200MHz.
Can we get from 10,000,000 samples to the 1,000 pixel available width on the screen without throwing anything away... ?
-
Below, I used a 250Hz carrier to demonstrate the downconverted case. But here's the crunch - I also had to lower the frequency of the modulation and increase the sweep time to make the waveform look like the first example...
Using my limited math skills -
The first example with the 20MHz carrier could have been sampled at 40MHz for a sweep of 50ms, so would consist of 2,000,000 samples.
The downconverted 250Hz carrier is samled at 500Hz for 4,000 seconds, with a modulation of 0.0015Hz. This also has 2,000,000 samples and looks exactly the same.
Down-conversion (down-mixing) does not preserve the overall time-domain waveform shape. For the given AM signal it only preserves the envelope. I would not use an IF of of only 250 Hz, though, but rather say 10 kHz (->leading to still 100 carrier cycles per envelope period then). Down-conversion is not generally useful, but only for particular signals. You need some a priori knowledge about the signal in order to decide whether it can be used at all, and to establish a suitable frequency plan.
On the other hand, if you would decimate to a samling rate < 4.99995 39.9998 MSPS with ideal boxcar AA-filtering, then you lose the signal completely.
Decimation with peak detector instead of AA filter still remains a very useful alternative to get what you want to see.
EDIT: Just noticed that the given carrier was 20MHz.
-
Can we get from 10,000,000 samples to the 1,000 pixel available width on the screen without throwing anything away... ?
When you decimate for sceen display, you are not limited to a 1-dimensional destination signal vector, but you have more opportinuties. E.g. you can light up multiple pixels in the same column, and you can assign different intensity and/or color to the pixels.
You could, for instance, map the first 10,000 samples to colum 0, and set all pixels in this column which correspond to the 10,000 values, then map the next 10,000 samples to column 1, etc. The pixel brightness/color could depend on the number of values which hit the pixel. Just be creative ;)
-
Please do not raise issues which are amply addressed by references I provide. If you are not willing to read the references,..
Well, it's possibly rather the missing willingness to buy a dozen of (printed) books, just in order to read one chapter then. If you could provide references as web links to (freely) downloadable documents (which focus on one particular subject, and not the whole DSP world in hundreds or thousands of pages), then I guess that more people were willing to read them ;)
-
Won't its amplitude change?
Nothing in the passband will change amplitude.
And why would it 'disappear' using decimation provided its bandwidth is under Nyquist?
When you throw away samples you're doing very bad things in the frequency domain.
No, I have to correct you there. Decimation (by throwing away samples, periodically) will have absolutely same effect as sampling with lower frequency. Samples are not integrated voltage between intervals, but discrete point in time voltages, points, not little lines. Dot mode on scope is showing exactly what is being sampled. So at any sample frequency, you get same samples (in theory) only more or less dense or sparse.
Actually, faster A/D will have closer to ideal (shorter) sample window and in theory be more mathematically "correct".
So sampling at 5 GS/s and keeping every 5th sample, and sampling at 1 GS/s and keeping every single one is SAME.
By applying averaging (as in Hires), you can get signal with less noise, so converting from 5 to 1 GS/s by filter will be more beneficial that simply decimating.
But sample per sample decimation gives same result as lower sample rate (if we exclude fact that slower A/D sampling on REAL A/D will probably have better ENOB by virtue of lower operating frequency).
-
Can we get from 10,000,000 samples to the 1,000 pixel available width on the screen without throwing anything away... ?
Of course.
Try even one time (example Siglent) so that you capture (example) 14ms long capture to 14M memory and so that whole captured length is displayed. Because there is 14 div time scale 1ms/div. Just single sho, exmple with wideband high level noise and then scope stopped. You have there every single sample mapped to screen. Trace length is 700 pixel. So first displayed signal area column have 20000 samples. After capture done then start zooming in and look carefully what all kind of things you can see.
It must not decimate! Do you know why. Because it is for analyzing signals and not for make entertainment images or lottery.
Yes old time there was once mistake and there was "accidentally" some decimation before display mapping... perhaps some peoples who are still angry to Siglent remember this quite short time case before it was corrected...
-
No, I have to correct you there. Decimation (by throwing away samples, periodically) will have absolutely same effect as sampling with lower frequency. Samples are not integrated voltage between intervals, but discrete point in time voltages, points, not little lines. Dot mode on scope is showing exactly what is being sampled. So at any sample frequency, you get same samples (in theory) only more or less dense or sparse.
OK, let's sample a sine wave at 2x frequency. We might get values [0, 1, 0, -1, 0, 1, 0, -1, ....]
Now let's throw away half the samples. Depending on where we start we might get 0,0,0,0 or we might get 1,-1, 1,-1. It's not the same thing at all.
If we apply a very simple low pass filter to those same numbers we might get [0.5,-0.5,0.5,-0.5, ...] and that will always work no matter where we start in the sequence.
-
By applying averaging (as in Hires), you can get signal with less noise, so converting from 5 to 1 GS/s by filter will be more beneficial that simply decimating.
In fact it is a digital low-pass filter, applied before the decimation, usually having a boxcar pulse response. This filter kernel has a -3dB point in the frequency domain is at ~0.9 * Nyquist (of the decimated rate). As long as the resulting roll-off in the pass-band does not matter, one could use it generally instead of plain decimation. Still it is by far not a suitable replacement for a proper AA filter, in case AA filtering is desired.
OK, let's sample a sine wave at 2x frequency....
2x frequency is not enough. It still violates the sampling theorem. It must be greater than 2x frequency.
Now let's throw away half the samples.
This violates the sampling theorem even more.
In order that you don't suffer from aliasing, the signal frequency must be < half of the decimated sampling rate.
-
For those who think scope manufacturers are stupid and don't know about DSP some material:
http://cdn.teledynelecroy.com/files/whitepapers/dsp_in_oscilloscopes.pdf (http://cdn.teledynelecroy.com/files/whitepapers/dsp_in_oscilloscopes.pdf)
https://www.edn.com/advantages-and-disadvantages-of-using-dsp-filtering-on-oscilloscope-waveforms/ (https://www.edn.com/advantages-and-disadvantages-of-using-dsp-filtering-on-oscilloscope-waveforms/)
Also a document on methods of establishing GUM for dynamic signals..
-
No, I have to correct you there. Decimation (by throwing away samples, periodically) will have absolutely same effect as sampling with lower frequency. Samples are not integrated voltage between intervals, but discrete point in time voltages, points, not little lines. Dot mode on scope is showing exactly what is being sampled. So at any sample frequency, you get same samples (in theory) only more or less dense or sparse.
OK, let's sample a sine wave at 2x frequency. We might get values [0, 1, 0, -1, 0, 1, 0, -1, ....]
Now let's throw away half the samples. Depending on where we start we might get 0,0,0,0 or we might get 1,-1, 1,-1. It's not the same thing at all.
If we apply a very simple low pass filter to those same numbers we might get [0.5,-0.5,0.5,-0.5, ...] and that will always work no matter where we start in the sequence.
Read again. Decimation is same as decreasing sampling frequency. If you sample at 100 MS/S, and take every 100th sample, it is same as if you were sampling at 1 MS/s. Off course it will be different than original. Statement is that if you sample at higher clock and decimate it is same as slower sample rate..
And that is what scopes are doing most of the time when timebases get long. They don't switch to Hires mode without asking. Which is what you get when filtering.
This is what I do manually. If I'm looking at something fast I sample normaly. If I'm looking at slow signals, where sample rate drops, I manually switch to Hires mode because efective sample rate will be the same, and Hires will be less noisy.
-
If we apply a very simple low pass filter to those same numbers we might get [0.5,-0.5,0.5,-0.5, ...] and that will always work no matter where we start in the sequence.
Yes, you have just invented the basic principle of an anti-aliasing filter ;)
(As said above (https://www.eevblog.com/forum/testgear/scope-wars/msg3117412/#msg3117412), your example signal is too fast and suffers from aliasing.)
EDIT:
[A perfect AA filter would have cut-off your signal completely, though, so that the sampled signal were 0,0,0,0... and the decimate signal were 0,0,... too.]
If you retry your example with a sine wave signal having e.g. 5x 1/5 the frequency of the initial samping rate then you'll be able to reconstruct the sine wave exactly with sinc interpolation, even after decimation by a factor 2 (granted that you capture either an infinte number of samples, or an exact integral number of signal periods).
-
Read again. Decimation is same as decreasing sampling frequency. If you sample at 100 MS/S, and take every 100th sample, it is same as if you were sampling at 1 MS/s. Off course it will be different than original. Statement is that if you sample at higher clock and decimate it is same as slower sample rate..
I thought the statement was that applying a low pass filter is the same as throwing away samples.
-
Read again. Decimation is same as decreasing sampling frequency. If you sample at 100 MS/S, and take every 100th sample, it is same as if you were sampling at 1 MS/s. Off course it will be different than original. Statement is that if you sample at higher clock and decimate it is same as slower sample rate..
I thought the statement was that applying a low pass filter is the same as throwing away samples.
No it wasn't. Sorry for confusion.
The OP is constantly shifting what is being discussed, ignoring warnings that despite math is being correct most of the stuff mentioned is either not applicable to GP scope, or wouldn't serve much purpose, or would be expensive, or would simply convert scope to some new category of instrument (scopalyzer ?) that is debatable whether it is something users would want.
It reminds me of : "To a man with a hammer everything looks like a nail..." Or however that saying goes...
-
Mostly I'm ignoring the fact that people are saying things that are so wrong as to be simply silly.
I did not shift the topic. I started this as a comparison of DSOs, not a debate about what people *think* a scope should do. My frame of reference is a good analog scope. So I am comparing DSOs to good analog scope signals.
The purpose of a DSO is to faithfully represent an *analog* signal. That requires low pass filtering to prevent aliasing.
Decimation without low pass filtering causes aliasing. It also causes fast impulses to not be captured at low sweep rates. With proper low pass filtering there is no need for "peak detection".
There is a vast amount of literature on DSP. I reference the books I use. I am not going to search the web *for you*. Do it yourself. I'm not interested in teaching DSP 101 to anyone, especially people who claim to understand it and then manifestly demonstrate they have not grasped the contents of the first 10 pages of a basic intro to the subject.
My preferred basic DSP reference:
An Introduction to Digital Signal Processing
John H. Karl
Academic Press 1989
It's the book I reach for first when I want to check something.
Have Fun!
Reg
-
Decimation without low pass filtering causes aliasing. It also causes fast impulses to not be captured at low sweep rates. With proper low pass filtering there is no need for "peak detection".
Decimation by discarding samples doesn't cause aliasing, it is fact that effective sample rate drops by doing it that does. If you want to give us crap about rigorous math, than be accurate.
But you are correct, doing full speed sampling and then downsampling it by filtering is an effective antialiasing method. And many scopes do it. Even little DS1054Z has it as a setting. Keysight 3000T does it too AFAIK because no aliasing is visible in normal use.
But no need for Peak detect?
Pray tell how does THAT work? MSOX3104T will show a 250 ps pulse on 1 second/div timebase in Peak detect mode.
If you lowpass filter samples you will filter OUT that peak. What am I missing? It will be gone.
Your solution is to pretend it's not there? Is that what Bendat and Piersol recommend ?
-
Mostly I'm ignoring the fact that people are saying things that are so wrong as to be simply silly.
I did not shift the topic. I started this as a comparison of DSOs, not a debate about what people *think* a scope should do. My frame of reference is a good analog scope. So I am comparing DSOs to good analog scope signals.
The purpose of a DSO is to faithfully represent an *analog* signal.
No. An oscilloscope is there to faithfully represent the shape of a signal. Your statement applies to digitizers which are a different beast compared to an oscilloscope even though at first sight there is a lot of overlap in functionality. On top of that some oscilloscopes are more on the digitizer side (the ones from Lecroy for example).
Your statement about peak detect also isn't true. Peak-detect retains the amplitude of a pulse. Decimation spreads the energy of that peak over a longer sample interval making the amplitude lower and thus distorting the shape of the original signal.
Bottom line: from a signal processing perspective you are right but an oscilloscope is not a device intended to do signal processing.
edit: typo
-
Decimation without low pass filtering causes aliasing. It also causes fast impulses to not be captured at low sweep rates. With proper low pass filtering there is no need for "peak detection".
What kind of humor is this, I do not recognize. I know parody, irony, satire etc but this I do not recognize.
If someone try tell me that digital scope do not need peak detect... and he try sell me this kind of scope. Just he can keep his scope and I keep my money. Well I have some experience about real works with oscilloscopes for many kind of purposes tens of years including also analog scopes and later digital ones, in real world in real practice... works with hands and tools. If some theory book writer or reader tell me no need peak mode I can as he go back to his cave reading more theory books and lets others do real practical works. But with ;) ;) ;)
-
Can we get from 10,000,000 samples to the 1,000 pixel available width on the screen without throwing anything away... ?
When you decimate for sceen display, you are not limited to a 1-dimensional destination signal vector, but you have more opportinuties. E.g. you can light up multiple pixels in the same column, and you can assign different intensity and/or color to the pixels.
You could, for instance, map the first 10,000 samples to colum 0, and set all pixels in this column which correspond to the 10,000 values, then map the next 10,000 samples to column 1, etc. The pixel brightness/color could depend on the number of values which hit the pixel. Just be creative ;)
Yes, that looks like exactly what my scope is doing.
Looking at the same 20MHz modulated by 100Hz, this time with vectors turned off (points mode, in other words)...
[attachimg=3]
If we crank up the Y axis gain, the screen turns into a snowstorm, showing that more than one Y point is plotted for each X...
[attachimg=1]
The scope's FFT function shows a blank display... doesn't seem to work when the buffer has more than one Y value per X coordinate.
[attachimg=2]
-
I started this as a comparison of DSOs, not a debate about what people *think* a scope should do. My frame of reference is a good analog scope. So I am comparing DSOs to good analog scope signals.
The purpose of a DSO is to faithfully represent an *analog* signal. That requires low pass filtering to prevent aliasing.
If you are going to compare or critique products without listening to anyone about how those products are used, your results will be meaningless to most people. Here, IMO, your premise is entirely wrong. It is not the purpose of a DSO to replace an analog scope nor to 'faithfully represent an analog signal', whatever that means. The purpose of a DSO is, like Forrest Gump in boot camp, to do exactly what I tell it to do. That might be 'faithfully represent an analog signal' if it is your scope and that is what you want. I'll agree that a typical entry level DSO should be able to be configured to do so. But the show doesn't stop there. DSOs are capable of much, much more and that is why we have them.
Also, since you are comparing and critiquing existing products, your statements on decimation and aliasing need to be reckoned with what the existing scopes actually do. Can you show us an example of one of those scopes showing aliasing on the screen due to decimation of a signal that was within Nyquist at the native sample rate but is now aliased because of decimation? Is that happening? If so, I'll grant that to be an undesirable result.
Decimation without low pass filtering causes aliasing. It also causes fast impulses to not be captured at low sweep rates. With proper low pass filtering there is no need for "peak detection".
Only if the signal wasn't already aliased and has a content above the new, lower Nyquist limit. That is what peak detect is for. On the last point, I think I can devise an experiment to prove you wrong. :)
There is a vast amount of literature on DSP. I reference the books I use. I am not going to search the web *for you*. Do it yourself. I'm not interested in teaching DSP 101 to anyone, especially people who claim to understand it and then manifestly demonstrate they have not grasped the contents of the first 10 pages of a basic intro to the subject.
Perhaps you could answer one simple question. I did look up 'folded LP filter' and I'm not finding much that I can access and understand. From your description, it is a clever, fast implementation that combines FIR, LP and downsampling all in one. Is that correct? Can I assume the the actual effect on a signal of your folded LP filter will be similar to any other LP filter? Or are there special characteristics that make it perform as you say, specifically eliminating the need for peak detection?
-
Perhaps you could answer one simple question. I did look up 'folded LP filter' and I'm not finding much that I can access and understand. From your description, it is a clever, fast implementation that combines FIR, LP and downsampling all in one. Is that correct? Can I assume the the actual effect on a signal of your folded LP filter will be similar to any other LP filter? Or are there special characteristics that make it perform as you say, specifically eliminating the need for peak detection?
It's folded FIR filtering, a specific technique of removing the number of taps / implementation complexity in exchange for adders.. this might help you
https://www.embedded.com/dsp-tricks-an-odd-way-to-build-a-simplified-fir-filter-structure/ (https://www.embedded.com/dsp-tricks-an-odd-way-to-build-a-simplified-fir-filter-structure/)
-
To prove that modern DSOs do in fact alias on long timebases:
Attached images show Rigol DS1074Z aliasing with 50MHz input signal on long timebase.
The "Anti-Aliasing" setting does not prevent aliasing. (Not entirely sure what it does other than slow down the scope and make the render look prettier: they probably should have called it "high-quality" or something similar.) Hi-res mode also shows an aliased signal: whatever hi-res averaging Rigol is using appears to be post-decimation. Peak detect does prevent aliasing.
-
Lessons for me in this thread so far:
1) A scope can plot more information on its screen than a pure analog signal (for example, by creating histograms of points for each X value). Paradoxically, an analog scope can do this too! Example: 100Hz modulation of 20MHz carrier, 50ms sweep.
2) The FFT function in the typical digital scope is limited by the need to display the time domain signal at the same time. Sometimes it is not possible to get a good compromise between the two and in some cases it doesn't work at all (example: 100Hz modulation of 20MHz carrier, 50ms sweep).
3) Peak detect appears to not really have an analog scope equivalent - a very short blip during a slow sweep will be displayed on the analog scope, but is likely hard to see without cranking up brightness - whereas a digital scope will plot a "fully formed pixel" that is as easy to see as any other. That pixel's width is going to be out of proportion to the actual signal and therefore not really a theoretically accurate representation of the duration of the signal. That doesn't make it any less useful, of course. Perhaps better to think of it as a marker than a part of the signal...
-
It's folded FIR filtering, a specific technique of removing the number of taps / implementation complexity in exchange for adders.. this might help you
https://www.embedded.com/dsp-tricks-an-odd-way-to-build-a-simplified-fir-filter-structure/ (https://www.embedded.com/dsp-tricks-an-odd-way-to-build-a-simplified-fir-filter-structure/)
Thanks! So the 'folding' just takes advantage of the fact that the multiplier coefficient for some of the taps is the same in some cases. I don't see how the OP's LP filter plan differs in function from ERES:
http://cdn.teledynelecroy.com/files/appnotes/an_006a.pdf (http://cdn.teledynelecroy.com/files/appnotes/an_006a.pdf)
Actually see no reason to think that manufacturers aren't already using a folded version of the filter. I hadn't heard of 'folding' in this context because I'm wholly ignorant of the inner details of DSP, but I'm sure the engineers at LeCroy have heard of it.
-
Perhaps better to think of it as a marker than a part of the signal...
All good :-+ except this last phrase. If it appears it's really part of the signal.
-
An oscilloscope is there to faithfully represent the shape of a signal.
:-+ But doesn't this phrase contradict your tagline? ? :D
BTW, I would add "...to try to faithfully.."
-
Perhaps better to think of it as a marker than a part of the signal...
All good :-+ except this last phrase. If it appears it's really part of the signal.
For example: you have a screen that is 1,000 pixels wide, and a sweep speed of 1 second, and you are using Points mode just to avoid the complication of talking about vectors in this example. This setting means each pixel width is "worth" one millisecond - each pixel always represents exactly 1ms of data.
Now turn on Peak Detect. Say that a 5ns pulse is detected. The lit pixel is still 1ms wide, which makes it look like the pulse was 1ms, which is really not correct... it is a big error! That's why I like thinking of the peak detect as a marker... there could be a pulse behind it anywhere from 5ns (or even less, depending on your scope) to 1ms, we can't know what it actually is until we look closer at a faster sweep.
A peak detected waveform could show incorrect results in the FFT view because of this error, not sure how scope manufacturers deal with that...
-
I started this as a comparison of DSOs, not a debate about what people *think* a scope should do. My frame of reference is a good analog scope. So I am comparing DSOs to good analog scope signals.
The purpose of a DSO is to faithfully represent an *analog* signal. That requires low pass filtering to prevent aliasing.
[...]
Can you show us an example of one of those scopes showing aliasing on the screen due to decimation of a signal that was within Nyquist at the native sample rate but is now aliased because of decimation? Is that happening? If so, I'll grant that to be an undesirable result.
Sorry for the interruption and although USB-oscilloscopes are probably off topic, but at this point I have to vent my disappointment about Agilent's U2702A (200MHz, 1GS/s, 32Mpts) in conjunction with Agilent Measurement Manager 2.2.4.0. This USB-oscilloscope can store a waveform from a single-shot-acquisition in its internal memory, then the software on the computer retrieves some amount of this data to display the waveform on the screen. You can zoom in and out, but unfortunately the decimation involved in this is done wrong, so aliasing occurs. The data is there (in the device), but the presentation is a failure. Unnecessarily. PicoScope handles this better.
The pictures show the same acquisition in different time/div settings (200µs/div and 500ns/div).
-
An oscilloscope is there to faithfully represent the shape of a signal.
:-+ But doesn't this phrase contradict your tagline? ? :D
BTW, I would add "...to try to faithfully.."
Well... after reading the dictionary I think you could have a very long semantic discussion on this. >:D
-
To prove that modern DSOs do in fact alias on long timebases:
Attached images show Rigol DS1074Z aliasing with 50MHz input signal on long timebase.
The "Anti-Aliasing" setting does not prevent aliasing. (Not entirely sure what it does other than slow down the scope and make the render look prettier: they probably should have called it "high-quality" or something similar.) Hi-res mode also shows an aliased signal: whatever hi-res averaging Rigol is using appears to be post-decimation. Peak detect does prevent aliasing.
OK, if you use the memory limitation to force the scope to reduce its sampling rate you should expect aliasing. The fact that the scope reduces its sampling rate by decimation rather then slowing the ADC clock is just the way it works--the results wouldn't change except maybe to the extent that a slower ADC might have a longer sample window.
The OP has proposed:
Data for time domain display is downsampled to sample rate appropriate to timebase setting using a folded LP filter
which I had interpreted as meaning that the LP/downsampling operation would be applied between the memory and the screen (post acquisition) not between the ADC and memory. However, if the folded filter can be applied in real-time, what would that look like?
Which of the three options posted here is the most 'correct' response--the one you want to see--to a signal at twice Nyquist of the posted sample rate?
If you apply the appropriate LP filter for a 20mS/div display to a 50MHz signal, what is the appropriate result? A flat line?
In this case, an LP filter might be the appropriate default. Aliasing has to be prevented by limiting BW before you impose the Nyquist limit by sampling or downsampling as the case may be, so if it can be done in real-time, great--especially if the user can have explicit control over the filter parameters. However, I think it will add some complexity that will be hard to justify in an entry level product. The FFT will need its own data channel and memory.
I cannot see how this can be claimed to eliminate the need for peak detection. I also don't have any high-end DSOs so I have no idea if any of this has already been implemented somewhere. Anyone?
-
An oscilloscope is there to faithfully represent the shape of a signal.
:-+ But doesn't this phrase contradict your tagline? ? :D
BTW, I would add "...to try to faithfully.."
I think the correct phrase is "... to endeavor to try to faithfully represent ..."
I prefer the more formal "... to attempt to endeavor to try to faithfully represent ..."
-
OK, if you use the memory limitation to force the scope to reduce its sampling rate you should expect aliasing. The fact that the scope reduces its sampling rate by decimation rather then slowing the ADC clock is just the way it works--the results wouldn't change except maybe to the extent that a slower ADC might have a longer sample window.
Decimation and slowing the ADC clock are functionally equivalent. It's probably doing a combination of both as the HMCAD1511 ADC inside the Rigol only supports sample rates down to 120MSa/s or so. In fact given the peak detect doesn't seem to fall foul of aliasing it may well just be doing the decimation on the FPGA - that's not entirely clear to me though.
The correct response would be to apply a filter once the decimation becomes necessary, but this is computationally expensive. The Spartan-6 used in the Rigol DS1000Z only has ~90 DSP cores, which limits it to ~90 tap filter. There may be a "cheaper" way to do it, but it is a limitation of hardware here, I believe. I'm not sure how complex of a filter you would need or if there is a way to implement one with an IIR or moving average.
-
Now turn on Peak Detect. Say that a 5ns pulse is detected. The lit pixel is still 1ms wide, which makes it look like the pulse was 1ms, which is really not correct... it is a big error! That's why I like thinking of the peak detect as a marker... there could be a pulse behind it anywhere from 5ns (or even less, depending on your scope) to 1ms, we can't know what it actually is until we look closer at a faster sweep.
Understand your thinking. Just remember, we are in peak mode so it's "more correct" (IMO) to display that 1 pixel than not displaying it at all. If you zoom out you'll still be in peak mode and it will be "even more correct".
And, if it appears (by any motive) we can be sure that there was a sample (as shorter as it could be) that had that value.
-
Well... after reading the dictionary I think you could have a very long semantic discussion on this. >:D
:D No semantics here. I think it's a perfect image of what this thread has been showing...
Another form: ;)
"An oscilloscope is there to represent the shape of a signal as faithfully as the owner expects and/or is willing to pay for."
-
On the subject of peak detection... I got interested in whether the scope treats peak detected pulses as "real" for the purposes of math operations probably including FFT.
It appears that at least in the case of the 54622D, that the peak detected pulses are NOT considered "bona fide signal material" for math operations.
Example: Channel A is a 100Hz sine. Channel B is a train of 5ns pulses coming in at 200Hz, which are not visible unless Peak Detect is turned on. The middle, unmarked channel is the Math channel, displaying A-B (scaled down to fit the display). As you can see, there are no signs of the peak detected pulses in A-B, meaning that Agilent appear to be using the peak detection only as a marker, not as a real signal.
[attachimg=1]
-
Decimation and slowing the ADC clock are functionally equivalent. It's probably doing a combination of both as the HMCAD1511 ADC inside the Rigol only supports sample rates down to 120MSa/s or so. In fact given the peak detect doesn't seem to fall foul of aliasing it may well just be doing the decimation on the FPGA - that's not entirely clear to me though.
Agreed as to equivalence as long as the sample window is the same in each. In this case it seems pretty clear that it can implement the peak detect in real time, so I would assume it does it at 1GSa/S--otherwise it might not work at other speeds. There would be no reason to slow it down.
The correct response would be to apply a filter once the decimation becomes necessary, but this is computationally expensive. The Spartan-6 used in the Rigol DS1000Z only has ~90 DSP cores, which limits it to ~90 tap filter. There may be a "cheaper" way to do it, but it is a limitation of hardware here, I believe. I'm not sure how complex of a filter you would need or if there is a way to implement one with an IIR or moving average.
Now that I know about folded DSP filters, I would think that the simplest, fastest way would be to use the same number of taps as the downsampling ratio all added together into one multiplier with a coefficient of 1/downsampling ratio. IOW, just an average over the samples. I'm guessing this is already widely done, folding and all.
-
Now turn on Peak Detect. Say that a 5ns pulse is detected. The lit pixel is still 1ms wide, which makes it look like the pulse was 1ms, which is really not correct... it is a big error!
A peak detected waveform could show incorrect results in the FFT view because of this error, not sure how scope manufacturers deal with that...
The screen can't possibly show you everything all at once--something has to be limited, whether it is resolution vs time or resolution vs amplitude. When you select peak detect, you are specifying that its priority is to show you the maximum and/or minimum amplitude of any signal it can detect, to the detriment of all other factors. It's not an error, it is an unavoidable feature.
Peak detect and FFT would seem wholly incompatible, or at least not accurate.
-
Now turn on Peak Detect. Say that a 5ns pulse is detected. The lit pixel is still 1ms wide, which makes it look like the pulse was 1ms, which is really not correct... it is a big error!
A peak detected waveform could show incorrect results in the FFT view because of this error, not sure how scope manufacturers deal with that...
The screen can't possibly show you everything all at once--something has to be limited, whether it is resolution vs time or resolution vs amplitude. When you select peak detect, you are specifying that its priority is to show you the maximum and/or minimum amplitude of any signal it can detect, to the detriment of all other factors. It's not an error, it is an unavoidable feature.
Peak detect and FFT would seem wholly incompatible, or at least not accurate.
Wonder if any scope manufacturers have fallen into that hole?
-
Hi,
I have scanned this (long) thread from beginning to end.
The impression that I get is that we are ridiculing comparing some 100MHz (class) scopes based only on their ability to perform FFTs without making (rookie) mistakes.
or we are expecting the FFT of a broadband source, impulse function, to reveal the filter shapes.
Given that a lot of these scopes have SW options to increase the BW to a maximum, that alone would imply the BW is determined by SW.
I tested my Tektronix MDO4104 ($30k) with an HP 8406A comb generator. The scope was behaving very nicely and not show FFT information above 1/2 the sampling frequency. The BW limiters worked as expected.
Did I miss something?
Regards,
Jay_Diddy_B
-
This is 50 MHz AM modulated with 100 Hz. It has very well functioning antialiasing filter.
[attach=1]
But, if you enable FFT, it starts aliasing.
[attach=2]
That means that it is mindful of the fact that FFT needs clean signal sampled under certain rules, so it disables mumbo jumbo and feeds FFT clean data. Antialiased "envelope display mode" is gone.
Also, on cursory glance, it doesn't seem Peak detect is fed to FFT. Averaging and Hires are, but for Peak detect and Normal FFT look pretty much the same.
So scope manufacturers know a bit about these things.
-
Hi,
I have scanned this (long) thread from beginning to end.
The impression that I get is that we are ridiculing comparing some 100MHz (class) scopes based only on their ability to perform FFTs without making (rookie) mistakes.
or we are expecting the FFT of a broadband source, impulse function, to reveal the filter shapes.
Given that a lot of these scopes have SW options to increase the BW to a maximum, that alone would imply the BW is determined by SW.
I tested my Tektronix MDO4104 ($30k) with an HP 8406A comb generator. The scope was behaving very nicely and not show FFT information above 1/2 the sampling frequency. The BW limiters worked as expected.
Did I miss something?
Regards,
Jay_Diddy_B
No, pretty much you're right.
-
Here are attempts at an eye diagram on the DS1102E, DS1202Z-E, MSO-2204EA and ZDS-2102A. Source is Keysight 33622A at 20 MHz 200 mVpp.
The Owon doesn't have an option to trigger on a rising or falling edge. So it's just here to be complete. The DS1202Z-E does, but it doesn't work.
Have Fun!
Reg
-
This is 50 MHz AM modulated with 100 Hz. It has very well functioning antialiasing filter.
(Attachment Link)
But, if you enable FFT, it starts aliasing.
(Attachment Link)
That means that it is mindful of the fact that FFT needs clean signal sampled under certain rules, so it disables mumbo jumbo and feeds FFT clean data. Antialiased "envelope display mode" is gone.
Also, on cursory glance, it doesn't seem Peak detect is fed to FFT. Averaging and Hires are, but for Peak detect and Normal FFT look pretty much the same.
So scope manufacturers know a bit about these things.
But the FFT is not showing anything at either 50MHz or 100Hz?
-
I have scanned this (long) thread from beginning to end.
The impression that I get is that we are ridiculing comparing some 100MHz (class) scopes based only on their ability to perform FFTs without making (rookie) mistakes.
Did I miss something?
No, you didn't miss anything. (Cheap) oscilloscopes make poor dynamic signal analyzers. Shocking.
-
Here is the DS1202Z-E and the MSO-2204EA being fed a 200 MHz and 300 MHz signal from an HP 8648C.
It's as good a demonstration of aliasing as you can get. I've got the scopes in dot mode with infinite persistence. Both scopes have all channels enabled to force 500 MSa/s so Nyquist is at 250 MHz.
I didn't think the Rigol was turning the anti-alias filter on and off when I was looking at the step response. The two photos of the Rigol screen prove that it does not work. I reset the persistence between photos and rest the frequency.
So, do you want a scope that shows you the truth or one that lies? These both lie. They indicate that the 300 MHz signal is 200 MHz.
Have Fun!
Reg
-
Given that a lot of these scopes have SW options to increase the BW to a maximum, that alone would imply the BW is determined by SW.
No, the BW in these scopes is determined by the band limiting amplifier IC in front of the ADC. All software does is sends commands to switch the band limits. Therefore in order to figure out a scope's front end filter shape all one needs to do is refer to that IC's datasheet.
-
This is 50 MHz AM modulated with 100 Hz. It has very well functioning antialiasing filter.
(Attachment Link)
But, if you enable FFT, it starts aliasing.
(Attachment Link)
That means that it is mindful of the fact that FFT needs clean signal sampled under certain rules, so it disables mumbo jumbo and feeds FFT clean data. Antialiased "envelope display mode" is gone.
Also, on cursory glance, it doesn't seem Peak detect is fed to FFT. Averaging and Hires are, but for Peak detect and Normal FFT look pretty much the same.
So scope manufacturers know a bit about these things.
But the FFT is not showing anything at either 50MHz or 100Hz?
I wasn't trying to get usable FFT, but demonstrating that switching on FFT changes how scope handles aliasing.
-
Are you saying it would be a bad idea for the scope to apply another, second filter that depends on the maximum frequency that the display is able to show at the selected sweep speed?
I think David Hess is referring to the situation where the samplerate has to drop because the memory is too short to capture a full screen at the maximum samplerate.
I am saying that applying a filter before decimation to prevent aliasing due to the lower sample rate after decimation is worse than not filtering at all because it corrupts the histogram of the input signal. That is what an anti-aliasing filter would have to do.
OTOH I wonder what good peak-detect does on the FFT operation anyway; do you want to have both enabled at the same time?
Peak detection is incompatible with many but not all measurements including FFTs; it also corrupts the signal histogram. Some very recent DSOs from R&S implement multiple decimation operations in parallel to produce multiple acquisition records, like sample, peak detect, high resolution, etc., simultaneously.
The design which I have been looking into produces multiple histograms whereas a DPO type of oscilloscope produces one, with the fastest histogram replacing peak detection and the others having higher resolution at lower sample rates. This makes the best use of limited hardware resources to produce an analog type display which captures as many characteristics of the input signal as possible.
I would prefer my digital scope to default the trigger position to one division from the left side of the screen, so you can see a little bit of "before" the trigger, and still use almost the whole screen for the rest of the signal.
You can of course manually move the trigger location horizontally, but that messes up with zooming in and out on a digital scope.
I am not sure what you are doing but most DSOs allow the trigger position within the acquisition record to be adjusted. In the photograph I posted earlier, it is difficult to see but the the trigger position is marked with a "T" one division from the left and about one division below the center horizontal graticule line.
As far as magnification not being aligned with the trigger, analog oscilloscopes always magnify starting at the center of the CRT which has nothing to do with the trigger position.
-
Here is the DS1202Z-E and the MSO-2204EA being fed a 200 MHz and 300 MHz signal from an HP 8648C.
It's as good a demonstration of aliasing as you can get. I've got the scopes in dot mode with infinite persistence. Both scopes have all channels enabled to force 500 MSa/s so Nyquist is at 250 MHz.
I didn't think the Rigol was turning the anti-alias filter on and off when I was looking at the step response. The two photos of the Rigol screen prove that it does not work. I reset the persistence between photos and rest the frequency.
So, do you want a scope that shows you the truth or one that lies? These both lie. They indicate that the 300 MHz signal is 200 MHz.
Have Fun!
Reg
And you would 'fix' this how? DSP isn't going to help.
These are budget brand devices that deliver a lot of 'bang for the buck' but can get you into trouble if you approach their limits unwarily. There are plenty of instruments that 'lie' to you if you don't understand their limitation and characteristics. Granted, the marketing on these things is borderline criminal, but I don't think there are any cheap fixes.
You have to respect ALL of the Nyquist criteria. You can use a sampling rate of 10X rated bandwidth with a gentler 6db/octave or so filter, or you can try to squeeze up closer to the limit with a fancier, more expen$ive analog front end. Either way, the filtering has to be done ahead of the ADC and either way, it costs money.
-
I started this as a comparison of DSOs, not a debate about what people *think* a scope should do. My frame of reference is a good analog scope. So I am comparing DSOs to good analog scope signals.
The purpose of a DSO is to faithfully represent an *analog* signal. That requires low pass filtering to prevent aliasing.
[...]
Can you show us an example of one of those scopes showing aliasing on the screen due to decimation of a signal that was within Nyquist at the native sample rate but is now aliased because of decimation? Is that happening? If so, I'll grant that to be an undesirable result.
Sorry for the interruption and although USB-oscilloscopes are probably off topic, but at this point I have to vent my disappointment about Agilent's U2702A (200MHz, 1GS/s, 32Mpts) in conjunction with Agilent Measurement Manager 2.2.4.0. This USB-oscilloscope can store a waveform from a single-shot-acquisition in its internal memory, then the software on the computer retrieves some amount of this data to display the waveform on the screen. You can zoom in and out, but unfortunately the decimation involved in this is done wrong, so aliasing occurs. The data is there (in the device), but the presentation is a failure. Unnecessarily. PicoScope handles this better.
The pictures show the same acquisition in different time/div settings (200µs/div and 500ns/div).
No DSO is "off topic". I'm on the warpath because I'm so ticked off at all the really bad DSOs from all the OEMs including Keysight. Thanks for the demonstrating the issue.
The limitation on what I test is having to buy the junk as it's hard to get a demo unit for a few months and I don't want to have my life dominated by a short loan period.
Have Fun!
Reg
-
This is 50 MHz AM modulated with 100 Hz. It has very well functioning antialiasing filter.
(Attachment Link)
But, if you enable FFT, it starts aliasing.
(Attachment Link)
That means that it is mindful of the fact that FFT needs clean signal sampled under certain rules, so it disables mumbo jumbo and feeds FFT clean data. Antialiased "envelope display mode" is gone.
Also, on cursory glance, it doesn't seem Peak detect is fed to FFT. Averaging and Hires are, but for Peak detect and Normal FFT look pretty much the same.
So scope manufacturers know a bit about these things.
But the FFT is not showing anything at either 50MHz or 100Hz?
I wasn't trying to get usable FFT, but demonstrating that switching on FFT changes how scope handles aliasing.
I am seeing exactly the same thing on the 54622D. Good to see that Keysight is still using the same code 20 years later...
20MHz carrier with 100Hz modulation.
Both the FFT and the time domain signals are deeply wrong and misleading here - is it trying to show us an 800Hz carrier, with side bands 100Hz away?
[attachimg=1]
-
Here is the DS1202Z-E and the MSO-2204EA being fed a 200 MHz and 300 MHz signal from an HP 8648C.
It's as good a demonstration of aliasing as you can get. I've got the scopes in dot mode with infinite persistence. Both scopes have all channels enabled to force 500 MSa/s so Nyquist is at 250 MHz.
I didn't think the Rigol was turning the anti-alias filter on and off when I was looking at the step response. The two photos of the Rigol screen prove that it does not work. I reset the persistence between photos and rest the frequency.
So, do you want a scope that shows you the truth or one that lies? These both lie. They indicate that the 300 MHz signal is 200 MHz.
Have Fun!
Reg
And you would 'fix' this how? DSP isn't going to help.
These are budget brand devices that deliver a lot of 'bang for the buck' but can get you into trouble if you approach their limits unwarily. There are plenty of instruments that 'lie' to you if you don't understand their limitation and characteristics. Granted, the marketing on these things is borderline criminal, but I don't think there are any cheap fixes.
You have to respect ALL of the Nyquist criteria. You can use a sampling rate of 10X rated bandwidth with a gentler 6db/octave or so filter, or you can try to squeeze up closer to the limit with a fancier, more expen$ive analog front end. Either way, the filtering has to be done ahead of the ADC and either way, it costs money.
Actually DSP can do the job. You can combine a modest analog filter followed by a digital filter to zero out the region in which the aliasing is present. Just give the user the control of the corner frequency of a digital filter applied to the live trace so they can control overshoot. It's only a problem for marketing and accounting. The former are inclined to lie and the latter are greedy. Neither the Instek nor Rigol is even 3 dB down from specified BW at Nyquist. That is sad. I'll measure the actual corner later. That's time consuming as I need to check the 8648C against my 438A and DDA-125 to make sure they all agree on levels at stated BW and Nyquist.
As I noted earlier, I've been working on this for some time and have more than enough experience to know what to do, how and what it costs in FPGA real estate.
It costs a bit. You can't claim that your 2 GSa/s scope has a 1 GHz BW. Realistically 500 MHz is the best you can do with a 1 GHz Nyquist when you take the cost of the analog filter into account. My issue is with $20K DSOs which have total crap performance.
When I bought a Keysight "Premium Used" MSOX3104T and put my <40 ps pulser on it I was shocked at the 450+ ps rise time. So I reread the datasheet. The "calculated" rise time is 0.35/BW *except* when you get to the 1 GHz version. Then it's 0.45/BW! Given my respect and trust of HP it was like discovering my wife had been cheating on me. The net result was I bought a huge pile of kit that says "HP". I've not been disappointed.
Have Fun!
Reg
-
Here is the DS1202Z-E and the MSO-2204EA being fed a 200 MHz and 300 MHz signal from an HP 8648C.
It's as good a demonstration of aliasing as you can get. I've got the scopes in dot mode with infinite persistence. Both scopes have all channels enabled to force 500 MSa/s so Nyquist is at 250 MHz.
I didn't think the Rigol was turning the anti-alias filter on and off when I was looking at the step response. The two photos of the Rigol screen prove that it does not work. I reset the persistence between photos and rest the frequency.
So, do you want a scope that shows you the truth or one that lies? These both lie. They indicate that the 300 MHz signal is 200 MHz.
Have Fun!
Reg
And you would 'fix' this how? DSP isn't going to help.
These are budget brand devices that deliver a lot of 'bang for the buck' but can get you into trouble if you approach their limits unwarily. There are plenty of instruments that 'lie' to you if you don't understand their limitation and characteristics. Granted, the marketing on these things is borderline criminal, but I don't think there are any cheap fixes.
You have to respect ALL of the Nyquist criteria. You can use a sampling rate of 10X rated bandwidth with a gentler 6db/octave or so filter, or you can try to squeeze up closer to the limit with a fancier, more expen$ive analog front end. Either way, the filtering has to be done ahead of the ADC and either way, it costs money.
Actually DSP can do the job. You can combine a modest analog filter followed by a digital filter to zero out the region in which the aliasing is present. Just give the user the control of the corner frequency of a digital filter applied to the live trace so they can control overshoot. It's only a problem for marketing and accounting. The former are inclined to lie and the latter are greedy. Neither the Instek nor Rigol is even 3 dB down from specified BW at Nyquist. That is sad. I'll measure the actual corner later. That's time consuming as I need to check the 8648C against my 438A and DDA-125 to make sure they all agree on levels at stated BW and Nyquist.
As I noted earlier, I've been working on this for some time and have more than enough experience to know what to do, how and what it costs in FPGA real estate.
It costs a bit. You can't claim that your 2 GSa/s scope has a 1 GHz BW. Realistically 500 MHz is the best you can do with a 1 GHz Nyquist when you take the cost of the analog filter into account. My issue is with $20K DSOs which have total crap performance.
When I bought a Keysight "Premium Used" MSOX3104T and put my <40 ps pulser on it I was shocked at the 450+ ps rise time. So I reread the datasheet. The "calculated" rise time is 0.35/BW *except* when you get to the 1 GHz version. Then it's 0.45/BW! Given my respect and trust of HP it was like discovering my wife had been cheating on me. The net result was I bought a huge pile of kit that says "HP". I've not been disappointed.
Have Fun!
Reg
That's why I ended up with an older model scope. - Basically, 450+ ps rise time is only ~10x better than a modest 20 year old 100MHz scope can deliver. But you pay much more than 10x the price for that.... and in return, you get a lower voltage rated, less solid feeling instrument.
I think a modern scope should be 100x or 1,000x better than 20 year old models... and not be ridiculously priced.
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
But that is an absolute minimum requirement.
As a thought experiment consider this model:
[attachimg=5]
It is a sample and hold circuit driven by a 1MHz sampling clock.
There is no need to worry about anti-aliasing filters because we are going to sample a pure sinewave so there is nothing to filter out.
The output of the sample and hold represents the data points that are stored in the scopes acquisition memory.
I have included a very simple reconstruction filter, a double pole filter at Fs/2.
This model helps us visualize the process.
I ran the model with different input frequencies.
100kHz
This is a fairly easy one. There are 10 samples of the input in the period and the sampling frequency is a multiple of the input frequency.
[attachimg=1]
The circuit can construct a reasonable representation of the input signal from the samples.
So a ratio of 10 sample per period gives very good results.
107kHz
There are 9.34 samples per period, not a multiple.
[attachimg=2]
You can still get a very reasonable representation of the input signal.
203KHz
Here just less than 5 samples period. Injecting a pure sinewave, so anti-aliasing will not help.
[attachimg=3]
There are barely enough data points to reconstruct the input signal.
303kHz
Still above the Nyquist limit about 3.3 samples period.
[attachimg=4]
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Regards,
Jay_Diddy_B
-
Actually DSP can do the job. You can combine a modest analog filter followed by a digital filter to zero out the region in which the aliasing is present. Just give the user the control of the corner frequency of a digital filter applied to the live trace so they can control overshoot.
If you have 500GSa/S on a 200MHz scope, a 350 MHz signal will fold down to 150MHz and I don't see how you can filter that out digitally without also zeroing out real 150MHz signals.
Neither the Instek nor Rigol is even 3 dB down from specified BW at Nyquist. That is sad. I'll measure the actual corner later.
How could they be? Even if the corner were at 200MHz, it would only be down another 1-2 db at 250MHz for a first-order filter. You would need a 24db/octave filter to even get close to proper solution to the specs they claim.
When I bought a Keysight "Premium Used" MSOX3104T and put my <40 ps pulser on it I was shocked at the 450+ ps rise time. So I reread the datasheet. The "calculated" rise time is 0.35/BW *except* when you get to the 1 GHz version. Then it's 0.45/BW! Given my respect and trust of HP it was like discovering my wife had been cheating on me. The net result was I bought a huge pile of kit that says "HP". I've not been disappointed.
Tektronix and others have application notes on this subject. They state that high bandwidth scopes that have lower sample rate to BW ratios need to use steeper low pass filters which affects the step response. They aren't lying--the scopes likely do have the bandwidth that they claim, they just won't have any 'extra' and step response is affected, thus the '0.45' rule for high bandwidth scopes. I don't think it is straight out specsmanship and my one Tek scope (200MHz, 2GSa/S) does actually roll off pretty hard. It's a flat line at 400MHz even with a 1GHz Nyquist.
-
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Regards,
Jay_Diddy_B
Bingo.. My minimum is at least 10:1 for this reason, preferably higher.. Yes mathematically 2 is the bare minimum but your structural resolution is shit and is really only good for a repeating fixed frequency signal.. The more dynamic the more resolution you actually need
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
But that is an absolute minimum requirement.
As a thought experiment consider this model:
(Attachment Link)
It is a sample and hold circuit driven by a 1MHz sampling clock.
There is no need to worry about anti-aliasing filters because we are going to sample a pure sinewave so there is nothing to filter out.
The output of the sample and hold represents the data points that are stored in the scopes acquisition memory.
I have included a very simple reconstruction filter, a double pole filter at Fs/2.
This model helps us visualize the process.
I ran the model with different input frequencies.
100kHz
This is a fairly easy one. There are 10 samples of the input in the period and the sampling frequency is a multiple of the input frequency.
(Attachment Link)
The circuit can construct a reasonable representation of the input signal from the samples.
So a ratio of 10 sample per period gives very good results.
107kHz
There are 9.34 samples per period, not a multiple.
(Attachment Link)
You can still get a very reasonable representation of the input signal.
203KHz
Here just less than 5 samples period. Injecting a pure sinewave, so anti-aliasing will not help.
(Attachment Link)
There are barely enough data points to reconstruct the input signal.
303kHz
Still above the Nyquist limit about 3.3 samples period.
(Attachment Link)
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Regards,
Jay_Diddy_B
Awesome simulation.
I guess matters improve a little bit if you introduce dithering and other magic tricks into the mix... but there's really no substitute for having some sample rate headroom.
When you get close to Nyquist, you are mostly looking at the characteristics of the reconstruction filter...
-
Basically, 450+ ps rise time is only ~10x better than a modest 20 year old 100MHz scope can deliver. But you pay much more than 10x the price for that.... and in return, you get a lower voltage rated, less solid feeling instrument.
I think a modern scope should be 100x or 1,000x better than 20 year old models... and not be ridiculously priced.
So, you think modern scopes should have 10 GHz (100x) or 100 GHz (1000x) bandwidth and not be ridiculously priced? Cool.
-
I posted demonstrations of the relationship of filter profile and overshoot a couple of weeks ago. No one seemed to pay any attention. Left me thinking, why bother. Someone did make an important point that if you see overshoot you need a faster scope, but in some cases a user controllable LP filter would do the trick. For example, if there is an impedance mismatch on a transmission line you can get a reflection that is pretty much impossible to see if the AFE rings too much. You can see it if you know exactly what to look for, but it is really hard. But if you suppress the ringing it's obvious.
A good friend from grad school in Austin is a manager for a seismic instrument maker. I'll ask him what their AFE performance is. With 10's of billions of dollars riding on the results, the seismic industry is very stringent about accurately reproducing the analog waveform. At 32 bits of resolution in today's market. Only possible if your data is below 120 Hz. And even then difficult and expensive.
Have Fun!
Reg
-
I posted demonstrations of the relationship of filter profile and overshoot a couple of weeks ago. No one seemed to pay any attention. Left me thinking, why bother. Someone did make an important point that if you see overshoot you need a faster scope, but in some cases a user controllable LP filter would do the trick. For example, if there is an impedance mismatch on a transmission line you can get a reflection that is pretty much impossible to see if the AFE rings too much. You can see it if you know exactly what to look for, but it is really hard. But if you suppress the ringing it's obvious.
A good friend from grad school in Austin is a manager for a seismic instrument maker. I'll ask him what their AFE performance is. With 10's of billions of dollars riding on the results, the seismic industry is very stringent about accurately reproducing the analog waveform. At 32 bits of resolution in today's market. Only possible if your data is below 120 Hz. And even then difficult and expensive.
Have Fun!
Reg
On the contrary you are forcing useful discussion of theory and practice.. I'm sure im not the only one that is finding it interesting but just staying out of the kitchen :-+ :popcorn:
-
Basically, 450+ ps rise time is only ~10x better than a modest 20 year old 100MHz scope can deliver. But you pay much more than 10x the price for that.... and in return, you get a lower voltage rated, less solid feeling instrument.
I think a modern scope should be 100x or 1,000x better than 20 year old models... and not be ridiculously priced.
So, you think modern scopes should have 10 GHz (100x) or 100 GHz (1000x) bandwidth and not be ridiculously priced? Cool.
Well, a Pentium Pro CPU of 1995 managed a 150Mhz clock frequency. I would guess (but haven't done the math) that a modern processor is roughly 100x to 1000x better, given increased clock frequencies, higher IPC (instructions per clock), and high amounts of parallelism / multiple cores.
Fair enough, it probably isn't fair to expect an entire instrument to evolve as fast as a single chip, especially ones that have as much interest and development budget behind them as PC CPUs.
Let's turn the question around, do you think the state of the art in oscilloscopes has advanced significantly over the last 20 years? Long obsolete models still command high prices on eBay, which indicates that the market doesn't seem to think so. Maybe it's the kind of work we do with them that hasn't evolved all that much?
-
Well, a Pentium Pro CPU of 1995 managed a 150Mhz clock frequency. I would guess (but haven't done the math) that a modern processor is roughly 100x to 1000x better, given increased clock frequencies, higher IPC (instructions per clock), and high amounts of parallelism / multiple cores.
Processor speed is not what limits the performance of an oscilloscope.
Let's turn the question around, do you think the state of the art in oscilloscopes has advanced significantly over the last 20 years? Long obsolete models still command high prices on eBay, which indicates that the market doesn't seem to think so. Maybe it's the kind of work we do with them that hasn't evolved all that much?
I paid over $30,000 for a new 1 GHz LeCroy just over 20 years ago. They can be purchased on eBay now for under $1,000. I wouldn't say they are commanding high prices.
At the time, 1 GHz was the fastest scope LeCroy made. Now you can buy one that has 100 GHz bandwidth. I'd say that's a significant advancement.
-
I recently bought a working 1 GHz DDA-120 for $303 delivered with tax. Quite amazing.
Of course, as they get faster it goes up rapidly. The 110 GHz Keysight is well over a million dollars.
Have Fun!
Reg
-
Well, a Pentium Pro CPU of 1995 managed a 150Mhz clock frequency. I would guess (but haven't done the math) that a modern processor is roughly 100x to 1000x better, given increased clock frequencies, higher IPC (instructions per clock), and high amounts of parallelism / multiple cores.
Processor speed is not what limits the performance of an oscilloscope.
That's an interesting comment - there are no opportunities for more processing, more parallelism - no opportunities for integration with other instruments (e.g. spectrum analyzer, VNA)?
Picture an instrument from the year 2030 or 2040. It isn't going to be depending on massive amounts of processing power?
Let's turn the question around, do you think the state of the art in oscilloscopes has advanced significantly over the last 20 years? Long obsolete models still command high prices on eBay, which indicates that the market doesn't seem to think so. Maybe it's the kind of work we do with them that hasn't evolved all that much?
I paid over $30,000 for a new 1 GHz LeCroy just over 20 years ago. They can be purchased on eBay now for under $1,000. I wouldn't say they are commanding high prices.
At the time, 1 GHz was the fastest scope LeCroy made. Now you can buy one that has 100 GHz bandwidth. I'd say that's a significant advancement.
So the high end is about 100x better? - why not the entry level.
-
I recently bought a working 1 GHz DDA-120 for $303 delivered with tax. Quite amazing.
Of course, as they get faster it goes up rapidly. The 110 GHz Keysight is well over a million dollars.
Have Fun!
Reg
That DDA-120 was quite a bargain!
How does something like that perform compared with a modern "affordable" model from A or B brands?
-
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350. How is that not an improvement? What would that have cost 20 years ago?
-
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350.
That's misleading as it needs be hacked.
-
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350.
That's misleading as it needs be hacked.
:wtf:
How the hell is that misleading? You can still buy the damn thing for the quoted price...
-
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350.
That's misleading as it needs be hacked.
:wtf:
How the hell is that misleading?
Show us a listing for a new 100 MHz 4ch DSO for $350. :popcorn:
-
Wow really? Snipping out the rest of my post to take it out of context? ::) You buy the $350 one and hack it. How is that possibly misleading??
Here's what I *actually* wrote:
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350.
That's misleading as it needs be hacked.
:wtf:
How the hell is that misleading? You can still buy the damn thing for the quoted price...
-
Wow really? Snipping out the rest of my post to take it out of context? ::) You buy the $350 one and hack it. How is that possibly misleading??
Again, portraying a 100 MHz 4ch scope is available for $350 IS misleading without mention of a hack.
That's what Fungus did.
So now do we all quote prices and BW of hacked scopes as standard conversation, I think not.
We go that road and $499 for 4ch 200 MHz, $619 for 2ch 350 MHz, $1400 for 4ch 500 MHz and $3500 for 1 GHz 4ch.
Best we stick with unhacked rated BW and retail prices to avoid all confusion.
-
Actually DSP can do the job. You can combine a modest analog filter followed by a digital filter to zero out the region in which the aliasing is present.
The keyword here is "modest", though, meaning there is still a minimum requirement. If we want to place the digital cut-off at (say) fs/4, then the analog filter must still provide a sufficient a priori attenuation at 3*fs/4 and beyond.
Anoter question is of course, how many dB to consider/define "suffcient"?
(Individiual requirements may vary)
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal. Because there needs to be some headroom for an anti aliasing filter most DSOs have a maximum bandwidth of fs/2.5 .
Best we stick with unhacked rated BW and retail prices to avoid all confusion.
Even I find that total nonsense. If you can get more bandwidth from a scope using a simple key generator then it has that extra bandwidth before you bought it.
-
Again off the rails...
Back to the topic. Reg KNOWS his DSP. I argue with him when he confuses seismography equipment with a 1 GHz (input bandwidth) scope.
Reg properly said that by choosing proper input filter you will get perfect impulse response.
He properly said that that filtering can be done partially in analog and partially in digital domain.
He properly said that using DSP you can "idealize" (shaping it to achieve proper calibrated response) scope pulse response.
He properly said that by using proper reconstruction filter you can fully reconstruct sinewave with more than 2,5 samples per period. Make note that reconstruction filter is not simply low pass filter.
He properly said that even if scope doesn't have Gaussian frequency response, by choosing reconstruction filter appropriate for it's response, you get the scope to reconstruct signal properly.
He properly said that FFT on most scopes is far from being optimal, not to mention easy to use, or giving directly usable results similar to SA.
He properly said that decimation by simple throwing away samples is wasteful. Data can be downsampled by filter and gain additional dynamic range and lower noise in usable bandwidth. That also filters out all higher frequency components (low pass filter) so it takes care of aliasing . If that is what you want to accomplish. (edit:that is important if want to further process data, if you simply show it on the screen it's not important)
Things he gets wrong are connected with not knowing how other people use scope, and generally having limited insight into oscilloscope use cases by industry.
Also we are getting back to people again saying " we need at least 10x oversampling". No you don't. If band limiting and reconstruction is done correctly, you don't. Because, if you do that correctly, there is nothing in the signal reaching the A/D converter in between those samples that doesn't fit perfectly on top of interpolated sine segment. On a 1 GHz scope, looking at a 1 GHz square wave, you have to see perfect sinewave on the screen.
Reason why Keysight itself used to recommend 5x oversampling (which I find good compromise) is that input filtering is not ideal...
On most 1GHz scopes and up nowadays they use brickwall (sharp rolloff) filtering and sampling factor of 2.5 with all channels on.
That brickwall filter takes care of aliasing. But pulse response of that filter is not the same as Gaussian response filter. It has pros and cons. Pulse response is worse, it overshoots. But, same 1GHz scope will have 15% more accurate rise/falltime measurements. Additionally, it's frequency response will be essentially flat, gaining excellent amplitude accuracy, all the way up to the cuttof. In practice, that means that on 900 MHz you get less than 5 % amplitude error, compared to almost 30% on Gaussian response scope.
Also that distorted pulse response? Well, it isn't there if you don't feed scope stuff over 1.2 GHz, in amplitude that is high enough to show on screen. That overshoot shows only when you feed it signal with components in gigahertz range. You need to buy (or make) special pulser to even be able to create pulse that fast, and connect pulser directly on scope input. Even a foot of coax cable will make things very different.
In practice that is pretty much not going to happen. There are no 40 ps pulses in Raspberry Pi, Arduino, switching and analog PSUs, and pretty much 90% of all electronics. Unless you are working on multigigabit data links, picosecond lasers, or designing MRI machines or such, no worry. And if you do, your employer bought you the fancy stuff. That is no small business or hobby territory.
People who work on 40ps pulses need 20GHz+ scopes and equipment. DUH.
Stuff is, of course, more serious when same things happens on a 100 MHz scope. That bandwidth is pretty much easily pushed into aliasing with anything on your desk.
So question here is how to make it to not alias. Again it is simple. Cheaper scopes are being made with 1x1GS/s A/D for 4 Ch (those that will get in trouble), and those that have 2x1GS/s (those that will be doing better). Also if aliasing and pulse response is more important than maximum bandwidth, you can always opt NOT to buy highest bandwith machine.
It is quite obvious really. You don't get Keysight MSOX3104T, but deliberately get MSOX3054T or even better MSOX3034T. Those will still sample at 5 GS/s but with maximum 350/500 MHz frequency are guaranteed not to alias. And step response will be perfect. Really, just buy Rigol DS1054Z and DON'T hack it to 100MHz. Take a look at that pulse response...It will be slow, but perfectly behaved.
-
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
The example fails to reconstruct the sampled signals exactly due to the following violations of the samling theorem:
1)
The reconstruction filter is not sufficient. The sampling theorem requires a reconstruction filter with a boxcar frequency response, which is equivalent to sin(x)/x interpolation. In the analog domain this is impossible to realize. Digitally it is well possible to up-sample to higher sampling rates with sinc interpolation (of course not to an infinite sampling rate, but in a DSO a finite resolution is still sufficient for the reconstructed waveform).
2)
The signals are not sine waves, but truncated (-> true sine waves had an infinite extent in the time domain). A sine wave modulated with a single boxcar impulse is no longer a band-limited signal, though, thus violating the sampling theorem.
Even if we consider the captured buffer periodic, assuming that it repeats over and over again, then the repetition of the buffer will indeed lead to an infintie (sampled) sine wave for the 100 kHz signal, but not for the 107 and 203 kHz signals, because there are discontinuities at the wrap around from the end of the buffer to the start of the buffer. An exact reconstruction of the infinite sine wave from a finite buffer is only possible if the buffer would contain an exact integral number of signal periods. If the captured buffer is large enough this can be addressed by truncating the sinc interpolation kernel with a (not too short) window function, which still leads to a reasonable approximate reconstruction then (at least at the center of the buffer).
Btw, the audio world also manages to recontruct a 0..20 kHz analog signal from 44 kSa/s CD data (with an accuracy inside the SNR limit which is dictated by the 16 bits anyway).
-
Best we stick with unhacked rated BW and retail prices to avoid all confusion.
Even I find that total nonsense. If you can get more bandwidth from a scope using a simple key generator then it has that extra bandwidth before you bought it.
The consideration is only valid, though, if the hacked frequency response is really the same as for the higher-priced regular models. This would either need to be verified in the first place, or proved otherwise (e.g. by reverse engineering) with sufficient confidence.
Otherwise the hacked one and the higher-priced regular one would rather need to be considered different "models".
-
Best we stick with unhacked rated BW and retail prices to avoid all confusion.
Even I find that total nonsense. If you can get more bandwidth from a scope using a simple key generator then it has that extra bandwidth before you bought it.
The consideration is only valid, though, if the hacked frequency response is really the same as for the higher-priced regular models. This would either need to be verified in the first place, or proved otherwise (e.g. by reverse engineering) with sufficient confidence.
Otherwise the hacked one and the higher-priced regular one would rather need to be considered different "models".
Nowadays the difference is in software only. The dead giveaway is that you can buy options in the form of license keys to increase the bandwidth.
-
Best we stick with unhacked rated BW and retail prices to avoid all confusion.
Alright, let's do that then.
How much would a 4-channel, 50MHz DSO have cost back then?
(It would have been sooooo easy for you to just say "100MHz is cheating but a 50MHz DSO would have cost XXXXX back then")
Nowadays the difference is in software only. The dead giveaway is that you can buy options in the form of license keys to increase the bandwidth.
Yep, and Rigol KNOWS we hack them and you can bet they're not losing money.
If you want to go to 2-channel DSOs you can a lot cheaper. Owon, Hantek sell 2-channel scopes for a lot less than $350. How much would they have cost back then?
Heck, you can even got one of these for under $150: https://www.eevblog.com/forum/testgear/fnirsi-1013d-100mhz-tablet-oscilloscope/ (https://www.eevblog.com/forum/testgear/fnirsi-1013d-100mhz-tablet-oscilloscope/)
That's probably a match for most 20Mhz analog CROs but it has batteries, touch screen and lots of measurement functions including FFT.
The claim that things aren't advancing in the o'scope world is ridiculous.
-
Let's keep the war at the technical level.
20 years ago the chinese weren't in the game so you can easily imagine how much the prices were inflated...
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal. Because there needs to be some headroom for an anti aliasing filter most DSOs have a maximum bandwidth of fs/2.5 .
Sure, there are ways to reconstruct a sinewave from 2.5 samples per period.
But I believe the signal has to be continuous and have no frequency content beyond fs/2
If you reconstruct a sinewave from a small number of data points, you are making the assumption that the signal was sinewave in the first place.
Regards,
Jay_Diddy_B
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal. Because there needs to be some headroom for an anti aliasing filter most DSOs have a maximum bandwidth of fs/2.5 .
Sure, there are ways to reconstruct a sinewave from 2.5 samples per period.
But I believe the signal has to be continuous and have no frequency content beyond fs/2
If you reconstruct a sinewave from a small number of data points, you are making the assumption that the signal was sinewave in the first place.
No. Just read more about sampling theory and Nyquist. The signal doesn't need to be continuous but there is a requirement that the original signal (before sampling) doesn't have frequency content above fs/2 which is what the anti-aliasing filter is for. See gf's example of digital audio where they push the bandwidth to fs/2.2 .
-
Well, a Pentium Pro CPU of 1995 managed a 150Mhz clock frequency. I would guess (but haven't done the math) that a modern processor is roughly 100x to 1000x better, given increased clock frequencies, higher IPC (instructions per clock), and high amounts of parallelism / multiple cores.
Processor speed is not what limits the performance of an oscilloscope.
Practically all DSOs are limited by processor speed, which is what leads to dead time between acquisitions even with double buffering. Producing a usable display from a multiple Gbyte per second stream of data takes massive bandwidth. The highest performance hardware manages it with heroic amounts of parallelism but nobody here is going to pay for that.
Display refresh rate has nothing to do with it because processing should accumulate every acquisition during each frame; slower frame rates just mean more acquisitions for each one.
I mentioned earlier designing a DSO taking processing limitations into account. The primary limitation is memory bandwidth for decimation or processing the acquisition record suggesting that on a modern high performance processor, record length will be limited by cache size for a given performance with higher level caches being significantly slower but allowing larger record lengths. Offhand I do not know of any DSOs which take advantage of this but I suspect some do without mentioning it.
-
Sure, there are ways to reconstruct a sinewave from 2.5 samples per period.
Or even less than that.
The problem is that the width of your reconstruction window (and the number of samples that need to be processed) approaches infinity as you head in that direction. 2.5 Is a practical limit in real life.
-
If you reconstruct a sinewave from a small number of data points, you are making the assumption that the signal was sinewave in the first place.
No, you're making the assumption that the signal is bandwidth limited to the Nyquist frequency.
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal. Because there needs to be some headroom for an anti aliasing filter most DSOs have a maximum bandwidth of fs/2.5 .
Sure, there are ways to reconstruct a sinewave from 2.5 samples per period.
But I believe the signal has to be continuous and have no frequency content beyond fs/2
If you reconstruct a sinewave from a small number of data points, you are making the assumption that the signal was sinewave in the first place.
Regards,
Jay_Diddy_B
You are absolutely correct. If you are looking at 100 MHz squarewave signal on 100 MHz scope you should get 100 MHz sinewave on screen.
Because your scope must not show 300 MHz and 500 MHZ and 700 MHz at any amplitude to be visible on screen. Built in filtering must take care of that that neither of those frequencies and sharp changes in signal ever reach A/D converter.
If you need to look at 100 MHz squarewave, you need scope with 1GHz bandwith, to see at least first 9 harmonics..
-
The signal doesn't need to be continuous but there is a requirement that the original signal (before sampling) doesn't have frequency content above fs/2
The problem in practice is that even a band-limited non-periodic signal still has an infinite extent, i.e. you'd need to capture an infinte number of samples in order to enable sinc reconstruction, which is not feasible.
The problem is that the width of your reconstruction window (and the number of samples that need to be processed) approaches infinity as you head in that direction. 2.5 Is a practical limit in real life.
It is in fact always infinite, since a sin(x)/x impulse has generally an infinite extent.
The only special case is when the captured signal is periodic and the captured buffer happens to contain an exact integral number of periods for each frequency contained in the signal. Then the (finite) buffer can be considered being repeated periodically, enabling e.g. exact sinc up-sampling via FFT.
In the general case, you always have to truncate the reconstruction window at some point in practice, enabling only an approximate reconstruction then.
-
In the general case, you always have to truncate the reconstruction window at some point in practice, enabling only an approximate reconstruction then.
I prefer the term "good enough".
-
Hi,
To change the direction of the thread a little, other considerations when using a scope:
1) What is the bandwidth at the probe tip?
2) Is the 50 \$\Omega\$ input really 50 \$\Omega\$?
2b) does it change with the attenuator settings?
2c) can you help a lot by putting a 6dB pad on the input?
I have measured the input impedance of my
DSOX 3034A
MDO4104
TDS3032
and there are some differences. :-BROKE
I don't have any 100MHz class scopes to measure.
Regards,
Jay_Diddy_B
-
Must admit chaps is thread still going :-DD my goodness can we not just settle for say six fundamental requisites that a novice could use to select a scope for their own specific purposes?
I mean I respect those that have the time to deep dive into this subject but do you really, really need to? Are the scopes these days that bad at all?
If I took into account every possible variant before choosing a scope, I would be left with a mental break down FFS.
Agree we do not wish to end up with the device for our own needs and yes some manufacturers are not so truthful with their figures, however figures only tell you so much the real world use is where the real point of decision is made imho.
More T&E companies should have more extensive loan stock for genuine potential purchasers of their products imho, some are really good other are 'Mey' as our esteemed leader would say.
The Keysight MXR demo has gone awfully quiet to.
-
The Keysight MXR demo has gone awfully quiet to.
LeCroy rulez?? ;)
-
Here's a 20 MHz, 200 mVpp square wave from my 33622A.
First the gospel according to the 485.
[attach=1]
Now the MSO-2204EA with normal sampling
[attach=2]
Then in peak detect mode. Notice that the aliases that fall below 200 MHz become visible in the FFT.
[attach=3]
So same thing with the DS1202Z-E in sample mode:
[attach=4]
And peak detect:
[attach=5]
I forgot to set the anti-alias filter to on, but turning it on/off has no affect. I'm using a triangle window.
I am forced to conclude that on the Instek "peak detect" turns off the anti-alias filter. I am unable to interpret the Rigol behavior. I'm using a 6k memory buffer for the FFT but unable to determine whether the spurious responses are internal noise or numerical artifacts. An off by one indexing error in an FFT produces artifacts that are very difficult to interpret unless you can do the FFT on a pure impulse and a constant.
If you have an analog filter which does no achieve -6 dB/bit at Nyquist, energy above Nyquist is folded back. If a digital filter that suppresses everything that gets through the anti-alias filter follows you can remove the aliased frequencies above the frequency where the aliased signals are less than -6 dB/bit. Naturally this leads to a steep skirt and lots of ringing as we see commonly.
I'm pretty well aware of what sorts of things people use a scope for, I've been using one for quite a long time and have repaired both a 60 MHz Dumont 1062 and a 100 MHz Tek 465 analog scope with major faults due to cracked solder joints. I'm also enumerating the full range of use cases, but that's a subject for another time.
What I'm doing here is performing tests which show the actual internal engineering of the DSOs. So I'm considering the various options for implementing a DSO and then making up tests to see what particular scopes are doing.
That's an iterative process. They are not all the same and so one has to devise new tests as information is gained.
The input signal to a scope is an analog waveform. The sina qua non is that the scope accurately display the analog waveform within the limitations of the instrument.
The Instek is clearly a better instrument than the DS1202Z-E. So the real question is will Rigol fix the bugs I find. However, at current prices the Instek is 2-3x the Rigol. You can hack a GDS-2072E to 200 MHz, but it's almost twice the cost of the DS1202Z-E. And if you buy the GDS-2202E it's over 3x the Rigol.
Most people neither need nor can afford a $10k+ instrument. For $300, the Rigol is not bad. I know how it compares to the Instek, so the remaining question is how does it compare to a Siglent in the same price class.
-
No, the BW in these scopes is determined by the band limiting amplifier IC in front of the ADC. All software does is sends commands to switch the band limits. Therefore in order to figure out a scope's front end filter shape all one needs to do is refer to that IC's datasheet.
I wish that were true, but experimentation shows otherwise. At 500MSa/S, my Siglent 1104X-E will alias a 300MHz or 400MHz signal quite clearly, in fact the aliased 400MHz signal has greater amplitude than the genuine 200MHz! :wtf:
However, if I shut off CH2, the 300MHz signal is completely suppressed. If the BW filters were all before the ADC, I don't see why that would happen. To me this is a semi-serious shortcoming and the lesson is that if you want a good 2-channel scope the 1104X-E is it.
The first 4 pictures are 1GSa/S and 100, 200, 300, 400MHz, then 500MSa/S and 200, 300, 400MHz. I appear to have missed the shot of 500MSa/S @ 100MHz, but it looked similar but with slightly greater amplitude.
[attachimg=1]
[attachimg=2]
[attachimg=3]
[attachimg=4]
[attachimg=5]
[attachimg=6]
[attachimg=7]
-
@rhb, you have two channels enabled on the Rigol, which will reduce the sample rate to 500MSa/s. Does it look any different with CH2 off?
-
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal.
As the audiophools like to say, Nyquist is just a theory. It isn't just your brain that struggles to make a signal from the dots. The few dots there are, the harder it is to reconstruct the signal accurately. Just look at the example he posted. Small amplitude errors can result in huge reconstruction errors at low sample counts while at fs/10 one slightly misplaced dot won't hurt too much. Yes, in theory you can reconstruct a sine wave at fs/2.1 with a 48-pole unity-gain filter or its digital counterpart, but in practice that becomes quite difficult. I can show you distortions on an actual A-brand scope that are similar to the fs/3.3 example, even though is well within Nyquist, like fs/6 or so. This particular scope is a bit older and has a sample rate 10X the BW. I guess they agreed that 10X was a good number.
And then, consider that Nyquist, AFAIK, doesn't cover triggering, an important part of oscilloscopy.
Sorry about the terrible picture, I had to use my phone to get the picture off the camera since I can't find my cable at the moment...
This is a 300MHz signal (the scope attenuated it quite a bit), it shows averaging in use because I forgot to take a picture without it, but it still had similar issues in the SAMPLE mode.
-
Must admit chaps is thread still going :-DD my goodness can we not just settle for say six fundamental requisites that a novice could use to select a scope for their own specific purposes?
I mean I respect those that have the time to deep dive into this subject but do you really, really need to?
[...]
Well... if you go to a track day with your car, you kinda have to expect people to talk a lot about cars... :D
-
To try to get some data about the anti-alias filter in the Instek I decided to really go up far past the rated BW and Nyquist.
Here we have the 8648C:
[attach=1]
Here's what the Instek shows:
[attach=2]
I have to turn up the gain on the scope to get it to trigger, but 700 MHz on a 200 MHz scope is rather interesting, though clearly there is somewhat of a sampling theory issue.
NB As you approach Nyquist the phase and amplitude resolution is reduced.
Have Fun!
Reg
-
Tv84
Well when I see one I will let you know ::)
-
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal.
As the audiophools like to say, Nyquist is just a theory. It isn't just your brain that struggles to make a signal from the dots. The few dots there are, the harder it is to reconstruct the signal accurately. Just look at the example he posted. Small amplitude errors can result in huge reconstruction errors at low sample counts while at fs/10 one slightly misplaced dot won't hurt too much. Yes, in theory you can reconstruct a sine wave at fs/2.1 with a 48-pole unity-
Sorry about the terrible picture, I had to use my phone to get the picture off the camera since I can't find my cable at the moment...
This is a 300MHz signal (the scope attenuated it quite a bit), it shows averaging in use because I forgot to take a picture without it, but it still had similar issues in the SAMPLE mode.
Well... from my own testing with oscilloscopes I've found that sin x / x reconstruction isn't always implemented correctly. I had to report this as bug to 2 different manufacturers (after which it got fixed). So it doesn't surprise me that there are oscilloscopes out there on which the sin x /x reconstruction doesn't work as goed as it can.
-
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350. How is that not an improvement? What would that have cost 20 years ago?
1989: HP54501A, 100MHz, 4 channels, GPIB, FFT - list price: $3,465 all inclusive (which is ~$7,200 in modern play money)
2020: DSOX3024A, 200MHz, 4 channels - list price: $5,174 (but software features are optional extras, at extra cost)
There are definitely improvements, but hardly mind blowing considering it is 30 years - and the price reduction is hardly mind blowing either.
Yes, I would rather have the 2020 model because it is a better scope in so many ways. But fundamentally, you could probably do most probing jobs with either model, if you had to... it's just more work with the old guy.
-
OK here is the DS1202Z-E sampling at 1 GSa/s.
First a 700 MHz input:
[attach=1]
Next an 850 MHz input:
[attach=2]
and finally a 900 MHz input:
[attach=3]
I'm not sure what to think about this yet. Rather clearly if you are working at UHF you can use a 200 MHz DSO quite successfully. Just be sure you understand what it is actually showing you as the labeling will be completely wrong.
NB: I have yet to see a DSO that did a minimum phase sinc(t) interpolation. So technically every scope I've seen does it wrong. Pretty sad in the $20K class.
Have Fun!
Reg
-
So the high end is about 100x better? - why not the entry level.
You can get a 4-channel, 100MHz DSO with lots of memory and features for under $350. How is that not an improvement? What would that have cost 20 years ago?
1989: HP54501A, 100MHz, 4 channels, GPIB, FFT - list price: $3,465 all inclusive (which is ~$7,200 in modern play money)
2020: DSOX3024A, 200MHz, 4 channels - list price: $5,174 (but software features are optional extras, at extra cost)
There are definitely improvements, but hardly mind blowing considering it is 30 years - and the price reduction is hardly mind blowing either.
Yes, I would rather have the 2020 model because it is a better scope in so many ways. But fundamentally, you could probably do most probing jobs with either model, if you had to... it's just more work with the old guy.
If push came to shove, you could probably still make do with an analog 'scope and a camera. Many of the great technological innovations of the 20th century were done with analog scopes. Heck, even the legend himself Jim Williams often used a loaded-with-vacuum-tubes Tek Type 556 (often with the 1S1 sampling plugin) till the end of his life. You can see scope photography from it in many of his Linear app notes.
The real advantages of modern DSOs are primarily in (1) the advanced triggering and protocol decoding stuff (2) single shot acquisitions (because the storage crts sucked when they were new, and they didn't get any better with age). I would wager a skilled operator with a good analog 'scope could do everything else.
-
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal.
As the audiophools like to say, Nyquist is just a theory. It isn't just your brain that struggles to make a signal from the dots. The few dots there are, the harder it is to reconstruct the signal accurately. Just look at the example he posted. Small amplitude errors can result in huge reconstruction errors at low sample counts while at fs/10 one slightly misplaced dot won't hurt too much. Yes, in theory you can reconstruct a sine wave at fs/2.1 with a 48-pole unity-
Sorry about the terrible picture, I had to use my phone to get the picture off the camera since I can't find my cable at the moment...
This is a 300MHz signal (the scope attenuated it quite a bit), it shows averaging in use because I forgot to take a picture without it, but it still had similar issues in the SAMPLE mode.
Well... from my own testing with oscilloscopes I've found that sin x / x reconstruction isn't always implemented correctly. I had to report this as bug to 2 different manufacturers (after which it got fixed). So it doesn't surprise me that there are oscilloscopes out there on which the sin x /x reconstruction doesn't work as goed as it can.
The big "problem" with Nyquist is that it assumes you are sampling the sine wave at the tops...
Try offsetting the sampler by Pi/2 and the entire signal disappears....
Then you increase the sample frequency slightly to make up for it... now you get aliasing instead, even though theoretically you are doing the right thing being over the Nyquist frequency...
Basically to be able to see fine phase changes in the signal, you need a lot more than the Nyquist sample rate, and/or dithering of the samples, and the rest of it... or so it seems to me! :D
-
I am unable to interpret the Rigol behavior. I'm using a 6k memory buffer for the FFT but unable to determine whether the spurious responses are internal noise or numerical artifacts.
The datasheet of the HMCAD1511 specifies SFDR as follows:
Spurious Free Dynamic range including interleaving spurs:
Single Ch Mode, Fs = 1000 MsPs: 49dBc
Dual Ch Mode, Fs = 500 MsPs: 44dBc
Quad Ch Mode, Fs = 250 MsPs: 57dB
As long as spurs don't exceed this amount significantly, I'd consider them within the specs.
If the raw data can be saved, they can be split into the samples from each interlaved ADC core (-> 8 cores @1GSPS, 4 cores @500MSPS and 2 cores @250MSPS), gain/offset adjustment for each core can be estimated (relative to the the first core), and then the adjusted samples can be re-combined. I tried this some time ago for the data from my scope, and it did improve spurs a little bit. Still not too much, IIRC.
EDIT: Just recognized that it was the 2CH model. Not sure if it is equipped with a HMCAD1511 too?
-
Well... from my own testing with oscilloscopes I've found that sin x / x reconstruction isn't always implemented correctly. I had to report this as bug to 2 different manufacturers (after which it got fixed). So it doesn't surprise me that there are oscilloscopes out there on which the sin x /x reconstruction doesn't work as goed as it can.
I don't actually know the cause of the errors in this case. Even if the interpolation implementation is perfect, the result is still more susceptible to amplitude and jitter errors if it has less samples. It would be not possible for me, with my equipment to verify complete accuracy of the reconstruction filter on even a 200MHz scope--group delay, step response and relative amplitude and so on. However, if I have 10X samples and my scope doesn't lie to me in the 'dots' mode, I can turn off the interpolation filter to verify what I'm seeing. So, at the low end of the scope market, I'll take more samples over clever interpolation.
-
1989: HP54501A, 100MHz, 4 channels, GPIB, FFT - list price: $3,465 all inclusive (which is ~$7,200 in modern play money)
2020: DSOX3024A, 200MHz, 4 channels - list price: $5,174 (but software features are optional extras, at extra cost)
There are definitely improvements, but hardly mind blowing considering it is 30 years - and the price reduction is hardly mind blowing either.
Not much difference? The newer scope gives you 200x improvement in bandwidth, 400x improvement in sampling rate, and a 4,000x improvement in memory:
single shot bandwidth: 1 MHz vs. 200 MHz
sampling rate: 10 MSa/s vs. 4 GSa/s
memory: 501 pts. vs. 2Mpts
-
The big "problem" with Nyquist is that it assumes you are sampling the sine wave at the tops...
Try offsetting the sampler by Pi/2 and the entire signal disappears....
No, Nyquist doesn't assume that. What you describe is what happens right AT Nyquist, when you actually need to stay below it.
Then you increase the sample frequency slightly to make up for it... now you get aliasing instead, even though theoretically you are doing the right thing being over the Nyquist frequency...
Basically to be able to see fine phase changes in the signal, you need a lot more than the Nyquist sample rate, and/or dithering of the samples, and the rest of it... or so it seems to me! :D
Frequencies over the Nyquist limit simply fold back. Frequencies just under the limit can theoretically be reconstructed perfectly, but this is difficult. 'Fine changes' would imply a higher bandwidth--and the sample rate required to properly capture and reconstruct them would be over twice the BW of the 'fine changes', which is not in any way the same as the periodic frequency of the signal.
-
Well... from my own testing with oscilloscopes I've found that sin x / x reconstruction isn't always implemented correctly. I had to report this as bug to 2 different manufacturers (after which it got fixed). So it doesn't surprise me that there are oscilloscopes out there on which the sin x /x reconstruction doesn't work as goed as it can.
I don't actually know the cause of the errors in this case. Even if the interpolation implementation is perfect, the result is still more susceptible to amplitude and jitter errors if it has less samples.
But those errors come from the signal so in the end you get what you feed into the oscilloscope. Jitter isn't a problem because the clock will need to have a low enough jitter to satisfy the ADC's requirements.
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal. Because there needs to be some headroom for an anti aliasing filter most DSOs have a maximum bandwidth of fs/2.5 .
Sure, there are ways to reconstruct a sinewave from 2.5 samples per period.
But I believe the signal has to be continuous and have no frequency content beyond fs/2
If you reconstruct a sinewave from a small number of data points, you are making the assumption that the signal was sinewave in the first place.
Regards,
Jay_Diddy_B
You are absolutely correct. If you are looking at 100 MHz squarewave signal on 100 MHz scope you should get 100 MHz sinewave on screen.
Because your scope must not show 300 MHz and 500 MHZ and 700 MHz at any amplitude to be visible on screen. Built in filtering must take care of that that neither of those frequencies and sharp changes in signal ever reach A/D converter.
If you need to look at 100 MHz squarewave, you need scope with 1GHz bandwith, to see at least first 9 harmonics..
Now, let's say we phase shifted the 100MHz square wave by 1 degree. What sample rate would it take to be able to accurately portray the difference between the original square wave and the one that we just phase shifted by one degree?
-
LoL!
Here's the DS1202Z-E with a completely absurd 1.25 GHz 0 dBm input from the 8648C. The Rigol timebase is not very stable, so I took a single shot. The 8648C has the high stability ovenized clock option.
[attach=1]
So can *you* figure out what the design error in the AFE of the Rigol and the Instek is?
Have Fun!
Reg
-
1989: HP54501A, 100MHz, 4 channels, GPIB, FFT - list price: $3,465 all inclusive (which is ~$7,200 in modern play money)
2020: DSOX3024A, 200MHz, 4 channels - list price: $5,174 (but software features are optional extras, at extra cost)
There are definitely improvements, but hardly mind blowing considering it is 30 years - and the price reduction is hardly mind blowing either.
Not much difference? The newer scope gives you 200x improvement in bandwidth, 400x improvement in sampling rate, and a 4,000x improvement in memory:
single shot bandwidth: 1 MHz vs. 200 MHz
sampling rate: 10 MSa/s vs. 4 GSa/s
memory: 501 pts. vs. 2Mpts
You are selectively looking at specific items that have improved the most, while ignoring the base spec of 4 channel/ 100MHz?
As @2N3055 perceptively posted earlier: "There are no 40 ps pulses in Raspberry Pi, Arduino, switching and analog PSUs, and pretty much 90% of all electronics. Unless you are working on multigigabit data links, picosecond lasers, or designing MRI machines or such, no worry. "
Could that be the real reason basic scope performance hasn't moved on so much... most people don't need it?
A bit like the reason a car from 1989 probably had about the same top speed as a 2020 car...
-
1989: HP54501A, 100MHz, 4 channels, GPIB, FFT - list price: $3,465 all inclusive (which is ~$7,200 in modern play money)
2020: DSOX3024A, 200MHz, 4 channels - list price: $5,174 (but software features are optional extras, at extra cost)
There are definitely improvements, but hardly mind blowing considering it is 30 years - and the price reduction is hardly mind blowing either.
Not much difference? The newer scope gives you 200x improvement in bandwidth, 400x improvement in sampling rate, and a 4,000x improvement in memory:
single shot bandwidth: 1 MHz vs. 200 MHz
sampling rate: 10 MSa/s vs. 4 GSa/s
memory: 501 pts. vs. 2Mpts
You are selectively looking at specific items that have improved the most, while ignoring the base spec of 4 channel/ 100MHz?
Single shot bandwidth, sampling rate, and memory depth are probably the most basic defining specifications of a general purpose oscilloscope.
The 1989 scope isn't even a true 4 channel scope, it's a "2+2" channel scope. So, you're really comparing a 2 channel, 1 MHz, 10 MSa/s scope to a 4 channel, 200 MHz, 4 GSa/s scope. Apples vs. oranges watermelons.
-
1989: HP54501A, 100MHz, 4 channels, GPIB, FFT - list price: $3,465 all inclusive (which is ~$7,200 in modern play money)
I had a HP 54501A as my first scope. 10MSa/s ADC (so 1MHz single shot B/W), 50 wfm/sec, 500 points memory. Also like a DOS terminal to use with single rotary encoder.
I can tell you it is a bitch to use... It was my main oscilloscope for about 4 years. It is hardly comparable to modern scopes.
Edit: others beat me to it...
-
The relevant questions are:
"What is the rise time of that GPIO on the Pi?"
"What do I need to measure it accurately?".
"Do I have problems with the impedance of the trace to the device at the other end?"
Embedded systems probably constitute the vast majority of electronic engineering tasks by a wide margin. An STM32Fxxx running at 120 MHz clock has some rather fast edges. They are not picosecond steps, but they are quick. Accurately probing and displaying those is not trivial. A Pi or Beagle is even harder.
Have Fun!
Reg
-
Now, let's say we phase shifted the 100MHz square wave by 1 degree. What sample rate would it take to be able to accurately portray the difference between the original square wave and the one that we just phase shifted by one degree?
The samplerate doesn't matter. But you'd need to resolve tens of ps in this scenario.
-
Now, let's say we phase shifted the 100MHz square wave by 1 degree. What sample rate would it take to be able to accurately portray the difference between the original square wave and the one that we just phase shifted by one degree?
The samplerate doesn't matter. But you'd need to resolve tens of ps in this scenario.
There are two factors: sample rate and bit depth.
You have to have enough resolution to resolve the phase difference. At 200 Msa/s you only have 180 degrees of phase sampling no matter what your bit depth. And you have no accurate amplitude information. However, you can resolve a 1 degree phase difference if you have 8 bit data, but not if you have 7 bit data.
At 400 MSa/s you have phase sampling of 90 degrees. That guarantees that you can accurately reconstruct the amplitude and phase to within the amplitude resolution of the data. With 8 bit data you have 90/256 degrees of phase resolution.
In between those extremes pick a sample rate and bit depth such that phase_angle/bit_depth = 1.
Have Fun!
Reg
-
Hi,
I am going to speculate that more fundamental issue is the ratio between bandwidth and the sampling rate.
The Nyquist Sampling theorem tell us the bandwidth is half the sampling frequency. Then you get folding and aliasing.
There is some modulation caused by the sampling frequency.
If you look at V(sample) waveform, You have to guess at the input waveform from the sampled data.
There isn't enough data to create the waveform with any degree of confidence.
A 100Mhz scope with 500Msps isn't enough data points.
Sorry, but that is the wrong conclusion. Nyquist says that all the information from a signal is there up the fs/2. It is just that your brain can't make a signal from the dots. But that is a problem in your brain and not in the number of samples. In order to help your brain DSOs have sin x/x reconstruction which connects the dots in a mathematically correct way to show you the signal. Because there needs to be some headroom for an anti aliasing filter most DSOs have a maximum bandwidth of fs/2.5 .
Sure, there are ways to reconstruct a sinewave from 2.5 samples per period.
But I believe the signal has to be continuous and have no frequency content beyond fs/2
If you reconstruct a sinewave from a small number of data points, you are making the assumption that the signal was sinewave in the first place.
Regards,
Jay_Diddy_B
You are absolutely correct. If you are looking at 100 MHz squarewave signal on 100 MHz scope you should get 100 MHz sinewave on screen.
Because your scope must not show 300 MHz and 500 MHZ and 700 MHz at any amplitude to be visible on screen. Built in filtering must take care of that that neither of those frequencies and sharp changes in signal ever reach A/D converter.
If you need to look at 100 MHz squarewave, you need scope with 1GHz bandwith, to see at least first 9 harmonics..
Now, let's say we phase shifted the 100MHz square wave by 1 degree. What sample rate would it take to be able to accurately portray the difference between the original square wave and the one that we just phase shifted by one degree?
You need 1GHz scope to see 9th harmonics of 100MHz squarewave. That would make decent approximation of squarewave, but still bumpy and not perfect. Sampling at 5 GS/s, at 200 ps intervals.
1% of phase shift according to what ? Scope have triggering system to synchronise to signal. Trigger point is synch point.
And that would mean shifting sample point 100ps. What do you expect to see? It won't make a squat of the difference on your 3.5 ns edges on your 100MHz squarewave. Because that is what edge will look like on 100 MHz scope even if you put 40ps edge in it.
On a 1GHZ scope it will have 300-400ps edge.
Every time this kind of topic pops up, I'm sad that in schools they don't plot math functions and solve graphical solutions by hand on millimeter paper and pencil anymore. That was great way to get good feeling for sense of scale.
Get piece of paper (or CAD program that works in scale if you wish) and plot 100MHz squarewave from horizontal points of 200 ps. And then from 400 ps points. That is respectively 5 and 2.5 GS/s. Than step away from plot and squint at it so it is roughly the size of scope screen form your perspective...
Writing this I think I start to understand where confusion is coming from. People who say that you need at least 10-20 points or more per period to reconstruct arbitrary shape signal are correct. But that is because to look at 100MHz sinewave you need 100 MHz scope (with more than 250 MHz sampling). In order to look at some arbitrary shaped signal repeating at 100 MHz, if you were to look that signal on SA, you would see, say, harmonics going to 2 GHz or more. So to look at that signal you need 2 GHz scope, although signal is periodic with repetition rate of 100 MHz, because that is signal with 2 GHz bandwidth.
Square wave consists of negative harmonics on quite simple formula, 3rd harmonic with 1/3 of amplitude, 5th harmonic with 1/5 of amplitude etc.. To infinity. Of course, very soon you don't need to go further, so in practice on scopes for visual representation, after you pass 11th or 13th harmonic you don't need to go further.. So that is your formula,. Repetition rate of your squarewave multiplied by 10-15. That is your scope bandwidth needed. And that at least 2.5 more will give sampling frequency.
So 100MHz squarewave, 1GHz scope, 2.5 GS/s sampling.
If you could afford 2GHz scope with 10 GS/s that would be better.
-
I will happily plot that 100Mhz SWF for you with my everyday lab scope 2.2Ghz and 10G/s sampling rate tomorrow I will produce the plot up to the 15th harmonic for you chaps
-
What does it take to measure a 1 degree phase shift of a 100 MHz sine wave?
That is *all* you need if you want to *measure* a 1 degree phase shift of a 100 MHz square wave.
Representing it correctly is an entirely other matter. And for that a DSO with over 1 GHz of BW and a clean step response are needed. So it needs to sample at 4 GSa/s or faster.
Have Fun!
Reg
BTW Could we please not have quotes nested 8-10 deep? Trim off the stuff not relevant to your point. My measurements have been buried under people arguing.
-
NB: I have yet to see a DSO that did a minimum phase sinc(t) interpolation. So technically every scope I've seen does it wrong.
Just a few thought - but I may be wrong:
Agreed that an anti-alias filter is basically supposed to be minimum phase (particularly if it is an analog one, of if it is supposed to mimic an analog one).
But why do you think that the interpolator (i.e. the reconstruction filter) needs to be minimum phase?
The aim of the reconstruction filter is to reconstruct exactly that (band-limited) signal waveform, which came out from the AA filter and entered the ADC.
Doesn't a zero phase sinc reconstruction filter do exactly that anyway? (w/o introducing additional delays, phase shifts or artifacts)
And would a minimum phase reconstruction filter with the same boxcar magnitude response actually reconstruct exactly the same waveform?
(If not, then it would not lead to an exact reconstruction of the (band-limited) signal which entered the ADC, and the reconstruction filter would miss its goal.)
I do not see why I should blame the interpolator for additional (pre)ringing which was not present in the output of the AA filter, if the actual root cause is the AA filter, which has done a bad job in the first place, not limiting the BW properly.
Please correct me if I missed anything.
-
BTW Could we please not have quotes nested 8-10 deep? Trim off the stuff not relevant to your point. My measurements have been buried under people arguing.
Funny.
-
So can *you* figure out what the design error in the AFE of the Rigol and the Instek is?
Unless they've left a big peak in the response, no. :)
In this one: https://www.eevblog.com/forum/testgear/scope-wars/msg3119138/#msg3119138 (https://www.eevblog.com/forum/testgear/scope-wars/msg3119138/#msg3119138)
Peak detect needs at last 2 ADC samples to work, to me it looks like the GW MSO-2204EA might be halving the displayed 500MSa/s rate to give the 2 peak samples in each X sample position. :popcorn:
-
So can *you* figure out what the design error in the AFE of the Rigol and the Instek is?
I'm not sure the error is in the AFE, although the fact that it isn't in the AFE might be the error. :-DD
-
It's pretty simple. The anti-alias filter is *before* the amplifier, rather than after it.
I guess you could call it a "feature" as it lets you see a 1.25 GHz signal on a 200 MHz scope.
Have Fun!
Reg
-
It's pretty simple. The anti-alias filter is *before* the amplifier, rather than after it.
I would have thought it would be 'part of' the amplifier. But don't these scopes depend heavily on the ADC chip for all sorts of amplification and gain? How much actual AFE is there?
I don't have a Rigol anymore, but on my Siglent which can be made to alias in the same way, albeit with significant (but not sensible) attentuation, but the behavior changes radically going from 500MSa/S to 1GSa/S by turning the paired channel off and on. At 1GSa/S, it behaves more or less properly, with nothing folding back as far up as 900MHz. My level sine wave capabilities stop at 990MHz.
Do any of your scopes change their response, say looking at a 400MHz signal which will be aliased back to 100MHz at 500MSa/S and displayed more or less properly at 1GSa/S, less attenuation?
-
In entry level scopes anti-aliasing filter is part of the variable gain bandwidth limiting amplifier IC. The ADC buffer also may have a rudimentary low pass filter on its output right before the ADC. Examples can be seen here
https://www.eevblog.com/forum/projects/project-yaigol-fixing-rigol-scope-design-problems/msg890966/#msg890966 (https://www.eevblog.com/forum/projects/project-yaigol-fixing-rigol-scope-design-problems/)
https://www.eevblog.com/forum/blog/eevblog-978-keysight-1000x-hacking/msg1233118/#msg1233118 (https://www.eevblog.com/forum/blog/eevblog-978-keysight-1000x-hacking/msg1233118/#msg1233118)
It can be seen there is no filtering before the VGA.
And there is no "amplification" in the ADC beside the stupid x2 magnification which simply stretches the signal vertically by displaying twice lower bits of resolution along the Y axis. All gain settings occurred in the front end before the ADC so the signal is scaled to the full ADC range before it hits the ADC.
-
The DS1202Z-E does what one would expect. The Instek also does, however, it sometimes gets into a weird state and displays completely bogus traces until you fiddle with the knobs and then it does what one would expect.
I ordered a pair of 935 MHz LoRa transceivers from Crowd supply today. I'll be very interested to see what level of signal resolution I can get on those.
Having cut my teeth on an analog scope, I'm used to being able to go well past the -3 dB corner by turning up the gain. But 1.25 GHz on a nominally 200 MHz DSO sort of blows my mind. The Instek doesn't provide as clean a trace as the Rigol at 1.25 GHz. Not yet sure why yet, though I suspect timebase jitter is the cause.
It would be really nice if someone built a low end DSO with the filter after the amplifier and circuitry to bypass the filter. I'm sure it would confuse the hell out of a novice, but it would be very useful and an insignificant increase in cost. Though one could use band pass filters to provide clean ETS operation with accurate displays of frequency to alleviate the confusion factor.
The major issue I have is that the user has almost no control over the parameters, especially for the FFT.
Once you've paid for the ADC, FPGA and display I look at a DSO and think, "It should be able to do everything." Obviously DACs, buffers and such needed, but the major HW is already there.
I wish the person who asserted that the amplifiers had BW filters would post an example part number.
Have Fun!
Reg
PS The filter is *before* the amplifier even if it's in the same chip. Putting the filter before the amplifier improves the noise spec, but without a filter *after* the amplifier lowers the system performance. You really need two.
-
And there is no "amplification" in the ADC beside the stupid x2 magnification which simply stretches the signal vertically by displaying twice lower bits of resolution along the Y axis. All gain settings occurred in the front end before the ADC so the signal is scaled to the full ADC range before it hits the ADC.
The HMCAD1511 ADC used in the Rigol DS1000Z series has 14-bit ADC chains internally. It can provide any magnification up to 6-bits without reduction in sample rate or bit depth (as the output is 8-bit truncated), if I understand the datasheet correctly.
I've yet to find a VGA amp, besides one Texas part, that has proper tuned analog filters on the input. The Texas part only provided a 20MHz LP filter for the B/W limit function of most scopes. I would expect that to be something that could be done with the right DSP.
-
Tektronix makes "Mixed Domain Oscilloscopes" that have real spectrum analyzers built into them, but the spectrum analyzer uses it's own hardware. Why? Because the hardware in a standard digital oscilloscope is not well suited for doing spectral analysis. Putting the FFT from a normal digital oscilloscope under a microscope strikes me as a waste of time -- it will always be compromised because the hardware is designed for an oscilloscope, not a spectrum analyzer.
-
The DS1202Z-E does what one would expect. The Instek also does, however, it sometimes gets into a weird state and displays completely bogus traces until you fiddle with the knobs and then
I wish the person who asserted that the amplifiers had BW filters would post an example part number.
Follow his links. Photos, part numbers, diagrams, everything is there in quite a lot of detail.
I don't know how the filters are implemented in the chip, but they clearly aren't working the way I would 'expect'. The data sheet for the LHM6518 shows selectable bandwidth filters of 20, 100, 200, 350, 650 and 750MHz. That explains a lot in terms of available low-end DSO types (except 70 is missing...) ::)
So assuming your DS1202Z-E is set to 200, what should the attenuation at 1.25GHz be? And what do you actually see?
-
The real advantages of modern DSOs are primarily in (1) the advanced triggering and protocol decoding stuff (2) single shot acquisitions (because the storage crts sucked when they were new, and they didn't get any better with age). I would wager a skilled operator with a good analog 'scope could do everything else.
Modern DSOs have many advantages over old ones but old DSOs work fine for single shot acquisitions in most applications as the photo I posted earlier shows.
Not much difference? The newer scope gives you 200x improvement in bandwidth, 400x improvement in sampling rate, and a 4,000x improvement in memory:
single shot bandwidth: 1 MHz vs. 200 MHz
sampling rate: 10 MSa/s vs. 4 GSa/s
memory: 501 pts. vs. 2Mpts
It is not quite so stark and the HP54501A was never intended for real time sampling which is reflected in its name, digitizing oscilloscope; they had other DSOs at the time with high single shot sample rates. They were not thinking in this way then but what the HP54501A does have is 10 Gsample/second equivalent time sampling which should be obvious because otherwise how could a 100 MHz bandwidth combined with 10 Msample/second sample rate make any sense?
It also lacks peak detection because its intended applications would never need it. It is not the type of DSO I would expect to use for general purpose work.
In entry level scopes anti-aliasing filter is part of the variable gain bandwidth limiting amplifier IC. The ADC buffer also may have a rudimentary low pass filter on its output right before the ADC. Examples can be seen here
That sure is not how the Rigol DS1000Z series was designed. They implement discrete switched bandwidth limiting immediately after the discrete preamplifier, and it is not for anti-aliasing.
Like I said earlier, if the filter was for anti-aliasing, then shouldn't it track the decimated sample rate? Doing this would have all kinds of negative consequences for how the DSO behaves.
-
So assuming your DS1202Z-E is set to 200, what should the attenuation at 1.25GHz be? And what do you actually see?
The output of the 8648C was 0 dBm into a 50 ohm thru so you should be able to work out the attenuation from the photo. It's rather severe. I'm just amazed I could see anything.
Have Fun!
Reg
I never saw a link or part number until tom66 posted one.
-
Tektronix makes "Mixed Domain Oscilloscopes" that have real spectrum analyzers built into them, but the spectrum analyzer uses it's own hardware. Why? Because the hardware in a standard digital oscilloscope is not well suited for doing spectral analysis. Putting the FFT from a normal digital oscilloscope under a microscope strikes me as a waste of time -- it will always be compromised because the hardware is designed for an oscilloscope, not a spectrum analyzer.
You clearly do not understand DSP. The Tektronix MDO lines have separate HW for SA because they don't want to spend the money on the time domain acquisition HW required to achieve the same level of performance. They also cost 10-100 times more money than the DSOs I'm testing in this thread.
I have been doing SA on non-stationary data for 35 years. That is pure DSP territory. So if an SA is doing it, it is DSP. All "real time" SAs are using DSP. So far as I know there are *no* SAs capable of analyzing non-stationary data as it's not germane to electronics except in very exceptional cases.
Have Fun!
Reg
-
And there is no "amplification" in the ADC beside the stupid x2 magnification which simply stretches the signal vertically by displaying twice lower bits of resolution along the Y axis. All gain settings occurred in the front end before the ADC so the signal is scaled to the full ADC range before it hits the ADC.
The HMCAD1511 ADC used in the Rigol DS1000Z series has 14-bit ADC chains internally. It can provide any magnification up to 6-bits without reduction in sample rate or bit depth (as the output is 8-bit truncated), if I understand the datasheet correctly.
I've yet to find a VGA amp, besides one Texas part, that has proper tuned analog filters on the input. The Texas part only provided a 20MHz LP filter for the B/W limit function of most scopes. I would expect that to be something that could be done with the right DSP.
LMH6518
" The following LMH6518 functions are controlled using the SPI-1 compatible bus:
• Filters (20, 100, 200, 350, 650, 750 MHz or full bandwidth) "
-
Thanks. The block diagram confirms my interpretation of the experiment. I look forward to studying the datasheet.
Have Fun!
Reg
-
Tektronix makes "Mixed Domain Oscilloscopes" that have real spectrum analyzers built into them, but the spectrum analyzer uses it's own hardware. Why? Because the hardware in a standard digital oscilloscope is not well suited for doing spectral analysis. Putting the FFT from a normal digital oscilloscope under a microscope strikes me as a waste of time -- it will always be compromised because the hardware is designed for an oscilloscope, not a spectrum analyzer.
The Tektronix MDO lines have separate HW for SA because they don't want to spend the money on the time domain acquisition HW required to achieve the same level of performance.
No, they use separate hardware for the spectrum analyzer because oscilloscope hardware is not optimized for spectral analysis. If you want the best performance, you use hardware designed to do the intended application. FFT capability on an oscilloscope is convenient, but the best oscilloscope cannot match the performance of the most basic spectrum analyzers and dynamic signal analyzers. That's why nit-picking the FFT performance of an oscilloscope is kind of silly exercise.
I have been doing SA on non-stationary data for 35 years.
Yes, we know. You mention it in almost every thread.
-
As I stated earlier, you don't understand DSP.
Have fun!
Reg
-
As I stated earlier, you don't understand DSP.
Oh, please.
-
I am saying that applying a filter before decimation to prevent aliasing due to the lower sample rate after decimation is worse than not filtering at all because it corrupts the histogram of the input signal. That is what an anti-aliasing filter would have to do.
So as I read along and try to keep up, sometimes I just don't get it. Where is the 'histogram' you mention? Is it an inherent property of the signal or are you assuming some transform has been done and it stored somewhere? I thought I knew what a histogram was, but I'm not sure what you mean here by 'corrupts the histogram'. As far as applying a digital filter before decimation, are you saying this is bad in general, bad for constructing a time-domain display (i.e. the oscilloscope trace) or for some other purpose, such as FFT? This is done all the time in audio, it's just called oversampling. When you record audio intended for 44kSa/S, you need to apply a brick wall filter, like 96db/octave to the input, which is commonly done by recording at 192kSa/S and then applying the brick wall digitally, then downsampling. I can't see what harm this would do in the context of an oscilloscope display as long as it was done consistently and there was no expectation of a response beyond the bandwidth of the filter.
Like I said earlier, if the filter was for anti-aliasing, then shouldn't it track the decimated sample rate? Doing this would have all kinds of negative consequences for how the DSO behaves.
There typically isn't a decimated sample rate on these scopes unless you set the timebase so slow that memory is limited--in which case you have to decide which path to take--or in the case of Rigol, more likely just ignore the issue. I don't have a DS1054Z to take apart, but on scopes that have the LMH6518, which has the BW filter in between amp stages, I think that it is intended to be the anti-aliasing filter and it just doesnt' work well enough. Looking at the datasheet, it appears to have approximately a 6db/octave rolloff, which makes sense for a simple RC circuit. In each case, the advertised BW of the scope matches the 2.5X minimum full-speed sample rate (or better) and again, the only time the sample rate slows is at low sweep rates where I guess they just give up in budget scope-land. So theoretically, most of the time, the BW filter gives you maybe -10db at Nyquist--quite consistently. It just isn't good enough. There isn't any other anti-aliasing that I know of unless they implement a digital filter when they do decimate at slow sweeps. What other negative consequences do you refer to?
Now my Siglent 1104X-E (100MHz, unhacked) is another story. It strongly aliases a 400MHz signal at 500MSa/S (both paired channels running) but completely suppresses it at 1GSa/S (one channel) indicating that something is changing between sample rates. If it were a front end issue, I'd expect to see it display the 400MHz at least as strongly as it aliases it.
-
Now my Siglent 1104X-E (100MHz, unhacked) is another story. It strongly aliases a 400MHz signal at 500MSa/S (both paired channels running) but completely suppresses it at 1GSa/S (one channel) indicating that something is changing between sample rates. If it were a front end issue, I'd expect to see it display the 400MHz at least as strongly as it aliases it.
Are you sure... is it possible your mind have dropped to one very common and easy trap...
Take a mm or other paper, draw 400MHz signal samples to paper using 2ns sample interval... draw as many times as need and also draw original signal shape there and then alias... think. And think again. After you find it... 1st Nyquist-Shannon lesson done. may I help... These our 400MHz signal samples are there, so it display your 400MHs perfectly same level as its display alias, think why you see alias instead of original input. Alias is nothing but original samples from original signal. Turn Sinc off...
Also note things you see in this image. Think twise... and this is unmod SDS1104X-E.
Btw, if you continue your pen and paper exercise yoy meet also nearly exactly one datail you find there in this image...
And last note. Only SDS1x04X-E hardware model Siglent make for export is SDS1204X-E afaicg.
(https://www.eevblog.com/forum/testgear/siglent-sds1204x-e-released-for-domestic-markets-in-china/?action=dlattach;attach=465945;image)
-
You clearly do not understand DSP. The Tektronix MDO lines have separate HW for SA because they don't want to spend the money on the time domain acquisition HW required to achieve the same level of performance. They also cost 10-100 times more money than the DSOs I'm testing in this thread.
The time domain acquisition HW required to achieve the same level of performance also does not exist.
Oscilloscope signal chains have extra requirements which are incompatible with spectrum analyzer signal chains including thermal balance and DC coupling which would compromise the performance of a *simpler* signal chain for RF spectrum analysis. Both are different from the signal chains required for audio and other high resolution low frequency instruments which demand very high linearity produced with shunt feedback (oscilloscope vertical amplifiers rarely use shunt feedback) and low flicker noise if extended to low frequencies. Oscilloscopes and spectrum analyzers typically have terrible flicker noise but that is what external low noise amplifiers are for.
I am saying that applying a filter before decimation to prevent aliasing due to the lower sample rate after decimation is worse than not filtering at all because it corrupts the histogram of the input signal. That is what an anti-aliasing filter would have to do.
So as I read along and try to keep up, sometimes I just don't get it. Where is the 'histogram' you mention? Is it an inherent property of the signal or are you assuming some transform has been done and it stored somewhere? I thought I knew what a histogram was, but I'm not sure what you mean here by 'corrupts the histogram'.
Whether it is explicitly identified or not, there is at least one histogram produced from the acquisition record and often two. One is the display record which you see and the other if present is what measurements use. In the case of DSOs which perform measurements on the display record, it serves for both.
What you see on an analog oscilloscope is also a histogram, graded in intensity by the gamma of the CRT which is oddly useful as it allows analog oscilloscopes to make accurate RMS noise measurements. On old DSOs the display histogram may be only 1 bit deep like in the example I gave earlier while measurements are made on a separate deep histogram.
As far as applying a digital filter before decimation, are you saying this is bad in general, bad for constructing a time-domain display (i.e. the oscilloscope trace) or for some other purpose, such as FFT?
It is bad for time domain displays and measurements. For displays, it will change the shape of the displayed waveform in non-linear ways which vary depending on the decimated sample rate. The same applies to measurements which is even worse; which measurement at what time/div (and record length) is the most accurate, or even correct?
I consider DSO FFTs to be so broken that they are gimmicks for marketing so I am not sure anti-aliasing matters. Where is the noise marker function? How are they calibrated for spot noise? Where are the phase results?
This is done all the time in audio, it's just called oversampling. When you record audio intended for 44kSa/S, you need to apply a brick wall filter, like 96db/octave to the input, which is commonly done by recording at 192kSa/S and then applying the brick wall digitally, then downsampling. I can't see what harm this would do in the context of an oscilloscope display as long as it was done consistently and there was no expectation of a response beyond the bandwidth of the filter.
That might be acceptable however it would mean sacrificing significant performance or economy for little gain. It is difficult enough now to handle the data rates from existing oscilloscope digitizers which have data rates at least 3 order of magnitude higher and only the highest end DSOs are that capable.
And I am not sure anything would be gained. Modern DSOs are function rich on the software side of things but they are not particularly more effective at displaying and making the same measurements which DSOs handled 30+ years ago, and I would argue that they are worse in many respect but admittedly they cost less. (1) So what deficiency does oversampling and filtering correct? If I could oversample, then why wouldn't I just make that the new higher maximum sample rate?
Like I said earlier, if the filter was for anti-aliasing, then shouldn't it track the decimated sample rate? Doing this would have all kinds of negative consequences for how the DSO behaves.
There typically isn't a decimated sample rate on these scopes unless you set the timebase so slow that memory is limited--in which case you have to decide which path to take--or in the case of Rigol, more likely just ignore the issue.
I agree that long record lengths cure many ills but they are not a universal solution and they limit performance. Somewhere that data has to be processed to be useful and it slows things down.
I don't have a DS1054Z to take apart, but on scopes that have the LMH6518, which has the BW filter in between amp stages, I think that it is intended to be the anti-aliasing filter and it just doesnt' work well enough. Looking at the datasheet, it appears to have approximately a 6db/octave rolloff, which makes sense for a simple RC circuit. In each case, the advertised BW of the scope matches the 2.5X minimum full-speed sample rate (or better) and again, the only time the sample rate slows is at low sweep rates where I guess they just give up in budget scope-land. So theoretically, most of the time, the BW filter gives you maybe -10db at Nyquist--quite consistently. It just isn't good enough. There isn't any other anti-aliasing that I know of unless they implement a digital filter when they do decimate at slow sweeps. What other negative consequences do you refer to?
That filter was never intended to be an anti-aliasing filter. It allows one part to be sold for multiple applications. It also allows for easy market segmentation. On the practical side, it limits excess noise and provides for a predictable transition band.
DSOs like the DS1000Z series implement the bandwidth filters early in the signal chain for 50 and 70 MHz which when combined make 20 MHz. That never seems right to me but I have done the math based on the values twice.
(1) As has been pointed out occasionally, this is not really fair. Those old DSOs are comparable to modern DSOs costing thousands of dollars and not modern low cost DSOs which provide a very high capability for their price.
-
If you have a 400 MHz signal and a 250 MHz Nyquist, the alias is at 100 MHz. If you have a 500 MHz Nyquist there is no alias.
The alias is at 2*Nyquist - frequency.
Anything that can be done in the analog domain can be done in the digital domain. Depending upon the required performance one or the other may be cheaper. At the extremes it gets ambiguous as to what can be done with either. I don't know of any analog scopes faster than the 1 GHz Tek 7104. There are plenty of DSOs which are faster.
State an SA performance specification that you believe cannot be met by time domain acquisition: center frequency, span, dynamic range, RBW and sweep time.
Have Fun!
Reg
-
Are you sure... is it possible your mind have dropped to one very common and easy trap...
I'm still trying to understand your message, but as to if I'm sure...look at the pictures in my post.
https://www.eevblog.com/forum/testgear/scope-wars/msg3119156/#msg3119156 (https://www.eevblog.com/forum/testgear/scope-wars/msg3119156/#msg3119156)
The 4th and last pictures are of the same input signal at all the same settings except the sample rate changed because I turned CH2 on.
I think it is obvious what is happening. The scope is using post-acquisition digital low pass filtering in addition to the puny 6db/octave pre-acquisition analog BW filter. The 200MHz and 300MHz-folded-back-to 200MHz are attenuated some, while the 400MHz and up are completely blocked by this filter. However, at the lower sampling rate, the 400MHz is folded back to 100MHz, where it can't be attenuated because the digital filter can't tell the difference. It is also possible that this digital filter doesn't work at the lower sampling rate, or works differently. The 400 MHz is still attenuated by the analog filter, but only by 10-12db or so.
Anyone have a different idea?
-
If you have a 400 MHz signal and a 250 MHz Nyquist, the alias is at 100 MHz. If you have a 500 MHz Nyquist there is no alias.
The alias is at 2*Nyquist - frequency.
With a 500 MHz signal the alias is at zero frequency. In seismic applications you don't care about DC, but for oscilloscopes DC performance is absolutely vital.
-
Are you sure... is it possible your mind have dropped to one very common and easy trap...
I'm still trying to understand your message, but as to if I'm sure...look at the pictures in my post.
https://www.eevblog.com/forum/testgear/scope-wars/msg3119156/#msg3119156 (https://www.eevblog.com/forum/testgear/scope-wars/msg3119156/#msg3119156)
The 4th and last pictures are of the same input signal at all the same settings except the sample rate changed because I turned CH2 on.
I think it is obvious what is happening. The scope is using post-acquisition digital low pass filtering in addition to the puny 6db/octave pre-acquisition analog BW filter. The 200MHz and 300MHz-folded-back-to 200MHz are attenuated some, while the 400MHz and up are completely blocked by this filter. However, at the lower sampling rate, the 400MHz is folded back to 100MHz, where it can't be attenuated because the digital filter can't tell the difference. It is also possible that this digital filter doesn't work at the lower sampling rate, or works differently. The 400 MHz is still attenuated by the analog filter, but only by 10-12db or so.
Anyone have a different idea?
If you looked my old test image it is told there, just as your tests with 200, 300 and 400MHz. It is all there. You can see attenuation there - roughly for different frequencies, also over fNyq (note x and y scales) with 1Gsa/s and 500Msa/s (with sidenote: every scope and test setup is some amount different). Of course real high grade test need do using leveling just in scope terminal. But no one is doing now rocket science... after this scope have published it have been quite clear there is post acquisition DSP for make cheap 100MHz version. Turn it to 200MHz model what it really is and it change things.
As I have previously told this aliasing is not so big problem in normal use with probes and usual DUT things. BUT use need know and understand it. User need know limits and how to ge more reliable results and how to avoid aliasing. Even hammer need some knowledge how to use it for better results and for avoid bad effects.
It is not so easy to get high amount over 500MHz to scope using probes with usual DUTs. 2 channel can keep 1Gsa/s. If channels selected so that 500Msa/s is max, it need teach to users that it produce aliases if there is over 250MHz enough strong frequency components going to ADC isput. It is made for 100 or 200MHz BW and just do not use it for 400MHz. So simple. If you usuallu use it in situations where your signals are far over what your scope can handle you have wrong tool.
Of course if it is designed more carefully and there is good LPF before ADC it is better. Example bit over 200MHz and 400MHz brick wall filters or others but...then some peoples start barking about risetime... and so on. Oscilloscope is compromice and all you see on screen is sum of errors and lies.
My opinion have been and is and stay: Only right place for anti alias filter for reduce ADC true sampling speed aliasing is before ADC. And this is analog side. It is only place! Period. Why. Because digital side can not differentiate alias from true, both looks exactly same. If we make down conversion using perfect ADC no one can see this product is down conversion. Sample do not have any other information but its value and known sample interval. Totally different thing is after ADC digital side where can handle decimated samplerates aliasing if we do not directly throw away samples what have information. So it is true that stupid simple decimation is stupid but simple and cheap.. This is totally other story.
-
State an SA performance specification that you believe cannot be met by time domain acquisition: center frequency, span, dynamic range, RBW and sweep time.
Many modern SAs obviously do time domain acquisition (particularly real-time ones). Frequently they are still hybrids, not sampling the RF directly, but sampling the IF band after analog down-mixing, and calulating the whole IF spectrum at once via FFT which is faster than a narrow-bandwidth sweep.
-
I wanted to add few more observation, looking at discussion here.
Scope can have two reasons for aliasing:
- first, when running at fast timebases and fastest sampling speed it can alias because analog antialias filter in front end before A/D doesn't filter properly, and frequencies above Nyquist criteria reach A/D converter.
This happens when you connect 1 GHz signal to scope with 200 MHz rated bandwidth and push amplitude of generator up more than input attenuator is set for until something is show on a screen. If analog antialiasing filter is done right, when you connect out of band 100mV P-P signal on scope input set at 100mv full screen sensitivity (that would be V/div time vertical divisions) you shouldn't see any appreciable signal. If you do, analog filter is not steep enough. Mind you, you can always make something show, if you push 10V P-P into 1mV/div input. So that test has to always be with exactly the same amplitude as in band signal that shows correctly.
- second, when scope is running slow timebases, because of limited memory, it has to slow down effective sampling rate of A/D, otherwise it won't be able to capture such a long capture.
It can do that that by:
1. Literally use slower clock for A/D converter. That is not always possible for various reasons.
2. It keeps on sampling at maximum sample rate as always, but decimates by throwing away samples, and keeping only every 4th or 5th or 10th one. That is effectively the same as slowing down sampling clock for the same factor.
3. It keeps on sampling at maximum sample rate as always, but downsamples by filtering. This will produce same number of samples as first two, by lowpass filtering.
This, slow timebase induced one, is tricky one, and more often to happen than first occasion of aliasing.
Here , option at 1. is rarely used. Many reasons for it, including A/D converters designed for fixed clocks or limited frequency range, complications in datapump design etc etc.
Option 2. is mostly used. It's cheap and very easy to do. It is also compatible with Peak detect mode, that can run at the same time. This is what most scopes do.
Problem with option 1. and 2. is that as effective sample rate goes down, so does Nyquist criteria frequency. But at the same time we keep input analog antialiasing filter the same, breaking Nyquist criteria. Hello aliasing. So instead of seeing solid block of signal (because periods are so close together they actually blend together horizontaly, keeping vertical amplitude size, showing signal envelope a that time scale) it starts showing downconverted imaginary signal.
Option 3. doesn't have that problem. Since it is averaging with low pass filter it applies to sampled signal a digital antialiasing filter that filters out anything above the Nyquist at target sample rate. By virtue of that, it will pass through only signals that can be properly reconstructed, and nothing more. And being digital, you can make it pretty ideal if you want.
So option 3. is best one. It automatically creates signal that is free of aliasing errors, right? And that is what you wan't to do?
Well, not so fast.
I will demonstrate on Keysight 3000T. Make sure you get stable trigger and use holdoff if needed. Scope might show it uses FFT. I do that to force it to disable built in antialiasing algorithm.
Let's take that simple 50 MHz signal AM modulated with 100Hz as an example. That is carrier with period of 20ns (1/f), and modulation with period of 10ms (1/f)
Let's put scope at 10 ns/div. We want to see the carrier. Scope samples at 5 GS/s.
If you have scope that does 1. or 2. or 3. you get nice 50 MHz sinewave on screen.
[attachimg=1]
Let's put scope at 10 ms/div. We want to see the modulation envelope. Scope samples at 20 MS/s.
If you have scope that does 1. or 2. , you get this:
[attachimg=2]
That is wrong right?
Now you might be tempted to say: let's use option 3. It will fix this problem.
No, it would be right thing to do for spectrum analyser, not scope. Why?
Here, this is what we want to see:
[attachimg=3]
It is nice amplitude envelope at 100Hz and a solid block of 50 Mhz carrier inside.
But if we go with option 3. (downsampling by filtering), as suggested, we will filter out anything above 10 MHz..
On spectrum analyser, we would be looking at scale of 0-10 MHz and see nice clean spectra of that without any folding from upper bandwidth. Perfect.
What we will see on the screen of oscilloscope ?
-
State an SA performance specification that you believe cannot be met by time domain acquisition: center frequency, span, dynamic range, RBW and sweep time.
Many modern SAs obviously do time domain acquisition (particularly real-time ones). Frequently they are still hybrids, not sampling the RF directly, but sampling the IF band after analog down-mixing, and calulating the whole IF spectrum at once via FFT which is faster than a narrow-bandwidth sweep.
RT SA are mostly hybrid design that direct convert (have real time bandwidth of) 20-100 MHz bandwidth . In attachment SH BB60C manual with operating principle explained.
-
I wanted to add few more observation, looking at discussion here.
But if we go with option 3. (downsampling by filtering), as suggested, we will filter out anything above 10 MHz..
On spectrum analyser, we would be looking at scale of 0-10 MHz and see nice clean spectra of that without any folding from upper bandwidth. Perfect.
What we will see on the screen of oscilloscope ?
It's all wrong. Look at your samplerates and then think about how that aliases with your 50MHz carrier. If you do these kind of tests you need to use an 'odd' frequency like 33MHz. You'll see you can only get a good picture with lower samplerates when you turn peak-detect on.
As mentioned before: if you low-pass filter a signal you'll lose the high frequency content. Think about how a basic AM demodulator works; that is not a low-pass filter but an envelope detector. Do the same test with 20MHz bandwidth on and off and you'll see.
[attachimg=1]
This is an AM modulated signal using an 8.33MHz carrier and 100Hz modulation. Peak detect is on. Trace 1 (top) is unfiltered, trace 3 (bottom) is filtered using a 600kHz low-pass filter. I hope that this makes it very clear that anti-aliasing filtering based on the samplerate is a really bad idea on an oscilloscope.
-
I wanted to add few more observation, looking at discussion here.
But if we go with option 3. (downsampling by filtering), as suggested, we will filter out anything above 10 MHz..
On spectrum analyser, we would be looking at scale of 0-10 MHz and see nice clean spectra of that without any folding from upper bandwidth. Perfect.
What we will see on the screen of oscilloscope ?
It's all wrong. Look at your samplerates and then think about how that aliases with your 50MHz carrier. If you do these kind of tests you need to use an 'odd' frequency like 33MHz. You'll see you can only get a good picture with lower samplerates when you turn peak-detect on.
As mentioned before: if you low-pass filter a signal you'll lose the high frequency content. Think about how a basic AM demodulator works; that is not a low-pass filter but an envelope detector. Do the same test with 20MHz bandwidth on and off and you'll see.
I honestly don't understand what are you sayin. I'm DEMONSTRATING the problem by deliberately provoking aliasing... To show how it looks..
Third picture IS how it SHOULD look. And Keysight seem to have some intelligent detector that senses aliasing and uses something like Peak detect mode to show proper waveform. Funny enough it stops working if you enable FFT, probably to be able to provide proper samples for FFT engine.
And I know how it will look after filter. It is a question for a reader. You know, discussion..
-
After reading your post 3 more times I see what you are trying to show; it is extremely hard to follow what you wrote because you didn't explain what is in your screendumps and how your oscilloscope behaves.
-
After reading your post 3 more times I see what you are trying to show; it is extremely hard to follow what you wrote because you didn't explain what is in your screendumps and how your oscilloscope behaves.
Thank you for pointing that out. I will try to explain better next time. Sorry!
-
(Attachment Link)
This is an AM modulated signal using an 8.33MHz carrier and 100Hz modulation. Peak detect is on. Trace 1 (top) is unfiltered, trace 3 (bottom) is filtered using a 600kHz low-pass filter. I hope that this makes it very clear that anti-aliasing filtering based on the samplerate is a really bad idea on an oscilloscope.
Correct!! If you do downsampling by filtering, in time domain plot you will see absolutely nothing once your timebase gets long enough. That is worse than aliasing. With aliasing you at least get a notion there is SOME signal. And as you correctly point out, enabling Peak detect mode (or if the scope has functioning antialiasing function), you will get correct display, similar to what you would see on the analog scope.
What is good for scope is not optimum for SA and vice versa.
-
Afternoon Chaps
As promised the screen images of a 100Mhz square wave generated by a Tek 31252 @2G/s and 14 bits
Using T-Flex 405 18Ghz BW respectable lab cables with SS matching BW terminations, scope calibration performed yesterday
FG generator output set to 2vpp @ 100 Mhz
Below no high res filtering used at all, basic measurements on screen followed by the FFT images from 1.5Ghz up to 5Ghz plus colour graded version. Peak table set at 15 results
-
Just as a matter of coarse I performed a full bandwidth check on the scope, via my Agilent VSG E4433B and a Bodnar 40ps rise time reference.
Measurements were started a 2Vpp @ 1Ghz and reaches -3dB (1.4 Vpp) @ 2.145Ghz plus the Bodnar rise time of 170ps. Pretty happy we have the correct bandwidth for this Rigol
-
Thanks for that!!
I think some images got scrambled. :o
But on FFT you can see nicely that 9th harmonic is already 55dB lower power than 100 MHz fundamental..
So for scope time domain display, 10x more bandwidth than squarewave frequency is quite ok.
-
(Attachment Link)
This is an AM modulated signal using an 8.33MHz carrier and 100Hz modulation. Peak detect is on. Trace 1 (top) is unfiltered, trace 3 (bottom) is filtered using a 600kHz low-pass filter. I hope that this makes it very clear that anti-aliasing filtering based on the samplerate is a really bad idea on an oscilloscope.
Correct!! If you do downsampling by filtering, in time domain plot you will see absolutely nothing once your timebase gets long enough. That is worse than aliasing. With aliasing you at least get a notion there is SOME signal. And as you correctly point out, enabling Peak detect mode (or if the scope has functioning antialiasing function), you will get correct display, similar to what you would see on the analog scope.
What is good for scope is not optimum for SA and vice versa.
Attached are three Octave scripts. Please remove the ".txt" and run them. Then look at the time series that are output.
fig0.m creates a minimum phase Nyquist limited impulse response
fig1.m low pass filters it with 3 different filter profiles
fig2.m downsamples the output for the 3 filter profiles
Please feel free to make the input impulse as high frequency as you like. But please stop spouting the BS about low pass filtering making things go away. The scripts prove you are wrong. The loss of amplitude after filtering is easily corrected.
This is a first week of class homework exercise.
Reg
-
[...]
I think some images got scrambled. :o
[...]
There seems to be a bug with the forum software, it changes the order of images as compared with what the author specified when writing the post.
-
Please feel free to make the input impulse as high frequency as you like. But please stop spouting the BS about low pass filtering making things go away. The scripts prove you are wrong. The loss of amplitude after filtering is easily corrected.
Well... it seems to me that what is lost due to limited ADC resolution AND noise can not be recovered.
-
Attached are three Octave scripts. Please remove the ".txt" and run them. Then look at the time series that are output.
fig0.m creates a minimum phase Nyquist limited impulse response
fig1.m low pass filters it with 3 different filter profiles
fig2.m downsamples the output for the 3 filter profiles
Please feel free to make the input impulse as high frequency as you like. But please stop spouting the BS about low pass filtering making things go away. The scripts prove you are wrong. The loss of amplitude after filtering is easily corrected.
This is a first week of class homework exercise.
Reg
Ok let's try this. Maybe this will help.
Create file that makes 50 MHz sinewave AM modulated with 100 Hz. Sample it with 5 GHz, for 50 ms. That will give you 250 megasamples. Make data 8 bit signed byte. -127 to +127
Resample that to 2.5 megasamples with whatever you think is best. We'll consider that screen buffer, it's a nice 100x decimation.
Give me that file to plot it in time domain.
In Octave ammod function in communications package can be used to create am modulated signal.
-
Thanks for that!!
I think some images got scrambled. :o
But on FFT you can see nicely that 9th harmonic is already 55dB lower power than 100 MHz fundamental..
So for scope time domain display, 10x more bandwidth than square wave frequency is quite ok.
Apologies that was down to me having a Sunday moment, all rectified now :)
-
Attached are three Octave scripts. Please remove the ".txt" and run them. Then look at the time series that are output.
fig0.m creates a minimum phase Nyquist limited impulse response
fig1.m low pass filters it with 3 different filter profiles
fig2.m downsamples the output for the 3 filter profiles
Octave & math n00b here... when I run fig0.m, I get two files: a0 containing 4096 Ones, and d0 containing "4096" followed by 4095 Zeros. Is that the expected output?
-
Intuition is not a substitute for mathematical analysis and as most people learn very quickly in Calculus and Physics , rarely is intuition correct. My last post was a demonstration of the obvious to anyone who knows anything about DSP. The only part that was moderately above the fundamentals of the discrete Fourier transform was use of the Hilbert transform to compute the minimum phase time domain impulse response.
Upon reflection since my last post I have concluded that I am wasting my time. If the simple tests I have performed so far are to be buried in such a high level of noise from the ignorant I see no hope for a useful discussion of the correct interpolation operator and more nuanced topics. So I shall not be reading or posting further.
Edit: Just in case there are people reading this who are actually interested in in my original project of comparing low end scopes, I created scope-wars@groups.io a while back when this thread got out of control. If you are interested in scope evaluations I invite you to join. It will be moderated to prevent the level of sheer idiocy that has dominated this thread. Join the group, but don't post and if enough people join I'll repost stuff from this thread and we can continue discussing what is right, wrong and a "feature" of DSOs
Have fun!
Reg
PS @SilverSolder There is one non-zero value in d0, the first value.
PPS @2N3055 A 50 MHz carrier modulated by a 100 Hz signal consists of 3 frequencies, 49.9999, 50.0000, and 50.0001 MHz. This is ARRL Novice license guide knowledge. Do your own damn DSP. I already know the answer.
-
Intuition is not a substitute for mathematical analysis and as most people learn very quickly in Calculus and Physics , rarely is intuition correct. My last post was a demonstration of the obvious to anyone who knows anything about DSP. The only part that was moderately above the fundamentals of the discrete Fourier transform was use of the Hilbert transform to compute the minimum phase time domain impulse response.
Upon reflection since my last post I have concluded that I am wasting my time. If the simple tests I have performed so far are to be buried in such a high level of noise from the ignorant I see no hope for a useful discussion of the correct interpolation operator and more nuanced topics. So I shall not be reading or posting further.
Have fun!
Reg
PS there is one non-zero value in d0, the first value.
It wasn't a waste of time. I clearly shows that you will do anything to glorify how much smarter you are than all of us combined, but still unable to solve simplest problem put in front of you from real world...
Make that AM signal as I asked and downsample it the way I asked so we can see what signal will look like. If it looks like it should, kudos to you. If not than we can discuss difference between knowing how to solve math and knowing which math needs solving.
-
Intuition is not a substitute for mathematical analysis [...]
Reminds me of the old Einstein quote, "Imagination is more important than knowledge."
But he probably didn't mean that imagining our own facts is a good idea... so yes, applying the scientific method is important once the discussion gets to a certain level of detail! :D
-
@SilverSolder,
I have no doubt Reg knows all that stuff in and out. Since 2017 he's been on a crusade about how scopes are stupid and don't work. He started this topic with sentence " My goal in this is to embarrass the OEMs into improving their products..."
And every time somebody asks concrete real world question like I did, he starts with ad hominem attacks how stupid is everybody and if we understood iota about DSP we wouldn't be asking stupid question.
He bought GW Instek 2000E series scope (very nice scope BTW) to improve it, and then within few weeks damaged it so it was useless afterwards. Our member here Nctnico, meanwhile, uses that same scope on daily basis (despite having more expensive scopes) in production environment, and sing praises to it all the time.
Look, RHB obviously knows about DSP math a lot. If he can solve some problems that we mentioned here, really solving something Keysight (or Siglent or Rigol) doesn't already know, I'm sure there would be a handsome consulting fee somewhere for him.
Problem is that all of the things he mentions are known things in DSP. He once trumpeted about sparse sampling for months and how scopes would be done that way if only manufacturers knew better. Until they managed to explain to him it doesn't work on real time scope because it works by virtue of repetitive sampling. And is so computationally intensive that it is simpler to just sample at high sample rate and be done with it.
So yeah he's very smart, but controversial...
-
I think he's just got too much ego. You asked him to do one thing and he replies back with you're all too dumb so I'm done with it. That's just ego. I enjoyed the thread while it lasted.
-
If the simple tests I have performed so far are to be buried in such a high level of noise from the ignorant I see no hope for a useful discussion of the correct interpolation operator and more nuanced topics. So I shall not be reading or posting further.
This topic started out as potentially being a very interesting technical review and discussion of the engineering, performance, and applicability of the FFT math function built into many oscilloscopes as compared to spectrum analyzers. If a collection of techniques and test methods had been developed and presented it would have provided the opportunity for many here to test their equipment and setups, providing data that could have been collected into a spreadsheet, similar to the one for multimeters.
Unfortunately, it quickly devolved into an appeal to your own authority by doing things like bringing up the size of your bookshelf and years of experience while belittling others. All we got from you were some screenshots without a lot of context, data analysis, consistent comparisons, deep inside the noise of a bunch of talking at people, not talking with people.
Have fun!
Reg
And what's with this sophomoric trolling in all your posts?
-
@SilverSolder,
I have do doubt Reg knows all that stuff in and out. Since 2017 he's been on a crusade about how scopes are stupid and don't work. He started this topic with sentence " My goal in this is to embarrass the OEMs into improving their products..."
And every time somebody asks concrete real world question like I did, he starts with ad hominem attacks how stupid is everybody and if we understood iota about DSP we wouldn't be asking stupid question.
He bought GW Instek 2000E series scope (very nice scope BTW) to improve it, and then within few weeks damaged it so it was useless afterwards. Our member here Nctnico, meanwhile, uses that same scope on daily basis (despite having more expensive scopes) in production environment, and sing praises to it all the time.
Look, RHB obviously knows about DSP math a lot. If he can solve some problems that we mentioned here, really solving something Keysight (or Siglent or Rigol) doesn't already know, I'm sure there would be a handsome consulting fee somewhere for him.
Problem is that all of the things he mentions are known things in DSP. He once trumpeted about sparse sampling for months and how scopes would be done that way if only manufacturers knew better. Until they managed to explain to him it doesn't work on real time scope because it works by virtue of repetitive sampling. And is so computationally intensive that it is simpler to just sample at high sample rate and be done with it.
So yeah he's very smart, but controversial...
Smart, controversial - and not boring! :D Like many other posters in this thread as well. The quality of many of the posts have been amazing, frankly.
Is there any doubt that scopes can, and eventually will, become even better than they are today? - That can only happen by questioning the basic assumptions and limitations that led to the current designs, good though they may be, as we have been doing here. Questioning long held assumptions is useful even if it leads full circle back to understanding why things are done the way they are within current practical limitations.
Tempers flare, usually we all cool off again? - this is the Internet after all... it is just a sign that what people are saying matters to them and they take it seriously. Makes a refreshing change from politicians etc. just saying what they think will please?
-
Hi,
I think it is a good topic, that has caused us to think about the limitations of DSOs.
Here are some random thoughts on this subject.
Marking or highlighting data points
In LTspice you can mark the data points. This is a nice feature, you can easily what data the program is using to create the waveforms.
I can do the same thing on my MDO 4K scope. I would post the picture, but you can visualize sparse dots. This is a handy feature.
The forum is a melting pot with people with different backgrounds
Some members of this group are very familiar with the design decisions made in designing a successful scope, their insights are invaluable.
Anti-Aliasing Filter (or lack of one)
The LMH6518 used in the Rigol DS2072A, measured by forum member Bud:
(https://www.eevblog.com/forum/projects/project-yaigol-fixing-rigol-scope-design-problems/?action=dlattach;attach=211935;image)
is not suitable for anti-aliasing filter. It is only about 20dB / decade.
Sampling Rate and memory depth
In my mind there is no substitute for sampling speed and memory depth.
If we take an AM modulated RF signal, to accurately reproduce the signal you need 5 to 10 samples per period of the RF carrier.
To resolve the AM modulation in an FFT you need about a sample length equal to about 5 cycles of the modulation.
The resolution, bin width, is 1/acquisition time.
You are trying to resolve three components, RF-mod, RF and RF+ modulation.
If we take the thought experiment proposed by 2N3055
50MHz with 100Hz modulation
a sample rate of 50MHz x 5 or greater is needed
to resolve the 100Hz modulation, chose 20Hz bins, so 50ms of acquisition.
It may be better to use a spectrum analyzer?
A spectrum analyzer still needs the acquisition time, but it can reduce the number of sampling rate by mixing with an LO.
Design Decisions
A scope manufacturer could invest time and effort and design software to extract every last drop of performance out of a signal chain.
or they could invest in the acquisition hardware, increase the sample rate, interleave ADCs etc.
And get similar improvements in performance per dollar.
Given the choice of
1GHz bandwidth 2.5Gsps and advanced DSP
or
1GHz bandwidth 5Gsps and modest signal processing.
I would chose the later. I might make the wrong choice.
I have a 1GHz scope, 5Gsps on my bench, not because I am looking at 500MHz signals, because I don't have to be concerned about the scope performance when looking at 100 MHz (and often much less) signals.
Regards,
Jay_Diddy_B
-
[...]
Sampling Rate and memory depth
In my mind there is no substitute for sampling speed and memory depth.
[...]
That reminds me of old school hot rodders saying "There's no substitute for cubic inches"... until a smaller, lighter car with a four cylinder engine and a high pressure turbo blows them off the track! :D
Basically you are right, of course, but making the best use of whatever sampling rate the hardware can provide is still going to make for a better scope.
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
-
@SilverSolder,
I have do doubt Reg knows all that stuff in and out. Since 2017 he's been on a crusade about how scopes are stupid and don't work. He started this topic with sentence " My goal in this is to embarrass the OEMs into improving their products..."
And every time somebody asks concrete real world question like I did, he starts with ad hominem attacks how stupid is everybody and if we understood iota about DSP we wouldn't be asking stupid question.
He bought GW Instek 2000E series scope (very nice scope BTW) to improve it, and then within few weeks damaged it so it was useless afterwards. Our member here Nctnico, meanwhile, uses that same scope on daily basis (despite having more expensive scopes) in production environment, and sing praises to it all the time.
Look, RHB obviously knows about DSP math a lot. If he can solve some problems that we mentioned here, really solving something Keysight (or Siglent or Rigol) doesn't already know, I'm sure there would be a handsome consulting fee somewhere for him.
Problem is that all of the things he mentions are known things in DSP. He once trumpeted about sparse sampling for months and how scopes would be done that way if only manufacturers knew better. Until they managed to explain to him it doesn't work on real time scope because it works by virtue of repetitive sampling. And is so computationally intensive that it is simpler to just sample at high sample rate and be done with it.
So yeah he's very smart, but controversial...
Smart, controversial - and not boring! :D Like many other posters in this thread as well. The quality of many of the posts have been amazing, frankly.
Is there any doubt that scopes can, and eventually will, become even better than they are today? - That can only happen by questioning the basic assumptions and limitations that led to the current designs, good though they may be, as we have been doing here. Questioning long held assumptions is useful even if it leads full circle back to understanding why things are done the way they are within current practical limitations.
Tempers flare, usually we all cool off again? - this is the Internet after all... it is just a sign that what people are saying matters to them and they take it seriously. Makes a refreshing change from politicians etc. just saying what they think will please?
RHB is right in theory. If all scopes start using 12 bit converters with very high sample rate (like using 10 GS/s for a 1 GHz scope) a scope could be made that would have digital corrections making it have perfect pulse response. If data pumps in scopes where doubled so we could independently feed data to time domain scope segment and to FFT based realtime SA segment, we could have both being perfect without compromises..
But all that stuff is nothing new, it's been done for years by Lecroy, Tektronix, Keysight. Even modest Rigol has some form of pulse response equalization in their new ASIC.
But most of that stuff means massive (20-50 fold) increase in sampled and processed data. And that is expensive.
Will Rigol come out with new version of chipset that will quadruple amount of data it can process? Will manufacturers eventually come out with something like Zynq UltraScale+ RFSoC but 10 times more powerful, so you can squeeze whole scope inside it, and start selling it for 20 USD a chip? They might. But I wouldn't hold my breath.
-
Very nice post Jay.
-
Just in case there are people reading this who are actually interested in in my original project of comparing low end scopes, I created scope-wars@groups.io a while back when this thread got out of control. If you are interested in scope evaluations I invite you to join. It will be moderated to prevent the level of sheer idiocy that has dominated this thread. Join the group, but don't post and if enough people join I'll repost stuff from this thread and we can continue discussing what is right, wrong and a "feature" of DSOs.
Suggestion: Since this isn't the first time you've gone this route, maybe it would be more efficient to launch your next thread on groups.io, so you can maintain control of the content from the beginning.
-
[...]
Sampling Rate and memory depth
In my mind there is no substitute for sampling speed and memory depth.
[...]
That reminds me of old school hot rodders saying "There's no substitute for cubic inches"... until a smaller, lighter car with a four cylinder engine and a high pressure turbo blows them off the track! :D
Basically you are right, of course, but making the best use of whatever sampling rate the hardware can provide is still going to make for a better scope.
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
Yeah, but increasing displacement is very simple and reliable and easy to manufacture. In real life it is cheaper to use bigger engine than high tech.
And easiest way to accomplish that scope newer aliases IS to have a scope that has massive oversampling and large memory. But that is expensive in BOM and processing power needed. So compromises must be made.
-
Just in case there are people reading this who are actually interested in in my original project of comparing low end scopes, I created scope-wars@groups.io a while back when this thread got out of control. If you are interested in scope evaluations I invite you to join. It will be moderated to prevent the level of sheer idiocy that has dominated this thread. Join the group, but don't post and if enough people join I'll repost stuff from this thread and we can continue discussing what is right, wrong and a "feature" of DSOs.
Suggestion: Since this isn't the first time you've gone this route, maybe it would be more efficient to launch your next thread on groups.io, so you can maintain control of the content from the beginning.
And if you are going to compare cheapest scopes available, it would help to compare them in terms that are going to useful to people in real life scenarios. I can tell you up front what you going to see if you shoot a 100 ps pulse into 100MHz cheap scope. Some useless squigle that will tell something about how scope handles signals it was never meant to see.....
On the other hand if this was meant to be a theoretical treatise on correlation of oscilloscope pulse response with scope design and filter performance, that would be very interesting to many people. But it wasn't driven in that direction.
-
[...]
Sampling Rate and memory depth
In my mind there is no substitute for sampling speed and memory depth.
[...]
That reminds me of old school hot rodders saying "There's no substitute for cubic inches"... until a smaller, lighter car with a four cylinder engine and a high pressure turbo blows them off the track! :D
Basically you are right, of course, but making the best use of whatever sampling rate the hardware can provide is still going to make for a better scope.
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
Yeah, but increasing displacement is very simple and reliable and easy to manufacture. In real life it is cheaper to use bigger engine than high tech.
And easiest way to accomplish that scope newer aliases IS to have a scope that has massive oversampling and large memory. But that is expensive in BOM and processing power needed. So compromises must be made.
The trend in the auto industry is towards smaller and more efficient engines.
E.g. the Ford Mustang 2020 Eco Boost four cylinder... not your grandfather's four cylinder. ( This is a 4-cylinder Mustang that hits 60 mph in 4.5 seconds, with a top speed of 155 mph, from its 332 horsepower, 350 lb-ft torque 2.3 liter four banger... ) And this isn't somebody's hot rod, it is a stock production car that you can buy today.
Is there any doubt that improvements in silicon will eventually make your $20 scope on a chip possible? - except I don't think the A brand manufacturers will end up shrinking today's tech onto such as chip (the B brands probably will). The A brands will be building new and improved stuff that pushes the new chip to its limits, instead!
-
Basically you are right, of course, but making the best use of whatever sampling rate the hardware can provide is still going to make for a better scope.
"making the best use of" samples also costs money. The point I've been trying to make regarding wanting fs/10, that Jay Diddy makes very well here, is that fs/10 works even if the implementation lacks sophistication is crude. A system with fs/5 and a 4-pole Bessel input filter or fs/2.5 with even more sophisticated systems may work acceptably, but if, for a given bandwidth, fs/10 with a lame input filter and some simple DSP you can accomplish the same thing for less money, that's the ticket, IMO.
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
ETS (Equivalent Time Sampling) resolution is only limited by the ability to time the samples, given a repetitive signal. I think earlier Rigols even had it, but they've dropped it as a feature, which I'm guessing is because of how the trigger is implemented now. You still get a sort of random dot-walk effect that will give you a picture similar to an old ETS scope but this is not controlled and is due to the varied relationship between the trigger point and the sampling points.
-
A few more points:
The purpose of the thread is to document what is wrong with DSOs on the market.
I am not interested in teaching DSP 101.
I am not willing to spend the time to address every misunderstanding of DSP someone has.
I appreciate having someone point out an error I make.
If you are not willing to take the time to demonstrate an alleged error on my part, I shall ignore you unless it is stated such that it is immediately obvious.
Please do not raise issues which are amply addressed by references I provide. If you are not willing to read the references, I'm not willing to explain it to you.
I'm sorry, but I'm not a "nice" guy. I am not willing to indulge adults who still behave like children.
Have Fun!
Reg
Reg, I'm sorry to say this but it was you who wrote these conditions... And it was you that throwed in the towel. :-//
I sincerely hope all the "sheer idiots" that have been around continue to post here because I liked to read most of the discussions. I only regret that my level of real "idiocy" doesn't allow me to understand many of the concepts right away but I surely apprecciate most of the comments and been learning some things.
OTH, I can easily recognize that doing a 1 or 2 chips scope is a massive endeavour. My deep respect for those guys (US, chinese, germans, etc). And, some times, we take these equipments to debug our own 1-processor circuit!! And we like to expect/demand the equipment to be flawless and be as cheap as possible!
-
And easiest way to accomplish that scope newer aliases IS to have a scope that has massive oversampling and large memory. But that is expensive in BOM and processing power needed. So compromises must be made.
Understood - and this is presumably the birth place of sample dithering and other creative methods of making the best of the hardware we can realistically budget for?
After all, the marketing department would massacre the competition if "our" scope e.g. had higher bandwidth or some other improved spec using the same level of hardware?
-
ETS (Equivalent Time Sampling) resolution is only limited by the ability to time the samples, given a repetitive signal. I think earlier Rigols even had it, but they've dropped it as a feature, which I'm guessing is because of how the trigger is implemented now.
Low end DSO manufacturers dropped ETS when digital triggering became less expensive than analog triggering. Digital triggering is less expensive now because increases in integration made digital logic less expensive. Analog triggering does not benefit from increases in integration.
What does analog triggering require than digital triggering does not?
1. Trigger pickoffs for each channel.
2. Analog multiplexing to select the trigger channel.
3. Analog multiplexing to select the trigger coupling.
4. A fast analog comparator.
5. An optional time delay counter to measure the difference between the sampling clock and trigger.
In comparison, digital triggering might as well be free with increases integration.
-
OTH, I can easily recognize that doing a 1 or 2 chips scope is a massive endeavour. My deep respect for those guys (US, chinese, germans, etc). And, some times, we take these equipments to debug our own 1-processor circuit!! And we like to expect/demand the equipment to be flawless and be as cheap as possible!
Agree!!
To be honest, I would like to apologize everybody for being persistent. I was really ticked off by " My goal in this is to embarrass the OEMs into improving their products.".
It just bothers me. The older I am, I seem to have harder time letting go.. I will be one grumpy old man one day... :-DD
-
To be honest, I would like to apologize everybody for being persistent. I was really ticked off by " My goal in this is to embarrass the OEMs into improving their products.".
Nahh. In the end it was 'much ado about nothing'. There ain't no such thing as a free lunch. Or better put: 'you can not break the rules of physics'. If you want to get maximum bandwidth from a given samplerate you'll need a steep anti-aliasing filter. Steep filters come with lot's of phase shift. However, if you set a bandwidth limit (higher frequency oscilloscopes usually have several settings) you can get a more gradual roll-off which also has a better phase performance. But in the latter case you'll need a higher samplerate (and likely spend more money) to achieve a higher bandwidth compared to steep filtering. Pick your poison...
edit: typo
-
As has been pointed out the discussion already has some really interesting technical discussions revolving around whether scope FFT's can actually be useful enough to make a meaningful contribution to your working environment.
My answer is yes, are they the best way of viewing frequency domains? to a degree yes for a good look at what's going inside a signal and its artifacts whilst looking at the voltage simultaneously then it gives you another aspect of investigation which is great. For a more in-depth analysis then a SA or if you can a RTSA will allow a more useful look at the signal, these are though at a cost.
These days I perform more multi domain analysis far more, by tying in the RTSA and 2Ghz scope via IF output on the RSA5000.
By enabling both time and frequency analysis together your ability to really look deep into stubborn problem is greatly eased.
The new Keysight MXR with its massive hardware horsepower and DDC 320Mhz RTA then maybe the tide is turning now.
However I am firm believe of raw horsepower (well torque actually!) yes the new cars have more horse per CC however they are incredibly complex and use many tricks and DSP side routes to achieving their desired results.
Decent sample rates and long memory and ideally with a 12 bit ADC and low noise front end will make a very positive difference.
Thanks to all of the contributors on this thread, though Reg may wish to take a short break and contemplate life and treating people like little semi trained minions does you no favours when making friends and asking for help. From time to time we all need a little assistance and EEV blog is great community for sharing
Considering the amount of experience on this subject shown here its been great thank you all
-
As has been pointed out the discuss has some really interesting technical discussions revolving around whether scope FFT's can actually be useful enough to make a meaningful contribution to your working environment.
My answer is yes, are the the best way of viewing frequency domains to a degree yes for a good look at what's going inside a signal and its artefacts whilst looking at the voltage simultaneously then it gives you another aspect of investigation great,
That is absolutely true as well. Every now and then I implement signal processing algorithms and looking at the frequency and time domain of a signal is very helpful so optimise the algorithm / see where the algorithm goes wrong.
-
That reminds me of old school hot rodders saying "There's no substitute for cubic inches"... until a smaller, lighter car with a four cylinder engine and a high pressure turbo blows them off the track! :D
Yeah, but increasing displacement is very simple and reliable and easy to manufacture. In real life it is cheaper to use bigger engine than high tech.
A lot of those hot rodders have high pressure turbos too. And nitrous.
In a straight line they'll have 1000ft/lb of torque to destroy you with.
-
As has been pointed out the discuss has some really interesting technical discussions revolving around whether scope FFT's can actually be useful enough to make a meaningful contribution to your working environment.
My answer is yes, are the the best way of viewing frequency domains to a degree yes for a good look at what's going inside a signal and its artefacts whilst looking at the voltage simultaneously then it gives you another aspect of investigation great, for a more in-depth analysis then a SA or if you can a RTSA will allow a more useful look at the signal, these are though at a cost.
These days I perform more multi domain analysis far more, by tying in the RTSA and 2Ghz scope via IF output on the RSA5000.
By enabling both time and frequncy analysis together your ability to really look deep into stubborn problem is greatly eased.
The new Keysight MXR with its massive hardware horsepower and DDC 320Mhz RTA then maybe the tide is turning now.
However I am firm believe of raw horsepower (well torque actually!) yes the new cars have more horse per CC however they are incredibly complex and use many tricks and DSP side routes to achieving their desired results.
Decent sample rates and long memory and ideally with a 12 bit ADC and low noise front end will make a very positive difference.
Thanks to all of the contributors on this thread, though Reg may wish to take a short break and contemplate life and treating people like little semi trained minions does you no favours when making friends and asking for help. From time to time we all need a little assistance and EEV blog is great community for sharing
Considering the amount of experience on this subject shown here its been great thank you all
I would also like to thank you for your very valuable insight and for sharing real life experience with equipment you use.
-
To be honest, I would like to apologize everybody for being persistent. I was really ticked off by " My goal in this is to embarrass the OEMs into improving their products.".
It just bothers me. The older I am, I seem to have harder time letting go.. I will be one grumpy old man one day... :-DD
Your reaction is understandable. There's nothing more insufferable than someone -- whose experience is in another field -- who thinks they know more than everyone else.
-
Fungus
A previous life used to run a race team
You may laugh my personal occasional ride has 1400lbs torque with two turbos and nitrous but propane to go with the gas it weighs 1300kg, here's the rub it's a diesel. The last set of rollers we tested it it broke the retardation device
Always exceptions to rule i would agree
Lots of ways of skinning a cat
Sorry back on topic
-
OTH, I can easily recognize that doing a 1 or 2 chips scope is a massive endeavour. My deep respect for those guys (US, chinese, germans, etc). And, some times, we take these equipments to debug our own 1-processor circuit!! And we like to expect/demand the equipment to be flawless and be as cheap as possible!
Agree!!
To be honest, I would like to apologize everybody for being persistent. I was really ticked off by " My goal in this is to embarrass the OEMs into improving their products.".
It just bothers me. The older I am, I seem to have harder time letting go.. I will be one grumpy old man one day... :-DD
Yes Reg certainly went out on a limb stating that.
Here's another thread investigating similar scope response characteristics:
https://www.eevblog.com/forum/testgear/oscilloscope-frequency-response-correction-program/ (https://www.eevblog.com/forum/testgear/oscilloscope-frequency-response-correction-program/)
Fungus
A previous life used to run a race team
You may laugh my personal occasional ride has 1400lbs torque with two turbos and nitrous but propane to go with the gas it weighs 1300kg, here's the rub it's a diesel. The last set of rollers we tested it it broke the retardation device
Always exceptions to rule i would agree
Lots of ways of skinning a cat
Sorry back on topic
Topic ! :o
Reg has left the nest in disgust of all of us.
Been involved with a tiny bit of drag racing too helping the neighbor reach deadlines for meets.
All rotary powered, some 2r, some 3r, NA and turbed.
20B on meth and 40lb boost. First runs on hub dyno getting a basic tune. At a later date pushed to 1400 HP. :o
(https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/?action=dlattach;attach=789591;image)
-
The discrete Fourier transform of a band limited impulse is a constant value.
No. If a impulse is band limited than the spectrum is confined to a finite frequency interval.
Best regards
egonotto
-
That reminds me of old school hot rodders saying "There's no substitute for cubic inches"... until a smaller, lighter car with a four cylinder engine and a high pressure turbo blows them off the track! :D
Yeah, but increasing displacement is very simple and reliable and easy to manufacture. In real life it is cheaper to use bigger engine than high tech.
A lot of those hot rodders have high pressure turbos too. And nitrous.
In a straight line they'll have 1000ft/lb of torque to destroy you with.
No question - big displacement AND all the tricks, is really at the edge of what is possible!
-
[...]
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
ETS (Equivalent Time Sampling) resolution is only limited by the ability to time the samples, given a repetitive signal. I think earlier Rigols even had it, but they've dropped it as a feature, which I'm guessing is because of how the trigger is implemented now. You still get a sort of random dot-walk effect that will give you a picture similar to an old ETS scope but this is not controlled and is due to the varied relationship between the trigger point and the sampling points.
I don't think the dithering that HP did (still does?) is the same at ETS?
If you have a constant e.g. 200MSa/s, and you can dither the samples e.g. by 5 discrete steps, you end up with a "slow" 1GSa/s.
Maybe that effectively is actually the same as equivalent time sampling, just done by random steps instead of one long linear sequence?
-
Fungus
A previous life used to run a race team
You may laugh my personal occasional ride has 1400lbs torque with two turbos and nitrous but propane to go with the gas it weighs 1300kg, here's the rub it's a diesel. The last set of rollers we tested it it broke the retardation device
Always exceptions to rule i would agree
Lots of ways of skinning a cat
Sorry back on topic
Sounds like adequate performance for the morning commute! :D
-
I don't think the dithering that HP did (still does?) is the same at ETS?
If you have a constant e.g. 200MSa/s, and you can dither the samples e.g. by 5 discrete steps, you end up with a "slow" 1GSa/s.
Maybe that effectively is actually the same as equivalent time sampling, just done by random steps instead of one long linear sequence?
AFAIK it is the same, although there might be different ways of implementing it. If you use a uniform ramp of time displacement, you can create artifacts--the details of that I can't explain very well. However, if you vary the sample timing randomly or at least pseudo-randomly, as long as you can accurately measure the time that it was captured, you can put it in it's place and slowly build a record. Or something like that.
-
That reminds me of old school hot rodders saying "There's no substitute for cubic inches"... until a smaller, lighter car with a four cylinder engine and a high pressure turbo blows them off the track! :D
The saying is "there is no replacement for displacement". When your turbo car can win against a Pro Stock team (6.5 seconds) you can claim turbo-victory. Then you have to race against mountain motor wildcats.
Basically you are right, of course, but making the best use of whatever sampling rate the hardware can provide is still going to make for a better scope.
Not if 'making best use of' costs more than greater sampling rate.
-
[...]
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
ETS (Equivalent Time Sampling) resolution is only limited by the ability to time the samples, given a repetitive signal. I think earlier Rigols even had it, but they've dropped it as a feature, which I'm guessing is because of how the trigger is implemented now. You still get a sort of random dot-walk effect that will give you a picture similar to an old ETS scope but this is not controlled and is due to the varied relationship between the trigger point and the sampling points.
I don't think the dithering that HP did (still does?) is the same at ETS?
If you have a constant e.g. 200MSa/s, and you can dither the samples e.g. by 5 discrete steps, you end up with a "slow" 1GSa/s.
Maybe that effectively is actually the same as equivalent time sampling, just done by random steps instead of one long linear sequence?
In the end it all comes down to aliasing. If your sample points are random then aliasing is less of a problem. But in the end the random samples will have a finite interval and that becomes your sampling interval (and thus sets the effective sampling frequency). The Tektronix DSOs I have owned didn't use random intervals for ETS; they just shifted the sample clock a little bit for each 'sweep'. As long as the anti-aliasing filter does its job it will work just find.
-
[...] making the best use of whatever sampling rate the hardware can provide is still going to make for a better scope.
Not if 'making best use of' costs more than greater sampling rate.
I'm not sure that is what we see in the real world. For example, it would have been cheaper for Ford to just put a stonking big V8 in the Mustang to get 330 horses instead of a tricked out 4 cylinder, but they chose not to do that...
In the same way, low to mid range scopes probably have to use hardware that is available at a reasonable cost. Engineers can choose to put it together in creative ways, and maybe double up on some of the chips to interleave or otherwise boost performance. Using software creatively is going to be irresistible to most teams!
-
I'm not sure that is what we see in the real world. For example, it would have been cheaper for Ford to just put a stonking big V8 in the Mustang to get 330 horses instead of a tricked out 4 cylinder, but they chose not to do that...
In the same way, low to mid range scopes probably have to use hardware that is available at a reasonable cost. Engineers can choose to put it together in creative ways, and maybe double up on some of the chips to interleave or otherwise boost performance. Using software creatively is going to be irresistible to most teams!
You can't analyze Ford's decision until you look at the regulatory picture--emissions and fuel economy--and know the actual costs involved. They've been stuffing 2.3 Turbos in Mustangs for 40 years, on and off. It is getting expensive to get larger engines through emissions as well. And believe me, if I know Ford (and I do) they have wrung out the costs in that EcoBoost pretty thoroughly.
As for scopes, my fairly old Tek TPS 2024 uses 2GSa/S. They have a 100MHz version that uses 1GSa/S. So to go to 200MHz, their best solution was to increase the sample rate to maintain fs/10. I'm sure they aren't idiots and I'm sure that increase cost some money--but that is what they did. Until I see actual acceptable performance from lower rates in an economy scope, I'm not going to assume it is easy or should be a given.
-
it would have been cheaper for Ford to just put a stonking big V8 in the Mustang to get 330 horses instead of a tricked out 4 cylinder
I bet you the entire Internet that the tricked out 4 cylinder was chosen to maximize their profits.
-
I'm not sure that is what we see in the real world. For example, it would have been cheaper for Ford to just put a stonking big V8 in the Mustang to get 330 horses instead of a tricked out 4 cylinder, but they chose not to do that...
In the same way, low to mid range scopes probably have to use hardware that is available at a reasonable cost. Engineers can choose to put it together in creative ways, and maybe double up on some of the chips to interleave or otherwise boost performance. Using software creatively is going to be irresistible to most teams!
You can't analyze Ford's decision until you look at the regulatory picture--emissions and fuel economy--and know the actual costs involved. They've been stuffing 2.3 Turbos in Mustangs for 40 years, on and off. It is getting expensive to get larger engines through emissions as well. And believe me, if I know Ford (and I do) they have wrung out the costs in that EcoBoost pretty thoroughly.
As for scopes, my fairly old Tek TPS 2024 uses 2GSa/S. They have a 100MHz version that uses 1GSa/S. So to go to 200MHz, their best solution was to increase the sample rate to maintain fs/10. I'm sure they aren't idiots and I'm sure that increase cost some money--but that is what they did. Until I see actual acceptable performance from lower rates in an economy scope, I'm not going to assume it is easy or should be a given.
Tek used to do that because interpolation is costly in processing power. "Take a sample-plot a dot" is simple as apple pie. And if you have to, if you linear interpolate with enough dots it's gonna look OK. So at that time economic way was to simply oversample. 5x sampling frequency is really OK if done right. But that interpolation via reconstruction filter and proper design of front end.
-
it would have been cheaper for Ford to just put a stonking big V8 in the Mustang to get 330 horses instead of a tricked out 4 cylinder
I bet you the entire Internet that the tricked out 4 cylinder was chosen to maximize their profits.
4 cilinder is cheaper to make. Less moving parts to assemble. 8 cylinder V engine literally has two heads compared to 4 cylinders one. And then crankshaft, bearings...
And one thing is also important: blown 4 cylinder is smaller and lighter... Not only cheaper engine, but easier to design it into car...
-
They likely never would have designed that engine if it weren't for emissions standards. Adding turbos that have high reliability is not as cheap and easy a job as throwing it on your civic which can eat an engine a year. The intake and exhaust plumbing, extra oil and water lines, carbon build-up due to direct injection, and the direct injection components(high pressure fuel system)... If emissions weren't such an issue they would definitely just use a high output V8, just like they did before when emissions weren't as stringent(and they continue to tighten).
For good measure add in the required oil cooler so you don't ruin turbos, for high output engine you need intercoolers, and turbos are not cheap either. I'm betting an equivalent output V8 could be manufactured for less. Probably far less and there might not even be any weight savings because most of those turbo 4's use iron blocks(I may be behind the times here) but an NA V8 has no issues with aluminum.
-
Just in case there are people reading this who are actually interested in in my original project of comparing low end scopes, I created scope-wars@groups.io a while back when this thread got out of control. If you are interested in scope evaluations I invite you to join. It will be moderated to prevent the level of sheer idiocy that has dominated this thread. Join the group, but don't post and if enough people join I'll repost stuff from this thread and we can continue discussing what is right, wrong and a "feature" of DSOs.
Suggestion: Since this isn't the first time you've gone this route, maybe it would be more efficient to launch your next thread on groups.io, so you can maintain control of the content from the beginning.
Wow, looks i missed the show grand finale, with the main character leaving the stage theatrically slamming the door. What a drama! The guy was already in cavitation mode when started the topic, it was clear the intent was to show everyone how stupid they are and how smart he is. Sure he will be doing fine on his .io board, some people are just born to run echo chambers.
-
...., some people are just born to run echo chambers.
LOL, love that one!!
-
[...]
For example, the dithering of samples that HP/Agilent introduced. It seems to me that by dithering the samples slightly, you end up with more fine grained results that would otherwise have to come from using a much higher sample rate - but the "punishment" is probably a longer acquisition time, as the dithered samples slowly build up the complete picture?
ETS (Equivalent Time Sampling) resolution is only limited by the ability to time the samples, given a repetitive signal. I think earlier Rigols even had it, but they've dropped it as a feature, which I'm guessing is because of how the trigger is implemented now. You still get a sort of random dot-walk effect that will give you a picture similar to an old ETS scope but this is not controlled and is due to the varied relationship between the trigger point and the sampling points.
I don't think the dithering that HP did (still does?) is the same at ETS?
If you have a constant e.g. 200MSa/s, and you can dither the samples e.g. by 5 discrete steps, you end up with a "slow" 1GSa/s.
Maybe that effectively is actually the same as equivalent time sampling, just done by random steps instead of one long linear sequence?
Just for fun..
When oscilloscope works with decimated samples many or some DPO scopes can do Sequential Acquisition Random Interleaving (SARI)
(https://siglent.fi/data/examples/sinc/TEK2465-1870kHz-pulse-as-DSO-interpolation-exercise-03.png)
This is used signal.. 20ns pulse repeating 1.87MHz.
Next images scope have forced to low sampling speed using decimation and zoomed window for demonstrate aliasing and other things more easy.
Look carefully sampling speed and risetime etc things...
(https://siglent.fi/data/examples/sinc/05--1870kHz-pulse-250Msa-fast-Sinc.png)
Sinc post interpolation on.
(https://siglent.fi/data/examples/sinc/08--1870kHz-pulse-250Msa-fast-vectors.png)
Linear post interpolation on.
It have many times specially by LeCroy and also others said that this is sometimes better for this kind of signals. No gibbs etc...
(https://siglent.fi/data/examples/sinc/03--1870kHz-pulse-250Msa-fast-dots-vs-TEK2465.png)
No interpolation, just pure samples and nothing else (SARI).
In live screen it looks better due to human eye-brain things etc.
If you turn Siglent scope to slow ack mode (what is also useful in some special cases, it works like conventional DSO) mode SARI is not possible.
Other signal, other case.
Specially in these gif animations live screen it looks better due to human eye-brain things etc. Also gif animation amount of frames and timing is not like in IRL where it looks better but this may give some idea for least some because here in forum are also beginners and scope many have many features, settings and adjusments. Not only these what we most use, example always Sinc on without even thinking if this is always best mode.
Signal 45MHz some kind of "squarewave" and things far over 25MHz Nyquist... (this is not very normal how we setup scope but it was more easy to demonstrate with tools what was just available at this time moment) But with this aliasing and not aliasing... just example about SARI with decimated samplerates. Aliasing, no aliasing. Just because random interleave. Of course this works only for enough continuous waveforms.
(https://siglent.fi/data/examples/sinc/Gif4-SincOn-45MHz-Square-totally-aliasing-fNyq-25MHz.gif)
Nothing but aliasing
(https://siglent.fi/data/examples/sinc/Gif5-Dots(SARI)-45MHz-Square--fNyq-25MHz-decimated-samplefreq-50MHz.gif)
Sari and no aliasing
Even simple cheap tools have many features what user need be familiar. Too many times we use example scopes like "this is how I always have done" instead of thinking if it is always wise. Many hate dots mode but do not overlook it even when it is best in some situations, do not too much think polished nice picture, think what you need know about signal and nothing else. Modern DPO scopes can do lot of better than conventional DSO
-
Well that went well. ::)
Perhaps new all users should have to go on a course and get a DSP (Digital Scope Proficiency) licence before they're allowed to drive a scope. :-\ For increased revenue in the new "value chain" there could be large fines for learners caught speeding - going over 50us/div, or they could even be banned for driving while aliased. ???
I'm here all week (http://www.quora.com/What-is-the-root-of-the-cultural-reference-Im-here-all-week-try-the-veal).
(https://siglent.fi/data/examples/sinc/Gif5-Dots(SARI)-45MHz-Square--fNyq-25MHz-decimated-samplefreq-50MHz.gif)
Sari and no aliasing
Even simple cheap tools have many features what user need be familiar. Too many times we use example scopes like "this is how I always have done" instead of thinking if it is always wise. Many hate dots mode but do not overlook it even when it is best in some situations, do not too much think polished nice picture, think what you need know about signal and nothing else. Modern DPO scopes can do lot of better than conventional DSO
I like that bottom only 1 sample per waveform dots+persistance waveform, but won't the trigger position still have 22 1ns samples(*) per waveform cycle to work with, is it still as good when the trigger only has 5 or 6 samples per waveform cycle to work with.
*= maybe only 11 trigger samples per cycle to work with while Ch 1 & 2 are both on.
-
I like that bottom 1 sample per waveform dots+persistence waveform, but won't the trigger position still have 22 1ns samples per waveform cycle to work with, is it still as good when the trigger only has 5 or 6 samples per waveform cycle to work with.
At that timebase scope takes one sample per 20 ns. For a 45 MHz signal, that is a little above 1 sample per period. But since sampling clock of scope is not synchronous with signal, on every trigger that dot will fall on different part of curve. 1000 triggers later + persistence you have reconstructed waveform. At effective 10s off Gigasamples /s... That is how RIS scopes of old reconstructed waveforms.... Dot by dot ( of course using different electronics principles, before someone reacts..). That will work for any scope with dot mode and persistence.
EDIT: I reread what you posted and realized it's not what you asked. Trigger engine works of full 1 GHz sampling clock. Trigger point is anyways interpolated, so it should work well with just a few samples (at 1 GHZ clock) .
-
OTH, I can easily recognize that doing a 1 or 2 chips scope is a massive endeavour. My deep respect for those guys (US, chinese, germans, etc). And, some times, we take these equipments to debug our own 1-processor circuit!! And we like to expect/demand the equipment to be flawless and be as cheap as possible!
Agree!!
To be honest, I would like to apologize everybody for being persistent. I was really ticked off by " My goal in this is to embarrass the OEMs into improving their products.".
It just bothers me. The older I am, I seem to have harder time letting go.. I will be one grumpy old man one day... :-DD
Yes Reg certainly went out on a limb stating that.
Here's another thread investigating similar scope response characteristics:
https://www.eevblog.com/forum/testgear/oscilloscope-frequency-response-correction-program/ (https://www.eevblog.com/forum/testgear/oscilloscope-frequency-response-correction-program/)
Fungus
A previous life used to run a race team
You may laugh my personal occasional ride has 1400lbs torque with two turbos and nitrous but propane to go with the gas it weighs 1300kg, here's the rub it's a diesel. The last set of rollers we tested it it broke the retardation device
Always exceptions to rule i would agree
Lots of ways of skinning a cat
Sorry back on topic
Topic ! :o
Reg has left the nest in disgust of all of us.
Been involved with a tiny bit of drag racing too helping the neighbor reach deadlines for meets.
All rotary powered, some 2r, some 3r, NA and turbed.
20B on meth and 40lb boost. First runs on hub dyno getting a basic tune. At a later date pushed to 1400 HP. :o
(https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/?action=dlattach;attach=789591;image)
I finally get around to seeing what this thread is about and as I expected, the OP went off the deep end, again. They do seem to enjoy the drama:
https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg2688777/#msg2688777 (https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg2688777/#msg2688777)
"Ego pimping" became part of the work culture.
Looks like a nice setup. I've worked on a few home made dynos. One was a toy model to play with the controls, two were larger water brakes limited to around 100HP or so. Mostly I was putting together software and the electronics for them.
You can see my toy one here. While I have zero interest in the RC toys, I came across the group from my research and decided to post there.
https://www.msuk-forum.co.uk/forums/topic/81512-self-built-rc-dyno/?do=findComment&comment=1013094 (https://www.msuk-forum.co.uk/forums/topic/81512-self-built-rc-dyno/?do=findComment&comment=1013094)
For the bike, I just collect the data from driving it. It hardly makes enough power to run your friends cars starter motor. :-DD
Sorry for adding to the noise. I have a soft spot for things that are fast.
-
Been involved with a tiny bit of drag racing too helping the neighbor reach deadlines for meets.
All rotary powered, some 2r, some 3r, NA and turbed.
20B on meth and 40lb boost. First runs on hub dyno getting a basic tune. At a later date pushed to 1400 HP. :o
(https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/?action=dlattach;attach=789591;image)
Looks like a nice setup. I've worked on a few home made dynos. One was a toy model to play with the controls, two were larger water brakes limited to around 100HP or so. Mostly I was putting together software and the electronics for them.
Yeah it's pretty cool Joe and just for you here's the rest of it including the 2nd display with the live engine data.
(https://www.eevblog.com/forum/testgear/scope-wars/?action=dlattach;attach=1017888)
You can see my toy one here. While I have zero interest in the RC toys, I came across the group from my research and decided to post there.
https://www.msuk-forum.co.uk/forums/topic/81512-self-built-rc-dyno/?do=findComment&comment=1013094 (https://www.msuk-forum.co.uk/forums/topic/81512-self-built-rc-dyno/?do=findComment&comment=1013094)
You and your SW at it again, nice. :)
For the bike, I just collect the data from driving it. It hardly makes enough power to run your friends cars starter motor. :-DD
Sorry for adding to the noise. I have a soft spot for things that are fast.
That's how the neighbor is improving his drag cars now too, it's all about the logging. ;)
Amazing what can be deduced from inspecting the 20 or so parameters in the pit and dialing in some engine tweaks that remove a problem and result in faster runs. Bloody interesting stuff. :)
-
Never did drag cars, just built the engines for the owners they gave the 1/4 mile what for :-DD
-
Never did drag cars, just built the engines for the owners they gave the 1/4 mile what for :-DD
Sad story about the setup above.....no not a prang but just as expensive in that we couldn't hold 20B's on the engine plate when they were making that sort of HP, even machine recessing them into it ! :o
Factory steel castings couldn't handle the grunt at launch and got ripped apart.
So many $1000's later and billet end plates, them too recessed 2mm precisely into the engine plate we might soon be ready to go again.
-
$$$ vs lower ET is some sort of exponential curve. The same is true with scopes!! :-DD
I don't think the groups.io is going to be of much help. I remember seeing him post about covid in the HP forums and when a few people posted jokes about the matter, he had a similar response. He reminds me of Kiriakos-GR, who also left to start his own forum. Too bad really as he had put together some fairly nice reviews.
https://groups.io/g/HP-Agilent-Keysight-equipment/search?ev=false&q=covid
-
Well that went well. ::)
Perhaps new all users should have to go on a course and get a licence before they're allowed to drive a scope. :-\ For increased revenue in the new "value chain" there could be large fines for learners caught speeding - going over 50us/div, or they could even be banned for driving while aliased. ???
I'm here all week (http://www.quora.com/What-is-the-root-of-the-cultural-reference-Im-here-all-week-try-the-veal).
(rf-loop edit: /removed image from quote/ look it in orig quoted msg.)
Sari and no aliasing
Even simple cheap tools have many features what user need be familiar. Too many times we use example scopes like "this is how I always have done" instead of thinking if it is always wise. Many hate dots mode but do not overlook it even when it is best in some situations, do not too much think polished nice picture, think what you need know about signal and nothing else. Modern DPO scopes can do lot of better than conventional DSO
I like that bottom 1 sample per waveform dots+persistance waveform, but won't the trigger position still have 22 1ns samples per waveform cycle to work with, is it still as good when the trigger only has 5 or 6 samples per waveform cycle to work with.
In real life it do not even need persistence on. This peristence is here for demonstrate better how it looks in real and also for better imagine trigger jitter. This is DPO. Do you mean 1 sample per waveform or near 1 sample per one cycle in signal.
Here in this image every waveform displayed in bottom window have 7 samples (displayed samples interval is 20ns). It depends situation how many acquisitions (wfm) it can do inside one TFT update period what is in this model 40ms. So one TFT frame, depending wfm speed, may have lot of acquisitions overlaid in one frame and then every acquistion samples are interleaved randomly between sequential acquisitions. In cases where example wfm update rate is 20kwfm/s (not in this image) every TFT frame may have 800 acquisitions and so randomly interleved it looks quite dense, nearly like just continuous line.
Now, if you look carefully this image. As can see, visible is only one trace. But, because I want reduce also real true ADC samplerate as low as possible I have turned also Ch2 on but out from screen because this time there was not trace display off/on feature. So, sampling interval is 2ns! Now bottom window time scale is 10ns/div. 2ns is 1/5 div. There is not at all this amount of signal "jumping" in time axis, of course, as can see.
This also partially answer your question about if there is even less real samples in one period in signal.
Of course it do fine interpolation between true ADC samples for position samples, related to trigger position, to display. I do not remember fine interpolation (fine positioning in time axis) max resolution (xx picoseconds).
In image there is bit over 11 samples for one period in signal, not 22. So, every ancquistion is fine positioned and overlaid to display and without sync to signal dots in every overlaid acquisition is randomly interleaved producing quite good image about signal.
Teledyne LeCroy have much more sophisticated real Random Interleaved Samplimg mode - RIS. This Siglent mode is nothing if compare this and this Siglent is nearly just normal, of course not even advertised. User need only remember to use it in these cases where it is useful. Turn interpolations off... and turn Sinc or Linear on only when really needed and you avoid many problems, if not like look this kind of images, then even when you suspect something, check with Linear interpolation and without any interpolation, just dots (in scopes where dots are real only sampled and in scopes where Sinc can turn off and what do not produce extra fake dots in dots mode.) Scope where Sinc can not turn off when ever and scopes what do not even have real dots mode, example producing fake dots between true dots without even possible to turn off this bullshit... ==>> Recycle.
-
Next images
This is how comparisons should be done... :-+
-
Gentlemen, start you engines elsewhere!!!! Or I'll be going groups.io... :-DD
-
Gentlemen, start you engines elsewhere!!!! Or I'll be going groups.io... :-DD
LOL :-DD
-
Gentlemen, start you engines elsewhere!!!! Or I'll be going groups.io... :-DD
:-DD
Haven't found anything old yet you asked about......still looking through old drives. ::)
-
it would have been cheaper for Ford to just put a stonking big V8 in the Mustang to get 330 horses instead of a tricked out 4 cylinder
I bet you the entire Internet that the tricked out 4 cylinder was chosen to maximize their profits.
4 cilinder is cheaper to make. Less moving parts to assemble. 8 cylinder V engine literally has two heads compared to 4 cylinders one. And then crankshaft, bearings...
And one thing is also important: blown 4 cylinder is smaller and lighter... Not only cheaper engine, but easier to design it into car...
It's all about the power to weight ratio... the founding principle of Lotus! :D
-
They likely never would have designed that engine if it weren't for emissions standards. Adding turbos that have high reliability is not as cheap and easy a job as throwing it on your civic which can eat an engine a year. The intake and exhaust plumbing, extra oil and water lines, carbon build-up due to direct injection, and the direct injection components(high pressure fuel system)... If emissions weren't such an issue they would definitely just use a high output V8, just like they did before when emissions weren't as stringent(and they continue to tighten).
For good measure add in the required oil cooler so you don't ruin turbos, for high output engine you need intercoolers, and turbos are not cheap either. I'm betting an equivalent output V8 could be manufactured for less. Probably far less and there might not even be any weight savings because most of those turbo 4's use iron blocks(I may be behind the times here) but an NA V8 has no issues with aluminum.
Agree, the V8 is a much cheaper way of making power, in a way that increasing the sample rate on a scope just isn't! Stretching the analogy even further, a V8 is increasing parallelism (8 one cylinder engines in parallel), whereas increasing the sample rate is like increasing the RPM of the engine so each cylinder does more explosions per second.
Increaing the RPM, like increasing the sample rate, will soon get you to the point where things start to become unreliable due to the forces involved...
Of course, you can still get V8 Mustangs... they are significantly more powerful than the tricked out 4 cylinder, but they are not stupidly much faster. They sound better! :D
-
Thanks for trying. I get the impression that the misinformed and no-things have to reduce our productive time with frivolous banter.
-
I wanted to add few more observation, looking at discussion here.
But if we go with option 3. (downsampling by filtering), as suggested, we will filter out anything above 10 MHz..
On spectrum analyser, we would be looking at scale of 0-10 MHz and see nice clean spectra of that without any folding from upper bandwidth. Perfect.
What we will see on the screen of oscilloscope ?
It's all wrong. Look at your samplerates and then think about how that aliases with your 50MHz carrier. If you do these kind of tests you need to use an 'odd' frequency like 33MHz. You'll see you can only get a good picture with lower samplerates when you turn peak-detect on.
As mentioned before: if you low-pass filter a signal you'll lose the high frequency content. Think about how a basic AM demodulator works; that is not a low-pass filter but an envelope detector. Do the same test with 20MHz bandwidth on and off and you'll see.
(Attachment Link)
This is an AM modulated signal using an 8.33MHz carrier and 100Hz modulation. Peak detect is on. Trace 1 (top) is unfiltered, trace 3 (bottom) is filtered using a 600kHz low-pass filter. I hope that this makes it very clear that anti-aliasing filtering based on the samplerate is a really bad idea on an oscilloscope.
This argument is all based on a 1950's use case for an analog scope testing for over modulation of an AM transmitter. It is *so* last century. A DSO should provide a spectrum analysis mode for this use case.
When you see spurious side bands the modulation is too high. Besides, very few people have a need to test the modulation of an AM transmitter in 2020 other than broadcast engineers who will be well equipped for the task
Have Fun!
Reg
-
This argument is all based on a 1950's use case for an analog scope testing for over modulation of an AM transmitter. It is *so* last century. A DSO should provide a spectrum analysis mode for this use case.
You are missing the point completely. The AM modulated signal is just easy to show the effect of filtering & peak detect when you are looking at a signal which has both low and high frequency content. Not how to analyse an AM signal. Who is using AM nowadays anyway?
-
This argument is all based on a 1950's use case for an analog scope testing for over modulation of an AM transmitter. It is *so* last century. A DSO should provide a spectrum analysis mode for this use case.
You are missing the point completely. The AM modulated signal is just easy to show the effect of filtering & peak detect when you are looking at a signal which has both low and high frequency content. Not how to analyse an AM signal. Who is using AM nowadays anyway?
My local radio market has a very active AM band. Still got some 50 kW stations on the air.
It is unlikely, however, that anyone is doing groundbreaking research or anything.
-
I thought Elvis had left the building?
-
This argument is all based on a 1950's use case for an analog scope testing for over modulation of an AM transmitter. It is *so* last century. A DSO should provide a spectrum analysis mode for this use case.
When you see spurious side bands the modulation is too high. Besides, very few people have a need to test the modulation of an AM transmitter in 2020 other than broadcast engineers who will be well equipped for the task
Have Fun!
Reg
AM modulated signal was simplest signal to explain that points to the problem.
Since you seem not to have answer to that, let's try something from this century.
How about switching power supply, 1 MHz switching frequency, responding to a load pulse from zero to max load and back, 200 ms cycle. What would you see on a output mosfet gate?
And list goes on.
You can have a system that uses sensors that gathers some useful signal and some noise, and than massaging data to extract only portion that is interesting to you.. Like crankshaft sensor, where we take some pulses and noise and extract pulses that we count to extract frequency /RPM. Here we try to discard as much of sampled data and keep only data that gains us RPM info.
These are types of system you worked on your whole life, just yours were insanely more complex and sophisticated.
If we connect scope to that sensor, we want to see signal coming from sensor with all of the crap. All of it. No filtering and such. We use scope not to extract RPM data, but to verify sensor signal, including noise level, and we are doing it in time domain, correlating it to other sensor edges etc.. If there is noise, we want to see it all. And correlate sources in time domain, i.e. temporal correlation.
Make note that in this context word noise is used for all spurious and unwanted signal injected into system from any source in addition to wanted signal.
We might want to analyse noise for it's spectra and distribution, if noise is persistent (cannot be temporally correlated to known source) so we might try to find it's source that way. This is where we use histograms and spectrograms and what not. Here, your expertise is pure gold, and these are thing nowadays only high end scopes have,each to different extent. Some less expensive scopes are starting to have histograms, but like FFT, if control is limited, it has very limited usability.
If someone is capable of adding features of high end LeCroy to inexpensive scope I'm all for it.
But you cannot break how scope works. You have to keep what is perfect the way it is and add to it.
I believe few years ago I gave you advice to get yourself a Picoscope. They are perfect for you. They have all hardware things sorted out and nice ready made API. So you just grab a chunk of data and start analysing. You could make software for it that puts LeCroy to shame. Hell, if you would make it good I would pay for it...
-
Good grief, his Holiness of un-quantifiable epic knowledge of digital signal processing and quasi quantum mechanics, has re-graced us with his wondrous presence sporting a more than smattering resplendent intelligence of all matters of oscilloscope design and critique.
Reg, we bow down in the presence of your awesome magnitude of mathematical equatia (yes Reg deserves his own unique words reflecting his total omnipotence of majestic ADC design and uber low noise front ends)We sir at at the ready to deliver your every poise, word and utterance as pure gospel.
Forgive us oh master of the signal processing universe, may our spaces between the one and noughts be as relevant as the data carrying bits themselves. Oh Reg we are not worthy please forgive our sins against the father of mystical data transmission lanes of omi-presence.
-
Let's keep playing the ball please. 8)
Maybe at some point we'll end up with a nice Python script which shows us the characteristic in a graph (phase, frequency, group delay) of an oscilloscope using a CSV dump from acquiring an impulse response. That does seem usefull to me. Bonus points if the CSV parser can deal with all (most) oscilloscopes out of the box (IOW: cutting off the headers automatically).
-
Good grief, his Holiness of un-quantifiable epic knowledge of digital signal processing and quasi quantum mechanics, has re-graced us with his wondrous presence sporting a more than smattering resplendent intelligence of all matters of oscilloscope design and critique.
Reg, we bow down in the presence of your awesome magnitude of mathematical equatia (yes Reg deserves his own unique words reflecting his total omnipotence of majestic ADC design and uber low noise front ends)We sir at at the ready to deliver your every poise, word and utterance as pure gospel.
Forgive us oh master of the signal processing universe, may our spaces between the one and noughts be as relevant as the data carrying bits themselves. Oh Reg we are not worthy please forgive our sins against the father of mystical data transmission lanes of omi-presence.
^-^
-
Good grief, his Holiness of un-quantifiable epic knowledge of digital signal processing and quasi quantum mechanics, has re-graced us with his wondrous presence sporting a more than smattering resplendent intelligence of all matters of oscilloscope design and critique.
Reg, we bow down in the presence of your awesome magnitude of mathematical equatia (yes Reg deserves his own unique words reflecting his total omnipotence of majestic ADC design and uber low noise front ends)We sir at at the ready to deliver your every poise, word and utterance as pure gospel.
Forgive us oh master of the signal processing universe, may our spaces between the one and noughts be as relevant as the data carrying bits themselves. Oh Reg we are not worthy please forgive our sins against the father of mystical data transmission lanes of omi-presence.
I always liked British humor humour.
-
Let's keep playing the ball please. 8)
Maybe at some point we'll end up with a nice Python script which shows us the characteristic in a graph (phase, frequency, group delay) of an oscilloscope using a CSV dump from acquiring an impulse response. That does seem usefull to me. Bonus points if the CSV parser can deal with all (most) oscilloscopes out of the box (IOW: cutting off the headers automatically).
There is plenty of scope [haha] for analyzing how a particular oscilloscope actually performs, how much its specifications can be trusted, if there is any headroom in the design, etc.
I don't think anyone would disagree that a scope that doesn't live up to its specifications does not reflect well on the manufacturer.
-
Is war over? Who wins?
-
Is war over? Who wins?
Everyone is still armed.
Just waiting for a trigger to start it again.
Then it will sweep over everything.
And we will have an image of the situation.
What will be our response?
-
Is war over? Who wins?
The people who make the 'scopes.
Apparently they know more about applying the math than the theorists do.