Author Topic: 4ch 100MHz 1GS/s per channel cheap-o-scope  (Read 12453 times)

0 Members and 1 Guest are viewing this topic.

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 18543
  • Country: nl
    • NCT Developments
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #25 on: November 03, 2016, 09:29:49 am »
Almost forgot - there must be option do display non-interpolated raw waveform. 10 samples per wfm is not much and when interpolated all the useful info will be lost for good.
Sorry but no information gets lost due to interpolation. Interpolation is there to make the signal interpretable with our eyes. Just read about the Nyquist theory and you'll learn that 10 times oversampling is overkill and you won't get any extra information. Just more money gone from your wallet.
I would like to live in your magical world where measurements have no noise in them and the front end antialias filters are perfect, the rest of us will happily continue buying scopes with sample rates 5-10x higher than required by nyquist (or more!).
You forget the filtering provided by the probing solution (every probing solution has a limited bandwidth). If your signal has so much noise that it is becoming a problem then you have different challenges to measure it because with or without interpolation or oversampling the noise isn't going away. AND you have to realise you can never get an accurate representation of a signal on an oscilloscope anyway. Limited bandwidth, probe influences (loading), variations in the frequency response, phase delays, limited resolution, non-linearities all sit between the signal and what is shown on your screen. All in all an oscilloscope is good for getting an idea on what a signal looks like especially in the lower end segment and with general purpose probes.
« Last Edit: November 03, 2016, 09:38:50 am by nctnico »
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 2241
  • Country: au
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #26 on: November 03, 2016, 09:47:08 am »
Almost forgot - there must be option do display non-interpolated raw waveform. 10 samples per wfm is not much and when interpolated all the useful info will be lost for good.
Sorry but no information gets lost due to interpolation. Interpolation is there to make the signal interpretable with our eyes. Just read about the Nyquist theory and you'll learn that 10 times oversampling is overkill and you won't get any extra information. Just more money gone from your wallet.
I would like to live in your magical world where measurements have no noise in them and the front end antialias filters are perfect, the rest of us will happily continue buying scopes with sample rates 5-10x higher than required by nyquist (or more!).
You forget the filtering provided by the probing solution (every probing solution has a limited bandwidth). If your signal has so much noise that it is becoming a problem then you have different challenges to measure it because with or without interpolation or oversampling the noise isn't going away.
The bandwidth is not hard limited which is whats required for sampling, there is a diminishing return as the sample rate is increased and aliasing reduces but its common to have a much higher sample rate than strictly required. Its even mentioned in the opening paragraph about antialiasing filters on wikipedia:
https://en.wikipedia.org/wiki/Anti-aliasing_filter
And secondly its not the signal noise but the measurement noise, a higher sample rate still reduces noise in the measurement even if its far beyond whats needed to reproduce the frequency components. You can gloat all you like about your understanding of nyquist theorem but the theorem has many assumptions which do not apply in the real world where the rest of us live and work. So perhaps you could contribute something practical to the discussion rather than your ego stroking theoretical arguments which are repeatedly torn down.
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 17030
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #27 on: November 03, 2016, 12:30:48 pm »
SDS2104X all 4 channels on 1GSa/s for all channel. 2 channels on, 2GSa/ch
Also enough memory for most use,
Powerful fast segmented memory acquisition.
Always backround working Waveform history buffer.
In your budget due to factory offer.
Also it keep raw data independent of interpolation and display settings.
In fact much better than in MrWolf's budget, € 1405 for the SDS2104X (100 MHz 4 Ch) and with the promotion for that price the SDS2204X (200 MHz) is supplied.
So if only 100 MHz is required, purchase a SDS2074X for € 1099 and the SDS2104X is supplied.
1 GSa/s (all channels) and 140 M memory depth, nothing else comes near for the price.
http://www.siglenteu.com/pdxx.aspx?id=1195&T=2&tid=1


I'm told there are some noticeable performance improvements in the next soon to released FW.

Edit.
Link added
« Last Edit: November 03, 2016, 01:14:56 pm by tautech »
Avid Rabid Hobbyist
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 18543
  • Country: nl
    • NCT Developments
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #28 on: November 04, 2016, 01:10:08 am »
So perhaps you could contribute something practical to the discussion rather than your ego stroking theoretical arguments which are repeatedly torn down.
Perhaps I should have noted that most of my paid projects involve sampling and some are data acquisition systems (oscilloscopes without knobs and fancy user interface). Nyquist's theory is the basis for all these projects in order to get a good balance between sampling frequency and anti-aliasing filtering. But you don't have to take my word for it. Keysight has a nice document about it: cp.literature.agilent.com/litweb/pdf/5989-5732EN.pdf. On page 5 you can read that 8x sampling versus 4x sampling shows minimal differences (you have to look really hard).

The summary:
As you’ve read in this application note, there’s more to oscilloscope signal fidelity than just sample rate. In some cases a lower-sample-rate scope may produce more accurate measurement results. To satisfy Nyquist criteria, you need a scope that samples at least three to five times higher than the scope’s bandwidth  specification, depending on the scope’s frequency roll-off characteristics. Achieving higher sample rates often requires that scope vendors interleave multiple  real-time ADCs. But if real-time interleaving is employed, it is critical that the interleaved ADCs be vertically matched and the timing of phase-delayed clocking  must be precise. It should be noted that the problem is not the number of interleaved ADCs; the issue is the level of precision of interleaving. Otherwise, Nyquist’s second rule (equally-spaced samples) can be violated, thereby producing distortion and often negating the expected benefit of higher sample rates.

Back to the OP's original question: in order to sample a signal with frequency components up to 100MHz a 100MHz scope with 500Ms/s is perfectly suitable.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 2241
  • Country: au
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #29 on: November 04, 2016, 02:06:58 am »
Back to the OP's original question: in order to sample a signal with frequency components up to 100MHz a 100MHz scope with 500Ms/s is perfectly suitable.
The OP is wishing to measure small phase shifts which the additional samples will help for, they asked for specific features and backed it up with a legitimate explanation of why.

I don't care what your experience or credentials are, because you're wrong.
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 18543
  • Country: nl
    • NCT Developments
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #30 on: November 04, 2016, 02:24:00 am »
Back to the OP's original question: in order to sample a signal with frequency components up to 100MHz a 100MHz scope with 500Ms/s is perfectly suitable.
The OP is wishing to measure small phase shifts which the additional samples will help for, they asked for specific features and backed it up with a legitimate explanation of why.
If you'd read the paper provided by Keysight you'd known that 500Ms/s + interpolation for 100MHz signals gives you a much higher timing accuracy than the sampling rate would suggest. In the paper Keysight shows their DSO7000 series which uses 2Gs/s to get 10s of ps timing resolution as an example. The answer is right in your face and you still don't see it!  :palm:
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 2241
  • Country: au
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #31 on: November 04, 2016, 07:23:56 am »
Back to the OP's original question: in order to sample a signal with frequency components up to 100MHz a 100MHz scope with 500Ms/s is perfectly suitable.
The OP is wishing to measure small phase shifts which the additional samples will help for, they asked for specific features and backed it up with a legitimate explanation of why.
If you'd read the paper provided by Keysight you'd known that 500Ms/s + interpolation for 100MHz signals gives you a much higher timing accuracy than the sampling rate would suggest. In the paper Keysight shows their DSO7000 series which uses 2Gs/s to get 10s of ps timing resolution as an example. The answer is right in your face and you still don't see it!  :palm:
Or you could come back to the real world (marketing material is easy to make with happy numbers) and actually measure these things, you've had plenty of chances to back down so here is a real measurement on 2 different scopes where they are measuring the skew of two channels. The probes are connected to slew limited logic signals with edges around 5ns which is close to the OPs 100MHz signals, and the scopes are in normal run mode with their onboard measurements providing the data (you can do better with offline analysis).

Thats what you call a correlation, as predicted by noise reduction. Real world once again meets with theory when that includes real world effects.
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 18543
  • Country: nl
    • NCT Developments
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #32 on: November 04, 2016, 10:16:38 am »
The OP specified 1Gs/s which probably means he wants/needs a 1ns time resolution. That is easy to achieve with 500Ms/s + interpolation. If noise is a problem you can use frequency domain filtering (input filtering, bandwidth limiting, hires mode) or time domain filtering using averaging. A much more important feature will be the short term stability of the sample clock.

Sure you can move the goal posts and show a much better and >20 times more expensive oscilloscope with a much higher samplerate has a much higher time resolution but that is a) not what the OP is asking for b) rather obvious and nobody will claim anything different.
« Last Edit: November 04, 2016, 10:21:25 am by nctnico »
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 2241
  • Country: au
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #33 on: November 04, 2016, 10:29:46 am »
The OP specified 1Gs/s which probably means he wants/needs a 1ns time resolution. That is easy to achieve with 500Ms/s + interpolation. If noise is a problem you can use frequency domain filtering (input filtering, bandwidth limiting, hires mode) or time domain filtering using averaging. A much more important feature will be the short term stability of the sample clock.
Try it with a scope and see what result you get. Averaging or high res did not reduce the noise below that of the full sample rate. Pretty pictures and accurate measurements are two different things.
Sure you can move the goal posts and show a much better and >20 times more expensive oscilloscope with a much higher samplerate has a much higher time resolution but that is a) not what the OP is asking for b) rather obvious and nobody will claim anything different.
The example is a 100MHz band limited signal into a 200MHz 4GSa/s scope and showing how the noise in the time measurement increases as you reduce the sample rate through and below 1GSa/s. Thats the conditions the OP is asking about and what you claimed repeatedly would make no difference, and there for all to see (and reproduce to prove to themselves) is the result that increasing sample rate improves the measurement of relative time. Instead of theory and your promises how about you present some data?
 

Offline MrWolf

  • Regular Contributor
  • *
  • !
  • Posts: 209
  • Country: ee
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #34 on: November 04, 2016, 10:33:58 am »
I will tell a story: First I had old russian analog scope. Real straight forward piece of equipment like AK47. But I needed more channels so went for PicoScopes. Soon had two of them with 4ch + 16digital in total, was rather pleased with initial tests (but no idea about digital trickery involved behind the scenes). Afterwards started building some experimental analog contraption involving heavily non-linear components. When started testing almost went mad. It did not produce designed signals no matter what I did, non-linearities were all wrong. For days I debugged and tested and calculated until finally I found the culprit deep in the PS software menus... Sin(x)/x on by default. After switching it OFF discovered that contraption was working as designed from the day 1. So maybe one can get away with 2.5 samples per wfm (PS actually had 4 at max freq, all ch in use) for very well known situations... but for heavily experimental stuff only thing that counts is raw data, period. When you start replacing raw data with math fantasy you usually get string theory or something, not maglev trains :)
Edit: Remembered that there was yet another culprit, "resolution enhancement", I turned it on because is seemed cool to have 12bits instead of 8 :) But this was quick to pinpoint because at low sample rate per wfm all left was straight line on screen :P
« Last Edit: November 04, 2016, 11:03:18 am by MrWolf »
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3143
  • Country: cn
  • Born with DLL21 in hand
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #35 on: November 04, 2016, 11:07:37 am »
Quote
only thing that counts is raw data, period. When you start replacing raw data with math fantasy you usually get string theory or something, not maglev trains :)

 :clap:
If practice and theory is not equal it tells that used application of theory is wrong or the theory itself is wrong.
-
Harmony OS
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 18543
  • Country: nl
    • NCT Developments
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #36 on: November 04, 2016, 11:33:26 am »
Edit: Remembered that there was yet another culprit, "resolution enhancement", I turned it on because is seemed cool to have 12bits instead of 8 :) But this was quick to pinpoint because at low sample rate per wfm all left was straight line on screen :P
In my experience with Picoscopes their resolution enhancement mode can have all kinds of strange effects. However I don't buy sin x/x interpolation was/is the problem because it is a well proven technology in general but you'll have to keep in mind that the samplerate must be high enough to capture the highest frequency component of the signal. Keeping peak-detect on helps a lot to prevent strange signals showing on the screen due to aliasing.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 3613
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #37 on: November 04, 2016, 11:55:18 am »
Quote
only thing that counts is raw data, period. When you start replacing raw data with math fantasy you usually get string theory or something, not maglev trains :)

 :clap:
I had a good laugh at that. Excellent way to put it!  :-DD
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 

Online Fungus

  • Super Contributor
  • ***
  • Posts: 10406
  • Country: 00
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #38 on: November 04, 2016, 12:07:35 pm »
Screenshots or it isn't true.
 

Offline MrWolf

  • Regular Contributor
  • *
  • !
  • Posts: 209
  • Country: ee
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #39 on: November 04, 2016, 12:22:22 pm »
samplerate must be high enough to capture the highest frequency component of the signal
It was rather complex signal. Main point was that without interpolation I immediately realized that I have not enough raw data and tweaked stuff accordingly. Interpolation masked lack of data. My "contraption" is long gone, but just for kicks I might try to reproduce this when I get new signal gen and scope. Seems every scope has its own character and bag of tricks:



So long story short and my advice for other beginners. Start with all fancy stuff OFF.

Edit: Would be cool feature to display data points on top of interpolated signal with different color. This way you get nice "expensive look" but can instantly decide if you have enough data or not.
« Last Edit: November 04, 2016, 12:38:04 pm by MrWolf »
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 18543
  • Country: nl
    • NCT Developments
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #40 on: November 04, 2016, 02:10:44 pm »
The OP specified 1Gs/s which probably means he wants/needs a 1ns time resolution. That is easy to achieve with 500Ms/s + interpolation. If noise is a problem you can use frequency domain filtering (input filtering, bandwidth limiting, hires mode) or time domain filtering using averaging. A much more important feature will be the short term stability of the sample clock.
Try it with a scope and see what result you get. Averaging or high res did not reduce the noise below that of the full sample rate. Pretty pictures and accurate measurements are two different things.
Lets put it to the test then with a GDS-2204E, a HP4421B RF generator (with a stable timebase) set to 101MHz, RF splitter, 2 identical cables, a 33cm cable and some tees & terminators.
First check skew between channel 1 and 2 with averaging (32 cycles) on:

Now with the 33cm extra cable for channel 2
1Gs/s without averaging:

1Gs/s with averaging (32 cycles):

500Ms/s without averaging:

500Ms/s with averaging  (32 cycles):


I think it is clear that the samplerate doesn't affect the time resolution at the same time/div setting and averaging does improve time delay measurement (less noise). It all has to do with the data which is used to do the calculation. HP/Agilent/Keysight scopes typically use the screen data and not the actual samples to do waveform calculations so a lot of the initial accuracy of the samples gets lost in translation at longer time/div settings (Tektronix has a nice application bash note on that BTW) and hence the yellow line in your graph slopes down at first but then levels. It may be different on a Lecroy or Tektronix scope.
« Last Edit: November 04, 2016, 04:58:08 pm by nctnico »
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 2241
  • Country: au
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #41 on: November 05, 2016, 01:53:14 am »
I think it is clear that the samplerate doesn't affect the time resolution at the same time/div setting and averaging does improve time delay measurement (less noise). It all has to do with the data which is used to do the calculation. HP/Agilent/Keysight scopes typically use the screen data and not the actual samples to do waveform calculations so a lot of the initial accuracy of the samples gets lost in translation at longer time/div settings (Tektronix has a nice application bash note on that BTW) and hence the yellow line in your graph slopes down at first but then levels. It may be different on a Lecroy or Tektronix scope.
Determining these sorts of small delays between signals requires some form of reconstruction to the the picosecond resolution from "only" a nanosecond sample rate, so its impossible the measurement is directly from the raw points of the data. Better would be to use the triggers of the channels as in a hardware timer, but it would still need some interpolation to resolve the picosecond resolutions. The 3024A is limited to only 2ns/div and the measurement noise is related to the reconstructed signal as you change up through 5ns/div and 10ns/div it roughly doubles at each step.

The 54845A was used as an example as you can change the sample rate manually without changing the timebase, with a 50 ohm source the signal noise should be lower and we could expect a reduction in the measured time noise. It was particularly interesting as there was a repeatable bump in the measurement noise at 4GSa/s so your comparisons of just 2 different sampling rates may be hitting against some similar oddity in the way the channels are sampled at different rates for interleaving.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 2241
  • Country: au
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #42 on: November 05, 2016, 05:53:14 am »
Swept a whole bunch of different configurations and picked up on some further gotchas around the averaging and quantising. The reconstruction of the waveform can use the extra bits from averaging acquisitions but the measurement engines are heavily quantised. On the Agilent 3024A the delay time measurement is quantised to a single pixel (best case 32ps), while on the above captures of the GW Instek GDS-2204E it appears to be quantised to at least 5ps. Once the acquisition averaging reduces the noise below this quantisation then the measurement engines can report a zero standard deviation when the signal sits in a single quantised value, sitting close to a transition you can get measurements reporting silly things like just femtoseconds of noise. This caused interesting sequences of measurements which didnt fit to funnel plot, exposing their measurement biases. Quantisation in each step of the process is causing some really ugly answers.
 

Offline MrWolf

  • Regular Contributor
  • *
  • !
  • Posts: 209
  • Country: ee
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #43 on: November 09, 2016, 02:45:41 pm »
Conundrum is solved. If found no 4ch 100-200MHz all-channel 1GS/s scope in my price range without severe limitations in one or another area. So instead I bought "disposable-cheap" 4ch scope and two 2-channel signal generators (can be synced!). After applying some slight tweaks I will have 4ch 25MHz complex waveform read/write and 4ch ~100MHz sine read/write capability. Quiz: what exact products I did choose while not abusing the piggy bank of about 1500€ :P
 

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 3613
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #44 on: November 10, 2016, 10:05:07 pm »
Quiz: what exact products I did choose while not abusing the piggy bank of about 1500€ :P
Let me guess: a Hantek 6104BC and two MHS-5200A-25M?  :-DD
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 

Offline MrWolf

  • Regular Contributor
  • *
  • !
  • Posts: 209
  • Country: ee
Re: 4ch 100MHz 1GS/s per channel cheap-o-scope
« Reply #45 on: November 18, 2016, 11:08:32 am »
Well by "not abusing" I meant it was a clean kill. Not stretching or any other way of mutilation.
2x Siglent SDG2042X
1x Rigol DS1054Z
Even 25EUR left for testing-time beer case :D
What I did find out so far:
SDG2042X -  :-+  :clap:
DS1054Z -  :--  :-BROKE  :palm:  |O
Noone to blame but myself, rf-loop did discover long time ago that latter is just a hack...
but Dave said "just buy it" so....  :-//
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf