Author Topic: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review  (Read 3445 times)

0 Members and 1 Guest are viewing this topic.

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3061
  • Country: fi
  • Starting with DLL21
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #50 on: June 18, 2019, 09:13:18 pm »
And in the "digital phospor" / intensity grading modes. I would assume the samples of previous acquisition are not really existing anymore, just the previous waveform(s) must be stored somewhere to process the grading.

I do not fully understand what you mean here.

Do you perhaps mean that after acquisition data is handled  for add to display memory (to digital phosphor) for produce gradation but after then it - older than last one - do not really exist except in display map.
This is not how Siglent works.
"I would assume the samples of previous acquisition are not really existing anymore" No, in Siglent, they still really exist.

Example, if think one display frame. It do acquisition to memory (and interpolate (this  can see when look trigger out signal with other scope, depending interpolation also every single acq. time change, not only time gap when it update new display frame, if 5000X works like previous models) and map it without decimation also to display memory map and then next and next and so on until is time to show display, all accumulated acquisitions overlaid,  but still every acquisitions also separately in wfm  fifo, not only mapped to display memory, even if there is 100 or thousands of individual acquisitions in one display frame producing also one part of intensity gradation.
Also when it produce display frame it do not decimate, it maps every sample. Siglent keep still also all these separate acquisitions in data memory (wfm fifo), not only in display map. Also these every individual acquisitions what one display frame include (or more) are available for user. User can look these separate acquisitions (user can example break last screen back into individual acquisitions and afterward change interpolation or turn interpolation off). There is always background working waveform history buffer (fifo) where are least last frame individual acquisitions every original sample.

How much it do work for produce one screen. It can somehow see when use sequence mode (what is nearly 5 times more fast than normal mode) In this mode it do not produce display screen (and do not interpolate) at all until whole sequence is captured. If capture example 50000 waveforms sequence. After these 50000 are all captured it stops and start produce display frame. If there is example sinc interpolation instead of dots alone, it  takes more time before all these are interpolated and mapped to display memory and displayed with intensity gradation (all 50000 overlaid at once, perhaps over 100M sample points where every single point to point need interpolate and draw )

----
About Dave's video. There can see SDS5000X intensity gradation is somehow broken and need urgent repair. It looks terrible. Also it looks other way terrible because very unstable trig to  AM modulating freq (but this is imho user intentional error)
« Last Edit: June 19, 2019, 03:27:36 am by rf-loop »
If practice and theory is not equal it tells that used application of theory  is wrong or the theory itself is wrong.
It is much easier to think an apple fall to the ground than to think that the earth and the apple will begin to move toward each other and collide.
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 15695
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #51 on: June 18, 2019, 09:18:06 pm »
@tautech: As I wrote in my previous post: so far the Rigol DS1054Z seems to be the only exception. Pehaps you should include the oscilloscope model  WinnieThePooh is testing (the Rigol MSO5000 series) and the problem he describes in you short (likely out of context quote) has nothing to do with sin x/x reconstruction. Don't make the same mistake rf-loop made in his eagerness to make things prettier than they are or make standard features look special. It sounds like a clueless salesperson; 'Look, this Siglent calculator is showing 2 when I punch in 1+1'. Real engineers will be put off by those kind of tactics.
Which is why I included a link to the post, so you can check the context and model.
Yes it seems the 5k series is afflicted by incorrect Dot mode implementation too and this info is from other sources other than rf-loop.  :P
Avid Rabid Hobbyist
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #52 on: June 19, 2019, 07:03:49 am »
But I've got to place some small criticism here -- your input noise comparison is somewhat dodgy (34:00 into the video): Since the R&S had some visible DC offset of the input signal (round about -300µV), this will clearly contribute to the RMS measurement. Try checking noise with the inputs in AC coupled mode. Depending on the internal processing, the average offset will get subtraced in this mode (not sure if it works this way on the R&S, at least the "mature" Rigol products do it like this).
Then I'm sure you'll find that the R&S is at least on par with the new Siglent scope.

AC coupling makes no difference.
But you are right, the DC offset matters.
Just powered it up again and it happened to have almost no DC offset and a I got 107uV RMS, and it's 87uV in standard deviation mode.
After a bit the offset rose and the standard deviation gets rid of it, still around 87uV

Looking at their incomplete manual, at least Siglent included Stdev to compute AC RMS and RMS to compute AC + DC RMS.  But the multimeter has separate AC RMS and DC RMS selections.  Is Siglent always so consistent?

Calculating AC + DC RMS when AC coupling is selected must be a dumb oversight.

87 microvolts RMS over 1 GHz is respectable; many DSOs are much worse.  The lowest noise oscilloscopes I have tested have a slightly lower broadband spot noise than that, about 25% lower.  But with a lower bandwidth, that is expected because higher spot noise comes with higher bandwidth devices.

Maybe I didn't get the point but I can't imagine that any scope on the market isn't interpolating the displayed waveform from "true" sample points. At least in single shot mode there simply isn't any choice, or is it? And in the "digital phospor" / intensity grading modes. I would assume the samples of previous acquisition are not really existing anymore, just the previous waveform(s) must be stored somewhere to process the grading.
Anyway, I still find it puzzling that any kind of interpolation divides the acquisition rate by a factor of four (or more).

Interpolation might be used for the display but not to fill in the acquisition record.

The original DPO mode of operation generated a 3D histogram of the data in hardware at the maximum sample rate with no interpolation.  Modern DSOs tend to generate the histogram (for display or measurement) from the acquisition record with the processor instead of in hardware and the record length then limits the sample rate.  This makes sense when less than the absolute maximum performance is required because processor performance is so much better now than 20 years ago and it makes for less expensive high performance DSO.

But this creates a conundrum.  If interpolation is used, then it corrupts the histogram and any measurements which rely on it like RMS, peak, or whatever if aliasing is present.  I suspect Rigol does something like this explaining various odd measurement behaviors.
 

Offline Performa01

  • Frequent Contributor
  • **
  • Posts: 799
  • Country: at
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #53 on: June 19, 2019, 09:25:23 am »
Maybe I didn't get the point but I can't imagine that any scope on the market isn't interpolating the displayed waveform from "true" sample points. At least in single shot mode there simply isn't any choice, or is it?
Yes, it's hard to believe, yet rumours have it that such scopes do exist. You can always calculate something completely new from a number of original samples, which then just get replaced by the result of the computation (like most modern digital cameras do).

And in the "digital phospor" / intensity grading modes. I would assume the samples of previous acquisition are not really existing anymore, just the previous waveform(s) must be stored somewhere to process the grading.
Yes, there are some scopes that use only the screen (or some small secondary) buffer for intensity grading.

For Siglent X-series scopes there is one irrevocable standard: In normal mode, it will always show all original sample data from within the last 40ms (if available) on the screen at once. In Sequence recording mode it will even show the complete history buffer, i.e. up to 100000 records  and/or a total of 440 Megasamples at once. This way you can detect any anomaly at a glance.

This is true for normal acquisition mode; Peak Detect might, Average and Eres will replace the original sample data by the corresponding processed data, i.e. min/max data pairs in case of Peak Detect, the sliding average of corresponding samples over a number of subsequent records in case of Average and the result of a FIR lowpass filter operation within one record for Eres. Nevertheless, the number of samples doesn't change though and we still get all the data displayed.

This is by the way the major difference to LeCroy scopes, which never ever manipulate the original sample data, thus cannot have peak detect and implement Average as well as Eres as Math functions.

Now for the mystery of intensity grading, which some confuse with persistence (of course we can have persistence as well, but that has absolutely nothing to do with intensity grading):

It is simply a matter of the number of samples falling into the exact same position on the screen – similar to an analog scope, where intensity is determined by the speed of the electron beam moving over the screen. Since each sample represents a certain time step (1/samplerate) the accumulation of these results in a total time which in turn determines the trace intensity.

For this process, a lot of data is involved in the display mapping. It is the record length multiplied by the number of acquisitions within a span of 40ms (which is the display refresh period).

Anyway, I still find it puzzling that any kind of interpolation divides the acquisition rate by a factor of four (or more).
With the explanation given above, we could measure/calculate the data rate for different timebase settings:

200ps/div => 120kpts/s => 4.8kpts/frame;
20ns/div => 110Mpts/s => 4.4Mpts/frame;
1ms/div => 950Mpts/s => 38Mpts/frame;

So while at even slower timebases we can exceed 1Gpts/s, there seems to be no dedicated graphics processor to handle such an amount of graphic data in a fast and efficient way.

The amount of additional dots required for interpolation depends on the signal amplitude, waveform and frequency. This can easily lead to a multiple of the original data to be drawn on the screen. This seems to be the bottleneck and both position mapping and interpolation require additional processing time.

Looking at their incomplete manual, at least Siglent included Stdev to compute AC RMS and RMS to compute AC + DC RMS.  But the multimeter has separate AC RMS and DC RMS selections.  Is Siglent always so consistent?
Well, these are different worlds, I suppose. Experienced DSO users should be familiar with the Stdev measurement (even though I’ve come across “experts” who confuse it with the standard deviation in the measurement statistics), whereas on a DMM the terms AC RMS (often just RMS) and AC+DC RMS are common. But that’s irrelevant, because other than a true scope meter, a pseudo DMM is just a marketing gimmick in my book, that doesn’t belong into a DSO – not in the midrange class anyway. Yet since it was introduced by one of the big boys, every manufacturer feels the need to offer one as well. I think Siglent is the last one to follow that trend.
 

Online nctnico

  • Super Contributor
  • ***
  • Posts: 17394
  • Country: nl
    • NCT Developments
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #54 on: June 19, 2019, 11:55:33 am »
Another option could be that Siglent is making the intensity grading actually useful. There is not much use to hide a small glitch in 255 shades of yellow because you won't see it (just like on an analog scope). 8, maybe 16 shades is more than enough to distinguish between often and more rarely occuring signals.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 3061
  • Country: fi
  • Starting with DLL21
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #55 on: June 19, 2019, 12:17:34 pm »
Another option could be that Siglent is making the intensity grading actually useful. There is not much use to hide a small glitch in 255 shades of yellow because you won't see it (just like on an analog scope). 8, maybe 16 shades is more than enough to distinguish between often and more rarely occuring signals.

Let's hope master of magicians protect us from this devastation of performance.
If practice and theory is not equal it tells that used application of theory  is wrong or the theory itself is wrong.
It is much easier to think an apple fall to the ground than to think that the earth and the apple will begin to move toward each other and collide.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #56 on: June 20, 2019, 05:19:04 am »
Looking at their incomplete manual, at least Siglent included Stdev to compute AC RMS and RMS to compute AC + DC RMS.  But the multimeter has separate AC RMS and DC RMS selections.  Is Siglent always so consistent?

Well, these are different worlds, I suppose. Experienced DSO users should be familiar with the Stdev measurement (even though I’ve come across “experts” who confuse it with the standard deviation in the measurement statistics), whereas on a DMM the terms AC RMS (often just RMS) and AC+DC RMS are common. But that’s irrelevant, because other than a true scope meter, a pseudo DMM is just a marketing gimmick in my book, that doesn’t belong into a DSO – not in the midrange class anyway. Yet since it was introduced by one of the big boys, every manufacturer feels the need to offer one as well. I think Siglent is the last one to follow that trend.

The documentation does not even discuss the difference and then of course there is the standard deviation calculation of the measurements so that is three different names for the same calculation on different data.  I think standard deviation for automatic measurements and AC RMS and DC RMS for acquisition record measurements by both the DSO and digital multimeter would be less confusing.  As evidence I present Dave having to go back and make the noise measurement a second time.

Some old oscilloscopes had a real digital voltmeter coupled to the vertical inputs with real digital voltmeter performance.
 

Offline Performa01

  • Frequent Contributor
  • **
  • Posts: 799
  • Country: at
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #57 on: June 20, 2019, 07:03:12 am »
The documentation does not even discuss the difference and then of course there is the standard deviation calculation of the measurements so that is three different names for the same calculation on different data.  I think standard deviation for automatic measurements and AC RMS and DC RMS for acquisition record measurements by both the DSO and digital multimeter would be less confusing.  As evidence I present Dave having to go back and make the noise measurement a second time.
You're right, the documentation could certainly be improved.

Apart from that, it should be obvious that Siglent is heading more in the direction of LeCroy (and R&S), so we certainly shouldn't confuse professionals by arbitrarily changing names of measurements they have long been familiar with. Not after all that time anyway.

I take it that professionals would know that standard deviation is the same as RMS without the DC component and that RMS includes DC by definition.

The misleading terms have been widespread for DMMs as well, with names like "True RMS". What's a "true" RMS (as opposed to just RMS)? Well, I remember at least one company that used RMS for Standard Deviation and "True RMS" for RMS.

Some old oscilloscopes had a real digital voltmeter coupled to the vertical inputs with real digital voltmeter performance.
Yes, we know that.

But these options used to be separate devices in a special housing, bolted on top of scopes that were prepared to display the DMM measurements. I think I've even seen some with their own display, back in the days when scopes didn't support alphanumeric readout on the CRT. In any case they had individual 4mm sockets. So pretty much the same as a scope meter, the latter with good portability for service technicians - and that certainly makes sense.

Apart from students on a tight budget, who would like to have as many features for as little money as possible in a single device, I cannot imagine anyone wanting a bolted-on or even integrated solution in a scope over a proper separate and independent device in a lab.
« Last Edit: June 20, 2019, 07:05:19 am by Performa01 »
 

Offline tinhead

  • Super Contributor
  • ***
  • Posts: 1901
  • Country: 00
    • If you like my hacks, send me a donation
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #58 on: June 20, 2019, 09:23:13 am »
- External brick (forget the SAG1021.. there's going to be the SAG1021I where "I" stands for isolated, +/-42V of isolation

giving the problematic offset with SAG1021, Siglent should replace all already sold devices by SAG1021I.
I don't want to be human! I want to see gamma rays, I want to hear X-rays, and I want to smell dark matter ...
I want to reach out with something other than these prehensile paws and feel the solar wind of a supernova flowing over me.
 

Online EEVblog

  • Administrator
  • *****
  • Posts: 29468
  • Country: au
    • EEVblog
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #59 on: June 20, 2019, 09:45:47 am »
The misleading terms have been widespread for DMMs as well, with names like "True RMS". What's a "true" RMS (as opposed to just RMS)? Well, I remember at least one company that used RMS for Standard Deviation and "True RMS" for RMS.

"True RMS" came about because all meters on AC are calibrated to display an RMS value, but one without "True RMS" are only valid for pure sine waves.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: EEVblog #1220 - Siglent SDS5000X 1GHz Oscilloscope Review
« Reply #60 on: June 20, 2019, 04:57:16 pm »
Some old oscilloscopes had a real digital voltmeter coupled to the vertical inputs with real digital voltmeter performance.

Yes, we know that.

But these options used to be separate devices in a special housing, bolted on top of scopes that were prepared to display the DMM measurements. I think I've even seen some with their own display, back in the days when scopes didn't support alphanumeric readout on the CRT. In any case they had individual 4mm sockets. So pretty much the same as a scope meter, the latter with good portability for service technicians - and that certainly makes sense.

No No No!  I am referring to oscilloscopes where the DVM was coupled directly into the vertical input and measurements are made through the oscilloscope probe!

The example below is from the Tektronix 2236.  Coupling to the DVM occurs immediately after the high impedance buffer through R20 as marked.  Further, the state of the input attenuators, probe coding (x1 or x10), and input coupling (AC or DC), are also forwarded to the DVM as marked.  (1)

Tektronix produced a special x10 probe for this model to preserve the accuracy of the DVM since standard oscilloscope probes are typically no better than 1% and are often worse.  Tektronix also made probes which could be trimmed for DC accuracy but except for the ones LeCroy inherited from Preamble and is discontinuing, I know of no such thing being made now.

(1) The vertical input coupling selection literally determines whether the DVM makes an AC RMS or average DC measurement and the schematic is even marked to indicate this.  Ambiguity was not a virtue back then.  The RMS converter is AC coupled to remove any DC offset from the internal circuits.

"True RMS" came about because all meters on AC are calibrated to display an RMS value, but one without "True RMS" are only valid for pure sine waves.

I would have said this if you had not.

"AC" makes sense, "RMS calibrated" makes sense, and "True RMS" makes sense.  But "RMS" alone sometimes confusingly could mean either or even AC + DC RMS as we have discovered.
« Last Edit: June 20, 2019, 05:03:10 pm by David Hess »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf