Siglent SDS2000X Plus

[Martin72]

Today at work, before I decided to take the SDS2104X+ with me, I had started with 5ns/div.
But I have to repeat that, as I said, too distracted.
Here is an example:
20ns/div: Sinc:
20048wfms/s, vector.
20139wfms/s dot.
x:
71103wfms/s dot.
20164wfms/s vector.

[BillyO]

Acclimating the SDS2104X+, maybe tomorrow when time allows I could measure the wfms/s.

It's not going to like being that close to the Rigol!!

[Martin72]

Here the tablet for one channel active, normal acquisiton.
Maybe I could make another table with 2 ch active, depends on the free time and monday I must took the scope back to work.
Wfms/s were measured with the DHO804, avg values taken.
Scope was bought in 2022, firmware is not the newest but in my opnion the last two after this version wouldn´t have any effect.

[rf-loop]

Here the tablet for one channel active, normal acquisiton.
Maybe I could make another table with 2 ch active, depends on the free time and monday I must took the scope back to work.
Wfms/s were measured with the DHO804, avg values taken.
Scope was bought in 2022, firmware is not the newest but in my opnion the last two after this version wouldn´t have any effect.

Are you sure it measure whole average or inside acquisition burst speed average. I ask because your numbers are higher than my. My numbers are always whole long term average until noted some other.

My data from SDS2000X Plus (HW05-05, FW1.5.2R2) was for 50ns/div input 600mVpp 5MHz, zero offset.
Dots, int x: 120 500 wfm/s 
Vect, int Sinc or x: 19 540 wfm/s
(inside burst averages are higher)

Example when I measure wfm/s values I use external oscilloscope but also external counter (usually HP53131A)  and enough long gate.

[Martin72]

Hi,

I used the rigol scope to measure and then the average value from its measurement statistics.
I don't know to what extent you can trust the frequency measurement of the rigol, as it is only burst packets that come from the trigger out of the siglent.
As a counter I would have the function of the SDG2122X, I would have to see if the gate can be set to a corresponding length.

[Martin72]

STB-3 I²C packet flaws reported to HQ. 

Anything new, Rob ?

[2N3055]

One way to measure wfms/s is to put other scope at 100ms/div and measure number of pulses..

[rf-loop]

And one  other way is use oscilloscope itself.
Normal always background working history buffer have time stamps.

[Martin72]

Sounds very interesting, must have a look at it on the forthcoming weekend when I could take the SDS2104X+ back at home again.
We have six of them, but they are constantly in use, so I have to take it back to work tomorrow.
Too bad.

[gwideman]

If you change the timebase the trigger position will off the display or closer to the 0s position but this can be addressed by using a fixed position setting where you can place it anywhere and it will remain there at any timebase setting. Utility>Reference Pos

Thanks so much for posting that note. I was sure that my SDS2014X Plus could do this, but for the life of me couldn't discover how to do it.

[Performa01]

One way to measure wfms/s is to put other scope at 100ms/div and measure number of pulses..

Exactly. It should be pointed out, that frequency measurement on the scope is not the same as a frequency counter - and even the trigger frequency counter is different. You can get totally confusing results by using the scope's frequency measurements.

To avoid this, you should use a 100 ms/div timebase, thus implicitely defining a 1 second time gate, and then measure the number of periods - this is your true frequency as it would be measured by a frequency counter at 1 second gate time.

Frequency measurement on the other hand has different strategies depending on the model. The bottom models like SDS1000X-E would still yield usable results because they do just what I've described above: they count the numbers of edges and divide it by the time span corresponding to the record length.

Higher class instruments - those that are able to measure CCJ (Cycle to Cycle Jitter) - have a different strategy; they measure each individual period separately and then provide the mean value of that. Why would there be a noticable difference, you might ask?

At least the Siglent scopes tend to process trigger events not evenly spaced, but there is often a double trigger at maximum speed, followed by a long break. If you simply count all trigger pulses over one second, then you have the true trigger rate. But if you use the frequency measurement of e.g. an SDS2000 for this task, the instrument sees one very short cycle followed by a looong cycle and presents the average of the two. thus you will get the average between the max. trigger rate (>100 kHz) and half the actual trigger rate.

[Performa01]

Here's some demonstration of a common exercise which cannot be solved without deep measurements.

Imagine a 16 bit PWM based on 20 MHz clock frequency. This results in a rather slow 304 Hz PWM signal that can resolve 65536 different levels of duty cycle. To analyze this, we should be able to have accurate time measurements with at least 0.001525878 % resolution

Let’s see if the SDS2000X Plus can do the job.

Timebase is set to 500 µs/div, so that we can capture at least one full PWM period.

At 2 GSa/s, this results in 10 Mpts record length.

Let’s use a PWM signal with 1 ns rise time and 0.001% resolution and try various duty cycles.

First the lowest at 0.001 %:

SDS2354X_Plus_Duty_00.001

Near full scale at 99.999 %:

SDS2354X_Plus_Duty_99.999

Half scale at 50.000 %:

SDS2354X_Plus_Duty_50.000

Finally one step higher at 50.001 %:

SDS2354X_Plus_Duty_50.001

As can be seen, the duty cycle measurement is spot on and rock stable (look at the standard deviation).

The period measurement is fairly stable too, with a standard deviation of only 11.9 ps (peak deviation 50 ps).

The Cycle Mean measurements gives an approximation of the resulting voltage level. It is far less precise than the duty cycle measurement though. No wonder – an 8 bit DSO is no precision bench DMM, hence measurement resolutions far below 1 mV are not going to be stable – this also shows in the standard deviation of ~30 µV.

Finally, the rise and fall times are considerably closer to the truth than what could be realistically expected at 2 GSa/s. The rather high peak deviation of ~500 ps already hints on the averaging of many individual measurements finally yielding a reasonably accurate result.

Of course we can get full accuracy for one additional detail like the rise time by adding zoom trace measurements:

SDS2354X_Plus_Duty_50.001_Rise

Now we get a highly accurate rise time measurement. The key for this is to use a timebase faster than 50 ns/div in the zoom window. Now the sinc reconstruction generates additional data points, thus increasing time resolution and reducing the standard deviation of the measurement to just 11.2 ps.

[Warhawk]

Hello everyone,
I've recently upgraded my lab with the SDS2104X plus scope. Additionally, years ago I created my version of the famous Jim Williams's short pulse generator. The main difference is that I just replaced the boost converter with a 555 timer. Back then, I was able to score a couple of the original Motorola transistors (NOS).



I guess it does not surprise anyone that my scope now appears as SDS2504X Plus. My other scope is HP 54825A (used to be 54815A). They both share 500Mhz bandwidth and 2 GSa/s sampling. I thought it would be interesting to compare them side by side. Fist I wanted to get some reference. Probably the best I could get my hands on was the SDA6020 from LeCroy. That is a 6GHz, 20 GSa/s scope. The input is 50 ohm with +/-4Vpp input voltage range. The pulse generator exceeds the range therefore I used 1:10 atenuator. I tried different ones, always with the same result. They were rated 0-3 GHz. The measurement gave me ~200ps rise time. I used Sinc(x) interpolation. In case of LeCroy it also makes sense to look at linear extrapolation but the waveform is pretty similar. I decided for persistence since not every pulse is identical and this gies the best impression on what is happening.




Now, with the reference, I tried my Infinium 54825A.



Finally, the Siglent with the 500MHz option.



I thought someone would find my experiments interesting. Feel free to point out if something is just not right.
Have a nice weekend everyone.


 

[rf-loop]

Also there can see how Siglent trigger accuracy is far ahead of the other two. 

[Performa01]

Also there can see how Siglent trigger accuracy is far ahead of the other two. 

This, which doesn't really come as a surprise, since the Siglent seems to be the only one with a fully digital trigger system in this bunch.

What is a surprise though, is the rather bad pulse fildelity of the Keysight. Don't even want to imagine how much noise some of the well known  "experts" would make if the Siglent showed such a behavior...

[Warhawk]

Also there can see how Siglent trigger accuracy is far ahead of the other two. 

I was definitely impressed how good the triggering mechanism is. I admit I had to "fiddle the knob" for both LeCroy and HP to get reliable trigger.

What is a surprise though, is the rather bad pulse fildelity of the Keysight. Don't even want to imagine how much noise some of the well known  "experts" would make if the Siglent showed such a behavior...

To be fair, both machines for the comparison are old. Still, the roller coaster after the falling edge is interesting. That's the filter response, I guess.<joke> Please don't call it keysight, this unit proudly carries HP logo !</joke>

[Martin72]

Of course there have been advances, but they are not so enormous that one could assume that a 1200€ scope would be on a par with or even superior to premium scopes from the 00s that were on a different planet in terms of price.
Siglent has done a good job with the SDS2000xplus.

[rf-loop]

Please don't call it keysight, this unit proudly carries HP logo !</joke>

R.I.P. Hewlett-Packard. Who assassin it - well...all know,  Lady Fiorina.   

[gf]

Also there can see how Siglent trigger accuracy is far ahead of the other two. 

This, which doesn't really come as a surprise, since the Siglent seems to be the only one with a fully digital trigger system in this bunch.

What is a surprise though, is the rather bad pulse fildelity of the Keysight. Don't even want to imagine how much noise some of the well known  "experts" would make if the Siglent showed such a behavior...

It looks like both of them suffer from aliasing (the 54825A badly, but the Siglent too - maybe a consequence of the 500 MHz upgrade?)
For comparison, the Siglent in points mode would be interesting, too.

Still, the roller coaster after the falling edge is interesting. That's the filter response, I guess.

Sure. If the original impulse is very narrow (so that it can be considered an approximation of a Dirac delta impulse), then the entire curve is the impulse response (but then the rise time you measured is not the step response rise time, but the rise time of the curve's integral would be the step response rise time). OTOH, if the original pulse width is not negligibly short, then it is hard to separate how much of the pulse shape is contributed by the frontend, and how much by the pulse itself.

[Warhawk]

R.I.P. Hewlett-Packard. Who assassin it - well...all know,  Lady Fiorina.   

At least she inspired Ginni Rometty... 

It looks like both of them suffer from aliasing (the 54825A badly, but the Siglent too - maybe a consequence of the 500 MHz upgrade?)
For comparison, the Siglent in points mode would be interesting, too.

Honestly, the point mode from Siglent does not show anything useful. I could not find a way how to show only points and the linear extrapolation interpolation looks silly. I can try with HP but later. Attached is SDA6020 if it helps.

General questions:
* Does SDS2000x+ have a timer that counts seconds waiting for trigger? I am debugging a Zigbee light and the bug happens one every couple hours.
* Isn't there a chance to have measurements window floating? The measurement feature is generally excellent but I wish I could have an overlay or so (see the HP printscreen).

[rf-loop]

Honestly, the point mode from Siglent does not show anything useful. I could not find a way how to show only points and the linear extrapolation looks silly. I can try with HP but later. Attached is SDA6020 if it helps.

It do not have any extrapolation... but interpolation

Display menu. Display mode "Dots" instead of Vectors

In acquire menu there is interpolation method: x or Sinc

If your display mode is selected dots, then it display only ADC samples alone. (And also do not add "interpolated dots" in dots mode as some oscilloscopes may do

[Warhawk]

Honestly, the point mode from Siglent does not show anything useful. I could not find a way how to show only points and the linear extrapolation looks silly. I can try with HP but later. Attached is SDA6020 if it helps.

It do not have any extrapolation... but interpolation

Display menu. Display mode "Dots" instead of Vectors

In acquire menu there is interpolation method: x or Sinc

If your display mode is selected dots, then it display only ADC samples alone. (And also do not add "interpolated dots" in dots mode as some oscilloscopes may do

Thank you for the tips and pardon my ignorant nomenclature. 
I try tomorrow.

[gf]

Honestly, the point mode from Siglent does not show anything useful. I could not find a way how to show only points and the linear extrapolation looks silly. I can try with HP but later.

Looking at the datasheet, the 54825A seems to support Equivalent Time Sampling.

[2N3055]

R.I.P. Hewlett-Packard. Who assassin it - well...all know,  Lady Fiorina.   

At least she inspired Ginni Rometty... 

It looks like both of them suffer from aliasing (the 54825A badly, but the Siglent too - maybe a consequence of the 500 MHz upgrade?)
For comparison, the Siglent in points mode would be interesting, too.

Honestly, the point mode from Siglent does not show anything useful. I could not find a way how to show only points and the linear extrapolation looks silly. I can try with HP but later. Attached is SDA6020 if it helps.

General questions:
* Does SDS2000x+ have a timer that counts seconds waiting for trigger? I am debugging a Zigbee light and the bug happens one every couple hours.
* Isn't there a chance to have measurements window floating? The measurement feature is generally excellent but I wish I could have an overlay or so (see the HP printscreen).

It has always running history mode that keeps time for every trigger event..

[Performa01]

Honestly, the point mode from Siglent does not show anything useful. I could not find a way how to show only points …

Well, let’s examine that a bit.

I don’t have a fancy pulser, but maybe a well defined pulse from a pulse generator, even if it’s a little bit slower at 500 ps rise time, might be even more appropriate for characterizing a moderate bandwidth oscilloscope.

So the test pulse has a repetition rate of 10 MHz, 500 ps transition times and is 1 ns wide.

The first screenshot has been taken with Sinc reconstruction, vector display mode, color grading and 10 s persistence.

SDS2354X_Plus_PR_10MHz_500ps_vect_CG

What do we see ? There are not any signs of aliasing. There is no pre-shot. There is a very moderate low frequency ringing at the end of the falling edge. The signal is rock stable, as the thin trace and the low standard deviation in the time measurements show. The amplitude measurement on the other hand cannot be very stable, as this fast pulse doesn’t have a well defined “roof”

Now let’s try the same thing in dots mode, which is claimed to “not show anything useful”…

SDS2354X_Plus_PR_10MHz_500ps_dot_IG

As can be seen, I’ve used the standard intensity grading this time, as color grading and not even persistence show anything spectacular at all. I wish every feature in any T&M gear were as “useless” as dots mode in a SDS2000X Plus.

Of course, for dots mode to yield nice results as in this picture, we need to have the DSO set to “fast acquisition” and keep it in Run mode, so that SARI (Sequential Acquisitions Random Interleaving) can do its job. With persistence enabled, we can even switch to Stop mode and still have a contiguous trace. A single shot acquisition on the other hand will always allow the closer inspection of the individual samples of a single acquisition:

SDS2354X_Plus_PR_10MHz_500ps_dot_IG_Single

Yes, we don’t see a lot in this picture, but it’s always nice to have the option to see the raw basis the instrument is working on.

Of course we can look at the history any time we like and there we can look at the individual records (up to 80000 at fast time bases like this) and try dots mode to show the true samples only, vector mode, where we can select between x-interpolation and Sinc reconstruction, as well as even persistence, which will collect display data as we browse through the history or play it back at any arbitrary speed.

and the linear extrapolation looks silly. I can try with HP but later. Attached is SDA6020 if it helps.

x-interpolation can avoid Gibbs ears in some situation, but is usually just an obsolete leftover from the old days of digital oscilloscopes, when there wasn’t enough processing power to run a true reconstruction (Sinc aka sin(x)/x). It is evident, that the sample rate should be about ten times the input (or signal) bandwidth for linear interpolation to work.

In our particular case, linear (x) interpolation isn’t terribly useful, yet it gives a nice demonstration of the intensity grading:

SDS2354X_Plus_PR_10MHz_500ps_vect_IG_Lin