Functional comparison of R&S RTB2000, Siglent SDS2000X and Keysight DSOX1000

[RBBVNL9]

Dear seronday,

It is possible to transfer a waveform from one of the vertical channels to the Arbitrary Waveform Generator.

This is done from the vertical channel menu dialog box.
One of the menu items is  " Apply To ".
Selecting " Apply To ", brings up a list of options, one of these is ARB.
This copies the current channel waveform to the Arbitrary Waveform Generator.

You can then go to the Arbitrary Waveform Generator and select "Stored" and then " Channel' to output the waveform.

That is wonderful! I just tested it and it indeed works. In fact, the very moment I tick "Apply To" in the vertical channel menu the wave appears on the output of my AWG (at least when it is turned on), so there seems to be no need in the AWG menu to go to "Stored" and "Channel". The latter only seems relevant if you in-between switch to another AWG waveform and want to go back to the captured one again. 

Just for the sake of completeness: a week before I said in my video I did not get this to work, I did already posted my problem in the EEVBLOG forum but no one seemed to be able to provide a solution at that time...

But you are the one who provides the solution so compliments go out to you!

This is mentioned in the User Manual in section 12.2, Channel Setup,  but the way it is explained is somewhat confusing.

True. In the last weeks, I went through the manual as well as all the device menu's quite carefully but missed this. Even searched the manual on keywords "ARB" and "Arbitrary' with no success, but I see in Section 12.2 they do not use these terms... Anyway, this information should really have been in the part of the manual that describes the ARB (like Section 28.3)! 

[pdenisowski]

Dear Egonotto,

I think they mean the frequency is max 10 MHz and 250 Ms/s.

Yes, that does make sense. The sample rate indeed needs to be more way than the 10 MSa/s R&S mention in the specifications in order to output the waveform you show (and assuming it would be the same on the RTB as the RTA). I agree with you it is likely the same 250 MSa/s as used for the regular waveform generation. Will update my overview document. Thanks for checking!

Hi Rudy - Not 100% sure about the question here, but the waveform generator in the RTB2000 (option RTB-B6) does have a sample rate of 250 MSamples/sec (as per the specifications)

https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTB2000_dat-sw_en_3607-4270-22_v1600.pdf

Incidentally, I just did a video walkthrough on the basics of using the RTB2000 function generator.

[RBBVNL9]

Dear pdenisowski, thanks for the reply.

The discussion above is (specifically) about the sample rate for the arbitrary wave function.

For this, the RTB2000 specifications (V.1600) you refer to mention: "Sample rate: max. 10 Msample/s" and "Memory depth: 16k points"  (see screenprint attached, yellow highlights)

Document R&S RTx-B6 ARBITRARY WAVEFORM AND 4 BIT PATTERN GENERATOR (Version 01.10, May 2020), for the RTB series, reads "Signal forms frequency ranges (arbitrary): max. 10 Msample/s; 32k points" (see screenprint attached, yellow highlights)

I think that in both documents, the 10 Msample/s is an error and should probably be 250 MSa/s (otherwise it could not generate the waveforms it actually does in practice!). Moreover, in the specifications, the 16k points is probably an error and should be 32k, as confirmed by tests in practice.

PS thanks for the walkthrough video, just looked at it!

[RBBVNL9]

Preparing for a new video on observing infrequent events and mask tests, and seeing huge differences there, so I thought I should first dig a bit more into actual triggering behaviour.

I see something I am not sure I understand well.

In short:
-   I feed all oscilloscopes with a 1kHz square wave on channel 1.
-   I set the horizontal time base such that I see two periods on the screen (200uS/DIV on the SDS and DSOX, and 170uS/DIV on the RTB as it has 12 instead of 10 horizontal divisions). Attaching a pic.
-   I activate the trigger out on each device and look at these on a fourth oscilloscope (a PicoScope 3405D). Channels are 1: RTB, 2: SDS, 3: DSOX and 4: input square to scopes.

All scopes are set to regular trigger settings (trigger on positive edge, level halfway square, no holdoff, DC coupling, no noise reject or filter). Record length / memory depth is chosen for best results, if there is any difference. Segmented acquisition off. Auto trigger or normal trigger makes no difference on any of the devices. Lastly, on the RTB I set the trigger out a pulse to 1mS to make it well visible (using SCPI command TRIGger:OUT:PLENgth 1E-3).

Ideally, I would expect to see a trigger every one out of three periods (where the positive edges of the two other periods are shown on the screen), so a constant 333.3 pulses per second on the trigger out bus. After all, 1kHz is such a slow signal and any eventual blank time these scopes need to write to memory etc. should be neglectable.

The results are in the attached screen print.
-   The Keysight DSOX behaves exactly as expected, triggering every third period.
-   The Rohde & Schwarz RTB mostly every third period but there are some (predictable) interruptions. Is the scope doing something else every once in a while ?!? 
-   The Siglent SDS triggers much, much less. Only 30 pulses per second instead of the expected 333.   

Can anyone enlighten me? Why does the RTB have periodic interruptions? And, more importantly, why is the SDS so slow to re-trigger ?!?
Do I overlook relevant device settings?

Thanks for your insights!

[egonotto]

Hello,

my conjecture: RTB and SDS take time to execute the samples. Meanwhile they do not sample.

RTA behave little wilder.

Best regards
egonotto

[Someone]

Can anyone enlighten me? Why does the RTB have periodic interruptions? And, more importantly, why is the SDS so slow to re-trigger ?!?
Do I overlook relevant device settings?

Double check the memory depth you have set, it often has a large effect on the update rate. Not sure if the SDS 2000X will approach "perfection", older models were similar to what you see:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg3169060/#msg3169060
(no idea if there is a set of measured data for the 2000X, so hard to search through the mountains of information on this forum).
If you look very carefully over a long period even on the Keysight there will be some gaps in the triggers (or use a pulse width trigger to catch it!).

Edit, found some measured values for the 2000X:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg3133558/#msg3133558
Not fully explained the methodology, but suggests better performance is possible in some configuration/situation.

[RBBVNL9]

Thanks egonotto, Someone and PeDre for taking the time to respond so swiftly. Some quick replies:

Double check the memory depth you have set, it often has a large effect on the update rate.

I tried each memory depth setting available in the SDS (see my post) but none of made a difference in seeing this pattern. But will do some more attempts later. (This setting did make a difference for the RTB though.)

If you look very carefully over a long period even on the Keysight there will be some gaps in the triggers (or use a pulse width trigger to catch it!).

That might be the case. But as can be seen from the PicoScope screen print, the Keysight DSOX’s average of 333.3 trigger out pulses per second was very stable; over 109 observations (of each ~150 pulses) there is an extremely low standard deviation. So, if it misses triggers, it’s not many. Indeed, a pulse width trigger (or mask test) could reveal this.

Edit, found some measured values for the 2000X:

Like in the post of Martin72 you refer to, I am indeed finding similar, very low waveform update rates on the SDS (orders of magnitude lower than advertised). That is the reason why I started to do these trigger experiments…

This has been asked before, here is Rich's answer:

Thanks for forwarding Rich’s answer. Before posting, I did a search on the forum but it’s not always easy to find what you are looking for, even if it’s there  :-| But anyway, questions answered!

RTB and SDS take time to execute the samples. Meanwhile they do not sample.

Yes, seems so. Perhaps the reason the DSOX is doing so well here is that it dedicated these tasks to different hardware resources so it can keep on triggering even when it’s buffering and plotting on the screen. 

RTA behave little wilder.

Can you perhaps elaborate on what you mean here?

Bottom line: While it’s now clear to me what the RTB is doing, the bigger concern here is the SDS. After a trigger, it misses something like 10 possible triggers before its ‘arming’ again. For a trigger signal as slow as 1 kHz, this is quite strange. Perhaps, like Martin72 suggests, there might be a serious bug here that is still waiting to be fixed?

[Someone]

RTB and SDS take time to execute the samples. Meanwhile they do not sample.

Yes, seems so. Perhaps the reason the DSOX is doing so well here is that it dedicated these tasks to different hardware resources so it can keep on triggering even when it’s buffering and plotting on the screen.

Bottom line: While it’s now clear to me what the RTB is doing, the bigger concern here is the SDS. After a trigger, it misses something like 10 possible triggers before its ‘arming’ again. For a trigger signal as slow as 1 kHz, this is quite strange. Perhaps, like Martin72 suggests, there might be a serious bug here that is still waiting to be fixed?

I do not think it is a bug, but just the way they work. As you say above these different scopes have different hardware separation of the work, DSOX puts most things in ASIC/FPGA which makes it fast but less flexible (examples: cannot reduce memory depth, or turn off interpolation, even though that would make it faster). Probably Siglent do much more of the display drawing in CPU/software for more flexibility/cheaper product (example benefit: colourised intensity view).

[RBBVNL9]

I do not think it is a bug, but just the way they work.

Well, even if these devices do work differently, Siglent does publish a waveform update rate. And if the actual update rate is orders of magnitude lower than what is published (see the posts of Martin72 and Performa01 referred to above, which are indeed in line with my own measurements, see my video Episode 7 at 47:27) then I think it is correct to speak of a bug... 

[2N3055]

Bottom line: While it’s now clear to me what the RTB is doing, the bigger concern here is the SDS. After a trigger, it misses something like 10 possible triggers before its ‘arming’ again. For a trigger signal as slow as 1 kHz, this is quite strange. Perhaps, like Martin72 suggests, there might be a serious bug here that is still waiting to be fixed?

Hi.

Could you check something on Siglent? If you go into Acquistion  menu, do you have Slow or Fast mode set there? You need to be in the dot mode also..

A small suggestion, if I may. To avoid confusion, may I suggest a wording specificum: a time that scope needs to be ready for a new trigger event can be called retrigger time.  We can also use a term rearm time or even a blind time to signify time that scope is "blind" to new events while trigger engine is being rearmed for another go.  I say that because saying that "scope is missing triggers" can be misconstrued as scope being ready for trigger but it didn't recognize it properly.
Difference is that first one is operating specification of the scope and the other one (missed triggers) is defect, a bug.

A bit of pedantry, I know, but nevertheless not unimportant sometimes.

As Someone correctly said, phosphor emulation works by virtue of sampling short bursts of acquisitions synchronised with screen refresh rate. Keysight also does it on certain timebase settings. Even KS 3000T that is clocked at more than 1 milion triggers per second does it at some settings.
But this is nothing new, really. This is pretty much only thing KS does better in Infiniivision scope series (Megazoom IV based). Whole architecture is based around this feature. With all the compromises that stem from it, like very small memory, use of decimated buffers for all measurements, all time interpolation with no user control etc. These are pretty much speciallistic scopes made to emulate something similar to analog CRT scope triggering performance as primary design goal.

Both R&S and Siglent, with their long memory architecture are closer to "analytic" scope type, where you capture longer sequences and then analyse it. In fact, we can argue that these scopes with long memory can achieve "zero" blind time for bursts of, say, 100ms. This way of thinking, though, does need for user to actually adjust way how they are using scopes, because it is slightly different to how you would use CRT scopes many people are used to.
That is why I always keep repeating that, for instance, KS 1000X series are not very capable scopes (they have very limited capabilities compared to even scopes many times cheaper) but are good for people that don't need advanced features (user wants to to only look at waveform on the screen and maybe use cursors and basic measurements) and that want as good as possible CRT emulation.

[Someone]

I do not think it is a bug, but just the way they work.

Well, even if these devices do work differently, Siglent does publish a waveform update rate. And if the actual update rate is orders of magnitude lower than what is published (see the posts of Martin72 and Performa01 referred to above, which are indeed in line with my own measurements, see my video Episode 7 at 47:27) then I think it is correct to speak of a bug...

I have not seen published/advertised waveform rates for different timebases, do you have a link/copy? Only value in the advertising/marketing is the peak:

Waveform capture rate up to 140,000 wfm/s (normal mode), and 500,000 wfm/s (sequence mode)

We usually only see these figures across many acquisition window sizes in "competitive comparisons" or user generated figures (such as on this forum).

Many scopes have advertised "headline" or show high rates in unusual/non-typical modes such as very short memory depth and/or interpolation disabled, getting lower rates with normal memory depth and interpolation on is not wrong/bug but normal/expected.

[dreamcat4]

so if this part of the series is all about detecting unusual events or glitches then...

it makes me wonder a lot how do the SDS and RTB compare to lecroy system. because in those products it seems like the most ideally executed feature. where it automatically recognized in the buffer any 'unclean' logic transition. or other such intermittent excursions. you just press a button and the scope does all of the work for you. and brings it up on the display

this feature makes me wonder what is the right lecroy scope to compare to the RTB. for  example if it includes that specific feature at a low enough price point to be an alternative option to the RTB 2004

[rf-loop]

I do not think it is a bug, but just the way they work.

 And if the actual update rate is orders of magnitude lower than what is published <clip>  then I think it is correct to speak of a bug...

In Siglent SDS2000X Plus data sheet (DS0102XP_E01A)

Waveform capture rates up to 120,000 wfm/s (normal mode) and 500,000 wfm/s (sequence mode)

Waveform capture rate (Max.)
Normal mode：120,000 wfm/s;
Sequence mode：500,000 wfm/s

Waveform capture rate
Normal mode: 120,000 wfm/s max.
Sequence mode: 500,000 wfm/s max.

And same oscilloscope but now Teledyne LeCroy brand.
Teledyne test tools T3DSO2000A Data Sheet, datecode 22july20

Key Features
New generation of high speed display technology
› Waveform capture rate up to 120,000 wfm/s
(normal mode), and 500,000 wfm/s (sequence
mode)

Waveform Capture Rate (Max.) 120,000 wfm/s (normal mode), 500,000 wfm/s (sequence mode)

Waveform Capture Rate Up to 120,000 wfm/s (normal mode), 500,000 wfm/s (sequence mode)

(bolded by me)

What is wrong now. Is it difficult to uderstand "up to"  or "max"
If this is difficult I think it is education problem or lack of enough knowledge and experience. I can predict this kind of problems may lead to severe problems when need read what ever  instruments or electronics, mechatronics, etc data sheets.  Is it correct to say... bug in understanding data sheets. 

[Someone]

do you have Slow or Fast mode set there? You need to be in the dot mode also..

A small suggestion, if I may. To avoid confusion, may I suggest a wording specificum: a time that scope needs to be ready for a new trigger event can be called retrigger time.  We can also use a term rearm time or even a blind time to signify time that scope is "blind" to new events while trigger engine is being rearmed for another go.  I say that because saying that "scope is missing triggers" can be misconstrued as scope being ready for trigger but it didn't recognize it properly.
Difference is that first one is operating specification of the scope and the other one (missed triggers) is defect, a bug.

First I think dot mode is not a normal situation to use a scope when wanting to look at waveforms! Its these sorts of "games" that can make comparisons silly, always best to try and find a common setting that all products can meet (and mention that they can do better in their special/preferred setting).

A problem with blind time as a measure is that it is not a constant/deterministic value in most (all?) scopes when in realtime mode, it may be accurate for sequence/segmented modes where nothing is drawn to the screen.

[dreamcat4]

what is the right lecroy scope to compare to the RTB. for  example if it includes that specific feature at a low enough price point to be an alternative option to the RTB 2004

ah, i remember now Martin72 had one...

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4054573/#msg4054573

it was the WS3024Z

[Someone]

What is wrong now. Is it difficult to uderstand "up to"  or "max"
If this is difficult I think it is education problem or lack of enough knowledge and experience. I can predict this kind of problems may lead to severe problems when need read what ever  instruments or electronics, mechatronics, etc data sheets.  Is it correct to say... bug in understanding data sheets. 

Peak value does not predict others! yes. A previously popular scope claimed best waveform rate but in real world performed badly:
https://www.eevblog.com/forum/testgear/rigol-mso4000-and-ds4000-tests-bugs-firmware-questions-etc/msg973064/#msg973064

[tautech]

what is the right lecroy scope to compare to the RTB. for  example if it includes that specific feature at a low enough price point to be an alternative option to the RTB 2004

ah, i remember now Martin72 had one...

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4054573/#msg4054573

it was the WS3024Z

Only LeCroy had rights to market SDS3000X (WS3000Z) DSO's to western markets and SDS3000 (WS3000) models before them as their development was a joint venture.

[2N3055]

what is the right lecroy scope to compare to the RTB. for  example if it includes that specific feature at a low enough price point to be an alternative option to the RTB 2004

ah, i remember now Martin72 had one...

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4054573/#msg4054573

it was the WS3024Z

AFAIK Martin has WS3024 and WS3024Z at work (among other LeCroys). At home he has SDS2104X+..
They recently bought Siglent SDS2104X+ at work too and it seems it fits nicely among its LeCroy "distant relatives"..

[tautech]

what is the right lecroy scope to compare to the RTB. for  example if it includes that specific feature at a low enough price point to be an alternative option to the RTB 2004

ah, i remember now Martin72 had one...

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4054573/#msg4054573

it was the WS3024Z

AFAIK Martin has WS3024 and WS3024Z at work. At home he has SDS2104X+..

And now at work too !
The lad should be paid a commission. 

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4067635/#msg4067635

[RBBVNL9]

Thanks all for digging so deep into this. And thanks to 2N3055 for suggesting clear terminology. I noted that sometimes specifications sheets are also not super clear on this. The SDS2000X+ spec sheets talk both of “waveform capture rate” and “waveform update rate”. They seem to mean the same thing with these two terms, but I’m not 100% sure and find it confusing.

And yes, the specifications of Siglent use terms such as “up to” or “max" (and they are not the only ones *). Not knowing under which conditions these maxima are reached, it’s hard to decide whether a device meets the specs, and when we can speak of a bug or not.

Having that said, if in a non-trivial case (triggering a simple 1kHz square wave) a device has a score that is orders of magnitude lower than what one would expect (and is orders of magnitude lower than other devices) I do think it’s a relevant thing for users to know about.

so if this part of the series is all about detecting unusual events or glitches then...

I know I called for input on this topic, but I should also mention that my comparison is certainly not aiming to focus on this aspect only, it’s one among many aspects I’m liking into.
 

* I love, however, the way Keysight provided its waveform update specs: “≥ 200,000 waveforms/sec” on page 12 of their Keysight InfiniiVision 1000 X-Series Oscilloscopes Data Sheet

[rf-loop]

Preparing for a new video on observing infrequent events and mask tests, and seeing huge differences there, so I thought I should first dig a bit more into actual triggering behaviour.

I see something I am not sure I understand well.

In short:
-   I feed all oscilloscopes with a 1kHz square wave on channel 1.
-   I set the horizontal time base such that I see two periods on the screen (200uS/DIV on the SDS and DSOX, and 170uS/DIV on the RTB as it has 12 instead of 10 horizontal divisions). Attaching a pic.
-   I activate the trigger out on each device and look at these on a fourth oscilloscope (a PicoScope 3405D). Channels are 1: RTB, 2: SDS, 3: DSOX and 4: input square to scopes.

All scopes are set to regular trigger settings (trigger on positive edge, level halfway square, no holdoff, DC coupling, no noise reject or filter). Record length / memory depth is chosen for best results, if there is any difference. Segmented acquisition off. Auto trigger or normal trigger makes no difference on any of the devices. Lastly, on the RTB I set the trigger out a pulse to 1mS to make it well visible (using SCPI command TRIGger:OUT:PLENgth 1E-3).

Ideally, I would expect to see a trigger every one out of three periods (where the positive edges of the two other periods are shown on the screen), so a constant 333.3 pulses per second on the trigger out bus. After all, 1kHz is such a slow signal and any eventual blank time these scopes need to write to memory etc. should be neglectable.

The results are in the attached screen print.
-   The Keysight DSOX behaves exactly as expected, triggering every third period.
-   The Rohde & Schwarz RTB mostly every third period but there are some (predictable) interruptions. Is the scope doing something else every once in a while ?!? 
-   The Siglent SDS triggers much, much less. Only 30 pulses per second instead of the expected 333.   

Can anyone enlighten me? Why does the RTB have periodic interruptions? And, more importantly, why is the SDS so slow to re-trigger ?!?
Do I overlook relevant device settings?

Thanks for your insights!

Please can you clarify every oscilloscope sampling speed and current acquisition true memory length used in this image.

[dreamcat4]

ah ok. but  then does that lecroy 3000 series actually have the feature in question? if it is actually a rebranded siglent and not a true windows based lecroy scope?

the  feature seems to be able to detect any arbitrary random glitches or anomalies. such as the failed or unclean logic transitions etc

[2N3055]

First I think dot mode is not a normal situation to use a scope when wanting to look at waveforms! Its these sorts of "games" that can make comparisons silly, always best to try and find a common setting that all products can meet (and mention that they can do better in their special/preferred setting).

A problem with blind time as a measure is that it is not a constant/deterministic value in most (all?) scopes when in realtime mode, it may be accurate for sequence/segmented modes where nothing is drawn to the screen.

I agree with you in general. But fact is that SDS6000H12 works well in dot mode. As soon as you have 1000 pixels horizontally, it looks like continuous line, without any interpolation artefacts. Retriggered events overlay on top and you get SARI, a RIS alike random repetitive sampling. So I keep jumping between line and dot mode.

But making clear specifications is important. It's a shame not even big ones are doing it right. There is a Keysight whitepaper where they compare 3000T wfms/s with competition, carefully choosing test to favour Keysight. And forgetting to mention that much more expensive Keysight scopes have like 100 Wfms/s trigger rate because they have large memories and are doing full buffer managements on large datasets.
They also mention casually memory sizes, but 3000T has only 500k of sample memory when doing 4ch+ digital normal mode (4 buffer-2 per ch-1 with digital -0.5  Mpts for ping pong buffers. 1Mpts for Single mode). Scope with 100Mpts will be 100x slower everything else being equal. Scope with 500MPts will have soo much more work to do.

To make it short, Keysight will have faster retrigger rate because it is specifically designed to do so. Siglent was made to work LeCroy way, and those don't maximize raw retrigger rate but analytic capabilities in long memory.

[2N3055]

ah ok. but  then does that lecroy 3000 series actually have the feature in question? if it is actually a rebranded siglent and not a true windows based lecroy scope?

the  feature seems to be able to detect any arbitrary random glitches or anomalies. such as the failed or unclean logic transitions etc

Not even windows based Wavesurfers have that feature fully implemented (parametric triggers).

But you have advanced triggers on Siglent Touch series and also always running (if enabled) search function that you can parametrize to look for several waveform parameters. Those can detect runts, non monotonic edges, slow edges, pulse widths out  of spec etc... Combined with advanced triggers it has (including zone triggers), measurements and history mode there are many things that can be detected.

[2N3055]

Thanks all for digging so deep into this. And thanks to 2N3055 for suggesting clear terminology. I noted that sometimes specifications sheets are also not super clear on this. The SDS2000X+ spec sheets talk both of “waveform capture rate” and “waveform update rate”. They seem to mean the same thing with these two terms, but I’m not 100% sure and find it confusing. .....

Just to make it clear, it wasn't a critique to you, but a suggestion to actually address that problem you're mentioning: a nonuniform language that confuses us all. If we agree to common language, we understand each other better..