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

[RBBVNL9]

>> The undocumented SCPI command can be used to read out the FFT window factor:

Very nice, thanks! Now we know the exact values the instruments uses and plug these into our calculations.

[Performa01]

>> The undocumented SCPI command can be used to read out the FFT window factor:

Very nice, thanks! Now we know the exact values the instruments uses and plug these into our calculations.

Yes - whatever it's worth.

Maybe you want to consider my analysis presented in reply #3561 in the following thread:

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4318822/#msg4318822

One might come to the conclusion that these factors are all very nice (and different, depending on the source), but rather meaningless in practice - with one major exception...

[RBBVNL9]

All true, but I find them useful to reconstruct the RTB FFT method / settings ;-)

[RBBVNL9]

In my last video, the Rohde & Schwarz RTB2004 was still holding secrets in terms of how its FFT analysis actually work. In this video, I go into greater detail unravelling them.

Thanks to various people that contributed to this thread with useful suggestions and ideas!

https://youtu.be/D1dVqcjbWm4

[kerouanton]

In my last video, the Rohde & Schwarz RTB2004 was still holding secrets in terms of how its FFT analysis actually work. In this video, I go into greater detail unravelling them.

Thanks to various people that contributed to this thread with useful suggestions and ideas!

https://youtu.be/D1dVqcjbWm4

I spent the whole evening (and partly night) watching your videos, reading the thread and your comparison document, that's a huge work, so thank you @RBBVNL9.

Wondering if you plan to continue updating the document or releasing new videos? I also watched your other recent videos on Anritsu hardware (wow, that's high-end gear!), but I was wondering if you were dropping this comparison topic.

Again many thanks.

[ebastler]

May I revisit this older thread? I have just started to work my way through Rudi's very helpful series of videos -- mainly with an interest in the hands-on demonstration of the SDS2104X plus, which I'm  seriously considering as my next scope.

One thing that struck me in episode #4 (Measurements) is that the SDS came out last, by an order of magnitude (!), regarding the speed of collecting measurement statistics. See the attached table and the video linked below. Given how the strength of quantitative measurements and analysis is always stressed when discussing the Siglent scopes (and their LeCroy role models), I had not expected this.

I did not find any comments on this in the present thread. Is the slow measurement update rate just a reality with the SDS2000X plus? Are there some memory or acquisition settings that should be changed to get much faster measurements? Has anything been improved in recent firmware updates?

Thanks for your comments!

[Performa01]

May I revisit this older thread? I have just started to work my way through Rudi's very helpful series of videos -- mainly with an interest in the hands-on demonstration of the SDS2104X plus, which I'm  seriously considering as my next scope.

One thing that struck me in episode #4 (Measurements) is that the SDS came out last, by an order of magnitude (!), regarding the speed of collecting measurement statistics. See the attached table and the video linked below. Given how the strength of quantitative measurements and analysis is always stressed when discussing the Siglent scopes (and their LeCroy role models), I had not expected this.

I did not find any comments on this in the present thread. Is the slow measurement update rate just a reality with the SDS2000X plus? Are there some memory or acquisition settings that should be changed to get much faster measurements? Has anything been improved in recent firmware updates?

Thanks for your comments!

You have already named it – this is a MSO with main focus on analytic capabilities. Of course there is a difference between analytic scopes, which require deep measurements, and standard scopes, which prioritize speed and try to measure (estimate would be more appropriate) on heavily decimated data. To understand this, here’s a simple example:

SDS2354X_Plus_Meas_Pulse_10Mpts

This screenshot shows just a pulse train. Pulse width is 10 µs, transition times are ~1 ns and repetition rate is 4 kHz. At a time base of 500 µs/div and 2 GSa/s, this results in a record length of 10 Mpts. The key parameters of the pulse train shall be measured: period, amplitude, rise and fall times and pulse width. Everything is measured reasonably correctly. Of course, at 2 GSa/s this scope isn’t fast enough to measure a 1 ns transition time precisely, yet it’s well within reasonable expectations. For more accurate measurements on such short time intervals, we’d have to zoom in up to the point, where sin(x)/x reconstruction calculates additional data points. We cannot use a much faster timebase right from the start, because at faster time bases we lose the measurements, one by one: At < 25 µs/div we could not measure the period anymore, at <1 µs/div we lose the pulse width measurement and can only measure a single transition, I.e. either the rising or the falling edge.

Now try the same exercise with the fast Keysight, which uses 64 kpts decimated data – good luck!

It is worth to take a closer look at this screenshot. We see advanced measurements in M1 mode, where statistics and Histicons (small histogram icons) are enabled. If we look at the “count” in the measurement statistics, it becomes obvious that only the vertical (amplitude) related measurements appear to be slow. For the time measurements, the scope analyzes the entire record, hence a single acquisition yields some 20 measurements of period, rise and fall times and pulse width. It just so happened that not a single time related measurement was shown in this comparison.

Apart from that the SDS2000X Plus firmware has seen a few optimizations since this comparison. My measurements yielded at least twice the speed as the numbers found in this review. If I make the conditions a bit more equal, then the vertical measurement rate reaches >16 measurements per second if I reduce the record length to 100 kpts, thus getting into the ballpark of the measurements on decimated data of the competing instruments.

Furthermore we have these M1 and M2 measurement modes (limited to max. 12 measurements) only in the “advanced measurements” mode. But there is also a “simple” mode, which allows a lot more than just 12 measurements in parallel:

SDS2354X_Plus_Meas_Simple

You should be able to count not 6, also not 12 but no less than 52 measurements at once. The glitch, that the measurement window closes as soon as we select an item (so that we cannot read the help text and have to re-open the window if we need to add more measurements) is long fixed either.

[ebastler]

Thank you, Performa01 -- carrying out measurements on the full acquired data buffer vs. a decimated subset explains a lot. I had not realized that some other scopes decimate the data. (Besides my DS1054Z, which takes this to the extreme of using the screen data only...)

So, in situations where the high resolution is not needed and I want to gather statistics quickly, I can simply reduce the number of data points to speed things up? I did notice in the video that the measurement rate went up when Rudi demonstrated the use of gated measurements, i.e. only a shorter sequence of the data was used for the measurements.

Also, thank you for pointing out that multiple measurements per capture are derived for the time-related measurements. That makes sense; one does want individual measurements per period when collecting statistics on jitter etc.

You should be able to count not 6, also not 12 but no less than 52 measurements at once.

Now you are just showing off! 

[AC360]

You have already named it – this is a MSO with main focus on analytic capabilities. Of course there is a difference between analytic scopes, which require deep measurements, and standard scopes, which prioritize speed and try to measure (estimate would be more appropriate) on heavily decimated data. To understand this, here’s a simple example:

SDS2354X_Plus_Meas_Pulse_10Mpts

This screenshot shows just a pulse train. Pulse width is 10 µs, transition times are ~1 ns and repetition rate is 4 kHz. At a time base of 500 µs/div and 2 GSa/s, this results in a record length of 10 Mpts. The key parameters of the pulse train shall be measured: period, amplitude, rise and fall times and pulse width. Everything is measured reasonably correctly. Of course, at 2 GSa/s this scope isn’t fast enough to measure a 1 ns transition time precisely, yet it’s well within reasonable expectations. For more accurate measurements on such short time intervals, we’d have to zoom in up to the point, where sin(x)/x reconstruction calculates additional data points. We cannot use a much faster timebase right from the start, because at faster time bases we lose the measurements, one by one: At < 25 µs/div we could not measure the period anymore, at <1 µs/div we lose the pulse width measurement and can only measure a single transition, I.e. either the rising or the falling edge.

Now try the same exercise with the fast Keysight, which uses 64 kpts decimated data – good luck!

It is worth to take a closer look at this screenshot. We see advanced measurements in M1 mode, where statistics and Histicons (small histogram icons) are enabled. If we look at the “count” in the measurement statistics, it becomes obvious that only the vertical (amplitude) related measurements appear to be slow. For the time measurements, the scope analyzes the entire record, hence a single acquisition yields some 20 measurements of period, rise and fall times and pulse width. It just so happened that not a single time related measurement was shown in this comparison.

Apart from that the SDS2000X Plus firmware has seen a few optimizations since this comparison. My measurements yielded at least twice the speed as the numbers found in this review. If I make the conditions a bit more equal, then the vertical measurement rate reaches >16 measurements per second if I reduce the record length to 100 kpts, thus getting into the ballpark of the measurements on decimated data of the competing instruments.

Furthermore we have these M1 and M2 measurement modes (limited to max. 12 measurements) only in the “advanced measurements” mode. But there is also a “simple” mode, which allows a lot more than just 12 measurements in parallel:

SDS2354X_Plus_Meas_Simple

You should be able to count not 6, also not 12 but no less than 52 measurements at once. The glitch, that the measurement window closes as soon as we select an item (so that we cannot read the help text and have to re-open the window if we need to add more measurements) is long fixed either.

Hi Performa01,

How to enable all the measurements at once in “simple” mode with SDS2354X Plus?

Need to upgrae the firmware to support this?

[Performa01]

How to enable all the measurements at once in “simple” mode with SDS2354X Plus?

Simple. In the "MEASURE" menu, select "Type". Then choose from the three tabs (Vertical, Horizontal, Miscellaneous) all the measurements you like - for this demo (some call it "show off" ) I've simply selected all of them.

Done!

EDIT: You can change the measurement source after the fact. This enables you to compile a nice set of measurements and then apply it to any possible input source you like.

[RBBVNL9]

Dear all,

In this series of videos, I now posted an episode on the Power Analysis functions of the Siglent SDS2000X plus. While also looking at power measurements using the R&S RTB2000 and Keysight DSOX1000, the primary comparison is with a dedicated power meter, the GW INSTEK GPM-8310.

Enjoy!

[2N3055]

@RBBVNL9
Hello,

just a note, you cannot do this kind of math (you used on efficiency calculations) on either Keysight or RTB2000. Keysight has only simple math and RTB2000 has different math architecture without formula editor.

[RBBVNL9]

Hi 2N3055,

Code: [Select]

just a note, you cannot do this kind of math (you used on efficiency calculations) on either Keysight or RTB2000. Keysight has only simple math and RTB2000 has different math architecture without formula editor.
It’s true that the Siglent has a formulae editor, which the RTB does not. Yet, the RTB has five math channels (compared to 2 in the Siglent). So, what you typically do, is to break down tasks. The formula you refer to can be broken down into several parts (see the attached screenshot), which have the end result.

I understand some will really prefer scopes with a math formula editor (I myself particularly like the one in the PicoScope). And there certainly are formulas that cannot be broken down into parts the RTB can handle. On the other hand, having five math channels, like the RTB, can be valuable in other situations. Moreover, if you use it to break down a formula, you can also monitor each element separately in a measurement function, which can be useful for tracking issues.

The Keysight DSOX1000 is much more basic when it comes to math: one channel only, and no formulae editor.

[2N3055]

Hi 2N3055,

Code: [Select]

just a note, you cannot do this kind of math (you used on efficiency calculations) on either Keysight or RTB2000. Keysight has only simple math and RTB2000 has different math architecture without formula editor.
It’s true that the Siglent has a formulae editor, which the RTB does not. Yet, the RTB has five math channels (compared to 2 in the Siglent). So, what you typically do, is to break down tasks. The formula you refer to can be broken down into several parts (see the attached screenshot), which have the end result.

I understand some will really prefer scopes with a math formula editor (I myself particularly like the one in the PicoScope). And there certainly are formulas that cannot be broken down into parts the RTB can handle. On the other hand, having five math channels, like the RTB, can be valuable in other situations. Moreover, if you use it to break down a formula, you can also monitor each element separately in a measurement function, which can be useful for tracking issues.

The Keysight DSOX1000 is much more basic when it comes to math: one channel only, and no formulae editor.

I apologize for being too terse.

RTB has 5 channel/functions. For simple stuff you can get 5 Math traces. For complicated stuff only one or you cannot accomplish it.
On SDS2000X+you have 2 Math channels but you can perform complex math on both.
So sometimes RTB can be better if you need more simple math.
True is that today, price range comparison of RTB2000 would be with SDS2000XHD, that has 4 complex channels....

Yes Picoscope is another level..  I have 3... 

[RBBVNL9]

I have now updated the series of comparison videos with an episode on math channels. Enjoy!

[Performa01]

I have now updated the series of comparison videos with an episode on math channels. Enjoy!

Thank you very much for this nice presentation!

As always, a few remarks 😉

Of course does the SDS2000X Plus support track plots. I only have examples with the SDS824X HD handy, but you should be able to do this with the SDS2000X Plus just as well. It just cannot be found in the math menu, but at Measure - Tools – Track. This has the advantage that we can track all time-related functions, e.g. also rise time changes over time.

https://www.eevblog.com/forum/testgear/sds800x-hd-review-demonstration-thread/msg5293765/#msg5293765

https://www.eevblog.com/forum/testgear/siglent-sds800x-hd-12-bit-dsos-coming/msg5436737/#msg5436737

https://www.eevblog.com/forum/testgear/siglent-sds800x-hd-12-bit-dsos-coming/msg5439530/#msg5439530

https://www.eevblog.com/forum/testgear/yet-another-which-oscilloscope-thread-(though-slightly-different-i-promise)/msg5346956/#msg5346956


As far as the never-ending discussion about the ERES/Average modes in the SDS2000X Plus, it’s been often stated (and I want to repeat it here) that there are some advantages in having them as math functions, mainly that you can turn it on for individual channels, combine them together and with other math functions and finally you don’t destroy the original data, which you could still compare and analyze in parallel.

Acquisition modes on the other hand are never channel-specific; whenever you get something like that in the channel menu, then it certainly is some math function in disguise, destroying the original data for no good reason.

The SDS2000X Plus is a very special device, as it is the only one of the new touch-screen generation, that has these limitations. All newer models have either four math channels and/or ERES/Average as both acquisition mode and math function. Higher class devices like SDS2000X HD and above (except SDS5000X) offer everything at once.


Finally, the “divide by zero” debate. I was quite baffled to learn that you think there is something wrong with the solutions chosen by PicoScope and Siglent. In fact, there is not a single drawback to this. The R&S approach is almost the same, as PicoScope and Siglent have to do something similar; otherwise the application processor in the scope would throw an exception. The difference might just be the concrete number a vendor uses to substitute the zero value. This should take in consideration the possible resolution enhancement by certain acquisition modes and math functions, hence the smallest number that can be represented by the internal number format used for math would be a good start.

We should never overdrive an input channel, because clipping can have a number of adverse effects. But there is no such limitation to the math channels. If a math result ranges from zero to a very high number (representing infinity) then it’s up to the user to chose a scaling factor that shows the region of interest best. I consider this even better than the scope trying to be smarter than its user and limit their choices.

And you cannot argue with measurements; any function result that involves a division by zero just cannot yield a useful result. The correct approach in such cases is once again to define the zone of interest and do time-gated measurements there, carefully excluding the illegal part that contains one or more divisions by zero.