FFT - Rohde & Schwarz RTB2002 vs. Siglent SDS2104X Plus

[Hardware]

I had the opportunity to see a Rohde & Schwarz RTB2002 70MHz and a Siglent SDS2104X Plus. I am looking for a replacement for my 30 years old Hameg HM412-5.
For this I took a look at the sine and square wave signal of my new SGS2042X.
So I am still inexperienced with the DSOs.
That the measurement results were not always so identical is probably due to the tolerances.
Now I have also looked at the FFT function and compared it. Of course I know that the built in spectrum analyzer is only a little bit comparable to a real spectrum analyzer.
But the difference between the R&S and the Siglent was so big that it is not clear to me what the cause is.
Regarding the shown levels in dBm and the spectrum I see serious differences.
Yes, I have one oscilloscope on one channel and the other oscilloscope on the other channel. But I also changed the channels but it did not lead to a fundamental change of the displayed problem.

Did I do something wrong or do I expect too much?

The first two pictures show the FFT of a 1kHz sine wave.
The second two pictures show the FFT of a 100kHz square wave signal.

[2N3055]

Did you put external  50 Ohm terminator on R&S for measurements?
It doesn't have built in 50 OHm termination.
Also you should try to make both measure with similar RBW and same FFT Window..

[Hardware]

Yes, I had a 50Ohm resistor on the R&S. Regarding the FFT window the settings are identical or nearly identical. You should be able to see that in the pictures.

[Fgrir]

For 1KHz you are comparing dBm and dBV, you should expect to see different results.

For both frequencies on the Siglent you have C2 off the top of the scale, are you sure you aren't clipping the waveform?

[Andreas]

Did I do something wrong or do I expect too much?

I fear you have a signal clipping on the Siglent (50mV/Div versus 200mV/Div)

with best regards

Andreas

[Andreas]

And also different amplitude settings in the FFT (dbV versus dbm)
...

[Hardware]

Yeah, you're right. With the 1kHz sine signal I set dBV on the siglent by mistake. The following picture shows it correctly. But it does not change the fact that I expected a peak in the FFT window. In my opinion the R&S shows it correctly.
Also the 100kHz rectangle I just took directly from the oscilloscope for better readability.
But I don't think we have a clipping problem here. Yes, the vertical settings of the R&S are a bit different from the siglent. But that is because I can only place the signal separately above the FFT window and not overlapping as with the siglent.

[tautech]

Hardware, your Span settings are vastly different. 
Also good to have a SA in the mix to compare results of each against.

[Hardware]

The display may be a bit confused now. In the first case I compare the FFT of the sine 1kHz of the siglent with the FFT of the sine 1kHz of the R&S.
In the second case I compare the FFT of the rectangle 100kHz of the siglent with the FFT of the rectangle 100kHz of the R&S. The SPAN range between the 1kHz (0-6kHz) and the 100kHz signal (0-600kHz) is different.

What surprises me is that the siglent shows more peaks than the R&S.
In my opinion the siglent shows too many peaks. Unfortunately I cannot check it with a real spectrum analyzer because I don't have one.
But the SGS2042X has so many that I am sure there is someone who has a spectrum analyzer and can show a photo.

I am interested in which oscilloscope is better or more correct.

[tautech]

Later when time is available I’ll post some FFT’s from a Plus and SVA but now I have some jobs to do outside while it is fine and sunny

[Andreas]

What surprises me is that the siglent shows more peaks than the R&S.

The R&S simply seems to clip the curves near the noise floor of -60 dbm .
The Siglent shows much lower amplitudes too. (how many points does the FFT have on each device?)

with best regards

Andreas

[maginnovision]

You should probably turn off 10bits mode on the siglent. Also is avg=4 averaging the waveform or avg of 4 ffts. The waveform should ideally fill the display as well. I'd also do standard sample over high res on R&S to be able to get more like for like comparison.

[Hardware]

Unfortunately I can't answer the question about the points anymore, because the R&S oscilloscope is on its way back. In the specifications of R&S I can't find any statement about the number of points per second at the moment.

[Andreas]

Hello,

to get a RBW of 33 Hz I have to use  a insane low number of FFT points.
At 8 kPts I also do not see any further harmonics.

with 1 MPts the bandwidth or in my case bin width decreases
so also the noise floor goes down making the harmonics visible.

Increasing the ADC resolution to 16 Bits further decreases the noise floor.
And there are other tricks like averaging several FFTs possible.

with best regards

Andreas

[2N3055]

The display may be a bit confused now. In the first case I compare the FFT of the sine 1kHz of the siglent with the FFT of the sine 1kHz of the R&S.
In the second case I compare the FFT of the rectangle 100kHz of the siglent with the FFT of the rectangle 100kHz of the R&S. The SPAN range between the 1kHz (0-6kHz) and the 100kHz signal (0-600kHz) is different.

What surprises me is that the siglent shows more peaks than the R&S.
In my opinion the siglent shows too many peaks. Unfortunately I cannot check it with a real spectrum analyzer because I don't have one.
But the SGS2042X has so many that I am sure there is someone who has a spectrum analyzer and can show a photo.

I am interested in which oscilloscope is better or more correct.

You didn't compare FFT at same settings. like I said before.
Siglent display  shows much lower noise floor and has much better RBW.  We also don't know if same windowing function was used.
Peaks visible on Siglent are buried  in noise on R&S. 

Looking at this Siglent looks better by a mile. Actually I think that with proper settings R&S should do better... FFT is quite complicated sometimes and need some time to get acquainted with how it's implemented on particular scope.

I just saw Andreas already said most of this..

[jjoonathan]

Is that a gate cutting down the RTB to a fraction of the acquired length?

[maginnovision]

Here is what I got with a 100kHz square wave and averaging(like the siglent seemed to be doing). Flat top window, dBm. The waveform fills the capture screen as well.

I'm attaching another image with the window cut down from 1.13ms to just 500us and switched off high res mode.

[2N3055]

Here is what I got with a 100kHz square wave and averaging(like the siglent seemed to be doing). Flat top window, dBm. The waveform fills the capture screen as well.

I'm attaching another image with the window cut down from 1.13ms to just 500us and switched off high res mode.

Yes, thank you, that's better..
FFT is tricky, settings are critical.
Here it is how it looks like, on 8bit scope with good FFT implementation. Averaging used.

If you sample with very low sample rate, and feed FFT with very wideband signal (like square wave with many harmonics ), those frequencies that are too high for sample rate will get folded down (basically downsampled) and falsely shown as low frequency components. On a perfect square wave there should not be any even harmonics.
Any even harmonics are either distortion (scope, siggen or interconnect) or aliasing from too low sample rate.

[tautech]

Here is what I got with a 100kHz square wave and averaging(like the siglent seemed to be doing). Flat top window, dBm.

Here zero averaging, instead Max Hold which does indeed do a type of average in each sweep.
Signal source SDG6022X 0dBm Tee'd to supply SDS2104X Plus and SVA1032X so unmatched load for the AWG but serves well enough for this demonstration.
It need be noted the SVA in FFT mode cannot see the even harmonics only the fundamental and its odd harmonics.
This is something about square wave harmonics I have not understood until recently where it was discussed in this post: https://www.eevblog.com/forum/beginners/what-an-oscilloscope-recommended-for-a-woman-passionate-about-electronics/msg3216564/#msg3216564

Why the scope can see the even harmonics IDK. 

[2N3055]

Here is what I got with a 100kHz square wave and averaging(like the siglent seemed to be doing). Flat top window, dBm.

Here zero averaging, instead Max Hold which does indeed do a type of average in each sweep.
Signal source SDG6022X 0dBm Tee'd to supply SDS2104X Plus and SVA1032X so unmatched load for the AWG but serves well enough for this demonstration.
It need be noted the SVA in FFT mode cannot see the even harmonics only the fundamental and its odd harmonics.
This is something about square wave harmonics I have not understood until recently where it was discussed in this post: https://www.eevblog.com/forum/beginners/what-an-oscilloscope-recommended-for-a-woman-passionate-about-electronics/msg3216564/#msg3216564

Why the scope can see the even harmonics IDK. 

Try enabling 20MHz limiting and try again with same settings.. We are looking at 1 MHz bandwidth anyways. Also enable averaging, not peak hold.
Try with different V/div settings. With spectrum, we want least distortion with best noise floor. Try capturing at 200mv/div. That will decrease dynamic range but lower distortion in scope frontend. What you see now?.

My point is that FFT must be understood first, than you must get used to implementation on your scope. After that you need play with it for a while to characterize it's strengths, weaknesses, where the spurs are etc. It is same as spectrum analyser. It is practically another whole instrument inside scope that you have to learn in detail to be able to use well..


EDIT: posted too soon,sorry.. in bold what was added later..

[tautech]

Here is what I got with a 100kHz square wave and averaging(like the siglent seemed to be doing). Flat top window, dBm.

Here zero averaging, instead Max Hold which does indeed do a type of average in each sweep.
Signal source SDG6022X 0dBm Tee'd to supply SDS2104X Plus and SVA1032X so unmatched load for the AWG but serves well enough for this demonstration.
It need be noted the SVA in FFT mode cannot see the even harmonics only the fundamental and its odd harmonics.
This is something about square wave harmonics I have not understood until recently where it was discussed in this post: https://www.eevblog.com/forum/beginners/what-an-oscilloscope-recommended-for-a-woman-passionate-about-electronics/msg3216564/#msg3216564

Why the scope can see the even harmonics IDK. 

Try enabling 20MHz limiting and try again with same settings.. We are looking at 1 MHz bandwidth anyways. Also enable averaging, not peak hold.

Can do. With SDS2000X Plus we can apply averaging to both the displayed waveform and an FFT average.
Be back in 5-10 as they all need booting up again. 

[2N3055]

I played with the settings a bit more..
Captured all the same, but with 200 mV/div.
Odd harmonics much lower. Worse noise floor, obviously...

[tautech]

SDG6022X 0dBm into SDS2104X Plus only now so matched feedline and square wave now clean as it should be.
FFT averages = 10.
Then I looked at the time....no not the DSO time which is still Shenzhen factory time = minus 4hrs NZ time so nearly 11pm. Nighty night all, got things to do tomorrow.

[Performa01]

I have tested the FFT on SDS1000X-E, SDS2000X Plus and SDS5000X extensively and the results have always been correct and accurate. Siglent FFT provides class leading 2 Mpts FFT length (number of FFT points), except for SDS1000X-E, which is "only" 1 Mpts. This enables low resolution bandwidths and low noise, but keep in mind it's still an 8 bit DSO, hence don't trust results below 48 dBFS blindingly. In most cases, the usable dynamic range with decent accuracy can be up to 70 dB though.

Some hints for proper setup of the FFT on Siglent DSOs:

FFT-Bandwidth and RBW
This is quite different to a real SA. There is no menu for the resolution bandwidth and also no direct setting for the FFT-bandwidth, even though we have a menu item for the horizontal scale in Hz/div, which ultimately specifies the visible span. But this is just for zooming into a longer FFT trace; for best speed and lowest RBW we need to make sure that no high zoom factor is required to get the display we want. The following rules apply:

• The analysis bandwidth (FFT-BW) is always half the FFT sample rate (FFT-SR).
• The frequency step (Δf or df) is the sample rate divided by the number of FFT points.
• The resolution bandwidth (RBW) is the frequency step multiplied by a factor specific for the
window function in use.
• The maximum number of FFT points depends on the record length, which in turn increases with slower timebase settings, but is ultimately limited by the maximum memory set in the Acquire menu and of course also the specified maximum possible FFT length. Apart from that, the max. number of FFT points can be further limited by the specific setting in the FFT menu.

RBW = df * k, where k is the 3 dB bandwidth factor in bins, depending on the window function: Rectangle 0.99, Blackman 1.74, Hanning 1.62, Hamming 1.64, Flattop 3.73.

Blackman and especially Flattop are the most universal and useful window functions in practice, whereas Rectangle is rather specialized and should be avoided unless you absolutely know that you actually need it (e.g. for short transients).

Thus: df = RBW / 4 (rounded) in case of the flattop window.

To get the proper settings for any given FFT-BW and RBW pair, proceed as follows:

Determine the FFT samplerate: SR = FFT-BW * 2 [Sa/s];
Determine the number of FFT points: FFT-Pts >= SR / df [-];
Determine the timebase: TB >= FFT-Pts / SR / 10 [s/div];


Setting up an FFT Measurement
Even from the best FFT implementation, we can only expect good results as long as the scope has been set up properly for that specific task. How many so called “reviews” have we seen where FFT has been engaged and some scope settings randomly altered just to get some halfway plausible but actually rather meaningless FFT graph, which was then either praised or criticized?

Of course we can get away with some quick & dirty setup if our requirements are low, but we should never ignore the most important parameters like FFT bandwidth. We won't see anything meaningful, i.e. just some aliasing artifacts, if, for instance, we try to get the spectrum of a 33 MHz signal with just 25 MHz FFT bandwidth. Furthermore, for optimal speed, frequency resolution and dynamic range, we need to put a little more effort into a proper setup, which has quite different requirements compared to the usual Y-t view. Below there is a complete checklist how to properly set up the DSO for analysis in the frequency domain (most of these topics should be obvious, but still listed for completeness):

• Set acquisition mode to normal. Use average only for a good reason and stay away from ERES. Avoid Peak Detect under all circumstances and without any exception!
• Use edge trigger in auto mode to make sure signal acquisition doesn’t stop even when the signal amplitude drops below the trigger sensitivity. FFT doesn’t require a stable trigger by the way.
• Determine the lower bandwidth limit for the FFT analysis. If it is >10 Hz, use AC-coupling for the input channel to ensure maximum dynamic range even with large DC offsets and/or high input sensitivities. If DC-coupling has to be used, use the vertical position control to compensate for the DC offset, so you can get maximum sensitivity, hence highest dynamic range.
• Determine the upper bandwidth limit for the FFT analysis. In order to avoid aliasing artifacts, this should not only cover the desired analysis bandwidth, but include the highest expected input frequency. In general, it’s best to start with a higher upper bandwidth limit and reduce it only after it has been confirmed that there is no significant signal content above the desired final limit.
• Choose the frequency step size according to the explanations given earlier in this article, which would be about one quarter of the required resolution bandwidth when using the Flattop window.
• Find an appropriate set of horizontal timebase setting and the number of FFT points; refer to the explanations given earlier in this article. You should watch the displayed FFT parameters and double check that the chosen timebase together with the selected FFT length (number of points) matches your expectations. Be aware that the desired resolution bandwidth might not be achievable due to the limited choice of sample rates and FFT lengths and/or the maximum specified FFT length of your specific instrument.
• Engage FFT mode, select the correct source channel and start with Split Screen mode.
• Set the vertical gain so that the peak amplitude of the input signal is between ±2 to ±4 divisions.
• Set the FFT center frequency to the arithmetic mean between lower and upper bandwidth limit.
• Set the FFT frequency scale so that the desired analysis bandwidth is displayed on the screen.
• Set the desired level units and make sure the external load impedance matches reality whenever working with power levels, i.e. dBm.
• Set the reference level and vertical scale so that the FFT amplitude range of interest makes best use of the available screen space.
• Setup automatic peak-peak (and maybe RMS) measurement for the input channel, as well as Max for the math channel. During frequency domain analysis, especially in Exclusive mode, keep an eye on the Vpp measurement for the input channel to make sure no overload occurs.
• Select an appropriate window function; refer to the hints earlier in this document.

Hint: stay in Split Screen mode until the amplitude setup is finished and the levels are reasonably stable,
then switch to Exclusive mode. By keeping an eye on the peak to peak measurement of the input signal,
you can still detect an overload condition instantly; the scope indicates that by displaying > instead of = in
front of the measurement value, e.g. Pk-Pk[4]>796.00mV instead of Pk-Pk[4]=640.00mV.

[mawyatt]

Performa01,

Thanks for the nice FFT tutorial.

I have some background with FFTs (Discrete Time Continuous Amplitude Chirp-Z Algorithm from ~1980 and more recently a device called an RF Spectral Imager, basically an analog FFT on a single chip using RF pixel techniques similar to a camera sensor (https://patents.justia.com/patent/20190101577) and curious about how Siglent implemented the ERES function (maybe Auto-Correlation)? Here's a simple test I did that shows some frequency domain effects of the ERES function using the FFT.

https://www.eevblog.com/forum/testgear/sds2102x-plus-enhanced-bit-function/

Best,

Edit: I have an old PP Presentation on the RF Spectral Imager if interested, don't want to dilute this thread tho.