Products > Test Equipment
Siglent SDS2000X Plus Bandwidth & Aliasing Application Note
Performa01:
For those who got unnerved by the fact that, apart from a very few exceptions, the sample rate in modern DSOs barely exceeds the Nyquist requirements for the top model in a product line – and even worse, some models even offer bandwidths so high that the Nyquist criterion can only be met by reducing the number of active channels – I did some research and wrote an application note specifically for the Siglent SDS2354X Plus in order to bring some proven facts into the debate.
The SDS2000X Plus is unique within the Siglent product line of DSOs, because it offers a 500 MHz bandwidth option that can only be used in half channel mode, where the sample rate is 2 GSa/s.
At this point it should be mentioned that Siglent is not alone with this.
Fans of the Rigol MSO5000 often praise the excessive sample rate of 8 GSa/s, which ultimately means 4x2 GSa/s – and this is still more than enough for only 350 MHz bandwidth. I have read statements suggesting that other DSOs have “barely enough” sample rate for their input bandwidth.
What about the MSO8000 from the very same brand? Its sample rate would be “barely enough” for 1 GHz with more than 2 channels active, and it can be limited to just 2 channels to still get “barely enough” sample rate for 2 GHz. “Barely enough” in this context means a sample rate to bandwidth ratio of 2.5.
This is not limited to the so-called B-brands: the Keysight 4154A has a very similar restriction, although with slightly more margin in half-channel mode:
All channels: 1 GHz bandwidth @ 2.5 GSa/s; Ratio = 2.5 (like Keysight 3104T)
Half channels: 1.5 GHz bandwidth, 5 GSa/s; Ratio = 3.33
By contrast, the summary for the Siglent SDS2354X Plus with 500 MHz option looks like this:
All channels: 350 MHz bandwidth @ 1 GSa/s; Ratio = 2.85
Half channels: 500 MHz bandwidth, 2 GSa/s; Ratio = 4.0
Yes, most entry level DSOs coming from A-brands have a lot more margin, but that’s obviously not because manufacturers all of a sudden worry about aliasing in entry level products of all things, but they don’t want their “cheap” (everything is relative!) 1000 and 2000 product lines to eat into the sales of the much more expensive higher class products with higher bandwidth.
Please find the application note attached: SDS2000X Plus Bandwidth.pdf
Here is the verdict of the article:
--- Quote ---As could be demonstrated, the SDS2304X Plus with 500 MHz option works without any artifacts in half channel mode at 570 MHz bandwidth even with fast signals like the <200 ps rise time square wave.
Special caution is required when using more than two channels, because then the sample rate drops to just 1 GSa/s and even the input bandwidth limit is rather ineffective for various reasons. Consequently, for happy tinkering and probing around without further considerations, 4-channel operation should be limited to signals slower than 1.6 ns rise time if pulse fidelity is of any importance.
On the other hand, pulse integrity checks which require a reasonable accurate waveform display usually need not be performed on all channels all the time, so the half channel mode can be used for this exclusively. After these tests have been completed with a positive result, even the digital logic channels are good enough for many tasks, all the more so the analog channels in full channel mode.
For the timing analysis, the true waveform is not important – we often use the digital channels for this anyway – and in contrast to the logic channels we can still see a coarse approximation of the signal shape in dots display mode even when visual signs of aliasing start creeping in – and we still get fairly accurate amplitude and time measurements.
As has been demonstrated, dots display mode eliminates the reconstruction error and works well even close to the Nyquist frequency, so this can save your day in certain borderline situations, where aliasing artifacts are limited to overshoot and ringing on the pulse flats, hence not affecting the trigger signal path.
At the end of the day, the SDS2504X Plus can handle all kinds of TTL and CMOS logic up to GTL. With some precautions even LVDS is not completely out of the question, even though a higher bandwidth midrange scope with active probe support like the SDS6000 would be much better suited and the obvious choice for this.
--- End quote ---
EDIT 20220105: Document revised, corrected some errors and added numbers to the pictures.
EDIT2 20220107: Document revised again:
* Misleading "Safe Signals" and "Conclusion" sections corrected - in half channel mode, the SDS2000X Plus can handle at least the 200 ps rise-time signals without artifacts.
* Quoted Conclusions in this posting adapted according to the revised document.
* Artifacts can be Gibbs phenomenon and/or aliasing, both caused by high frequency portions of the input signal that vastly exceed the Nyquist bandwidth of the DSO.
* Lower bandwidth models most likely do not have a digital filter.
* Version number added to the title page (V1.02).
EDIT2 20230413: Document revised and supplemented (V1.03):
* Some typos corrected.
* Quoted Conclusions in this posting adapted according to the revised document.
* "SDS2354X Plus with 500 MHz option" is now called "SDS2504X Plus".
* The effect of different bandwidths demonstrated by the SDS6204 using digital FIR filters (chapter "How much bandwidth do we need?").
tautech:
Nice work however some edits are needed in the PDF where on numerous occasions SDS2304X Plus is mentioned where I believe SDS2504X Plus is the intended script.
Thread now added into the POI list in the SDS2000X Plus thread OP:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/
bdunham7:
Very nice article. Two comments:
1) The issue of the BW in 350MHz mode being much higher than specified is something they should fix if possible. I understand the idea of maximal leveraging of the sample rate, but having the -3dB point be higher than the Nyquist limit seems to be particularly unhelpful.
2) With regard to the usefulness of the scope at varying risetimes, the width of the pulse is just as important a the risetime in how well it triggers, measures times, etc. For typical data signals, I suppose the two are closely related--you won't have signals with long periods and short rise times--but for some other applications you may have just that. The ability of the scope to accurately and reliably trigger, and thus measure fast edges and event timing are actually remarkably good. CH1 is a 33ps rise time 10MHz signal, CH3 is an 8.4ns rise time 10MHz signal.
nctnico:
--- Quote from: Performa01 on January 05, 2022, 12:24:02 pm ---For those who got unnerved by the fact that, apart from a very few exceptions, the sample rate in modern DSOs barely exceeds the Nyquist requirements for the top model in a product line – and even worse, some models even offer bandwidths so high that the Nyquist criterion can only be met by reducing the number of active channels – I did some research and wrote an application note specifically for the Siglent SDS2354X Plus in order to bring some proven facts into the debate.
--- End quote ---
On page 16 you seem to show the difference between sin x/x reconstruction (top) and dot mode (bottom) to show the effect of aliasing. But IMHO this isn't correct. If you have signal distortion due to aliasing, then the dot mode would also show the same signal. The effect you are showing in the top picture looks like the Gibbs effect to me.
Edit: it also seems to me that the sin x/x is broken from the picture on page 25 and/or the triggering is wrong. A 200MHz square wave should show as a sine-ish wave. At 1Gs/s the limit for sin x/x to stop working is over 400MHz (1Gs/s / 2.5) and 200MHz is far below that limit. The effect shown has nothing to do with aliasing or rise time!
tggzzz:
It would be helpful if each figure was numbered, and the text referred to it like "Figure 123 shows..."
Navigation
[0] Message Index
[#] Next page
Go to full version