Choosing between entry-level 12-bit DSOs

[2N3055]

I can zoom in on the top corner of a square wave, too...



And then add an FFT of that...

No you cannot zoom. You can use Math channel. It is something Performa01 had demonstrated, last time when this "can it zoom vertically ?" was discussed.
Math is valid technique and accomplishes the task of "visual magnification" but it is not a part of zoom mode of your scope.

But Performa01 demonstrated this in the view that math can be used to magnify signal vertically without zoom  mode, because on Siglent that allows you to show such "magnified" signal in main window with other traces, for comparison.

So this demonstration of yours actually shows two negative points of your scope  (and two positive points of Siglent concept):
- It cannot zoom vertically.
- It cannot show math channels in same window, just in separate window, on a very small and crowded screen.

[tggzzz]

No you cannot zoom. You can use Math channel. It is something Performa01 had demonstrated, last time when this "can it zoom vertically ?" was discussed.
Math is valid technique and accomplishes the task of "visual magnification" but it is not a part of zoom mode of your scope.

But Performa01 demonstrated this in the view that math can be used to magnify signal vertically without zoom  mode, because on Siglent that allows you to show such "magnified" signal in main window with other traces, for comparison.

So this demonstration of yours actually shows two negative points of your scope  (and two positive points of Siglent concept):
- It cannot zoom vertically.
- It cannot show math channels in same window, just in separate window, on a very small and crowded screen.

It also shows another point: how complex such scopes are, and how difficult it is for an experienced engineer/operator to understand exactly what they are and are not showing.

[ebastler]

I can zoom in on the top corner of a square wave, too...
And then add an FFT of that...

That "vertical zoom" via a math function can be used in a pinch, but it is much less convenient than Siglent's interactive zoom. Siglent lets you use the regular vertical (and horizontal) controls to adjust the magnification and position of the zoomed-in section. And it indicates the zoomed-in area as a nice little "you are here" rectangle in the upper, non-zoomed trace, so you don't lose track of where you are looking.

Combining FFT with Zoom mode is possible in the Siglent UI as well. Let me just say that I have found very few practical use cases where I really want to see the trace in detail and look at the FFT at the same time. To get decent FFT resolution, I typically want to capture a long signal segment, hence can't resolve any detail along the time axis anyway. I merely use the trace display to make sure I have chosen a good vertical amplification -- and then often disable it to see the FFT exclusively. (Not sure whether the Rigol UI allows for that?)

Edit: Adding a screenshot which shows how a combined Zoom + FFT looks on the SDS800X HD, to give a better idea. Note the "you are here" rectangle in the upper trace window.

As a side note: I find it much easier to get meaningful, "clean" axis tick values on the Siglent. The dual Flexknobs in Rigol's UI, with their dynamic assignment of functions, are a great idea IMO. But Rigol implemented them in an awkward way -- they typically change values in very odd increments, often on a scale which does not even include the even values. So you either use the on-screen keyboard a lot to set axis scales and offsets, or you end up with very odd values as shown in Fungus' screenshots above. Easily changed in firmware; I hope Rigol will do that at some point to make the Flexknobs more useful.

[shapirus]

I can zoom in on the top corner of a square wave, too...

Try to zoom in on a corner of a 2.5 Vpp unipolar square wave to 10 (or at least 20) mV/div without using math.

But, good idea on using math for this. At least it can serve as a workaround.

[shapirus]

On the other hand, every single screenshot of a trace on the Siglent that I have seen has some weirdly pixelated (or interlaced if you wish) trace rendering, unlike Rigol, where the trace rendering/display engine does its job near perfectly in that regard. I can see how this can be annoying until you get used to it, and some may never do. But that's about the only issue with it that I have noticed so far without owning it, and it still looks much more capable and oriented towards a competent user than the Rigol.

[ebastler]

On the other hand, every single screenshot of a trace on the Siglent that I have seen has some weirdly pixelated (or interlaced if you wish) trace rendering, unlike Rigol, where the trace rendering/display engine does its job near perfectly in that regard.

Yes, Siglent had to cut corners with the vertical trace resolution, due to FPGA capacity constraints: Vertical pixels are actually grouped in pairs. This only applies to CH1..CH4 traces, not Math functions. It has been discussed at some length in the SDS800X HD thread and Performa01's demo/review thread.

I initially found this a very disappointing design tradeoff. In practice I rarely notice it on the small screen -- but sometimes I do, especially on signals with a flat slope or weak modulation. The SDS1000X HD and 2000X plus apparently have the same constraint, and I would probably be disappointed if I had paid extra for the large screen only to see this resolution compromise better. Larger Siglent SDS models support the full vertical resolution.

[Mechatrommer]

dho900 can go down to 200uV/div...

[shapirus]

dho900 can go down to 200uV/div...

Yeah, but the usability of that is questionable because of noise. Everything below 1 mV/div at 1:1 probe ratio or 10 mV/div at 10:1 probe ratio is a software stretch, but that's probably fine with the 12-bit resolution. The noise level, however, limits usability -- yes you can use averaging to cancel the noise, and it can be a good solution in many cases, but not all. There's also a parasitic DC offset, at least in my case, which becomes an issue at those extreme scale levels (but not too big of an issue, since you can move the waveform up and down manually). But yes I did already use the 200 and 500 uV/div scales with success to obverve and even measure very weak signals (e.g. output of a toroidal inductor used as a current transformer measuring low-mA currents without any preamplification).

[2N3055]

dho900 can go down to 200uV/div...

No, it does not.
Best it can do is 1mV/div, and 500µV/div and 200µV/div is software magnification.

[ebastler]

dho900 can go down to 200uV/div...

No, it does not.
Best it can do is 1mV/div, and 500µV/div and 200µV/div is software magnification.

As shapirus said, I don't think that is the major limitation. If the preamp and ADC are set to 8 mV full scale (4096 counts), a software magnification by a factor of 5 will still give you more than one ADC count per vertical pixel.

But I wonder how practical 200 µV/div are from a noise perspective. Peak-to-peak noise would easily span a full division, I assume?

[Mechatrommer]

well magnified or not... imo good enough "bell and whistle" for low noise debugging or measurement for blurry eye (when tired) like me... use averaging to lower the scope's noise ymmv (attached is 50 ohm terminated input)

[ebastler]

Ok, not quite a full division of peak-to-peak noise, but pretty close. And I guess you chose favorable conditions with the relatively fast timebase, which excludes most of the 1/f noise. Could you show this at a slower time base too, say 10 ms/div?

[Mechatrommer]

Ok, not quite a full division of peak-to-peak noise, but pretty close. And I guess you chose favorable conditions with the relatively fast timebase, which excludes most of the 1/f noise. Could you show this at a slower time base too, say 10 ms/div?

see attached 01,02,03 (peak, ultra, average 4) input 50 ohm terminated. i also captured my UTG962 AWG output 1mVpp but terminated with 12.5ohm (4x inline 50 ohm) because its not so low noise enough, 50Hz and 100KHz not so stable trigger due to noises (normal acq and average) ymmv. (ps: last image... how to estimate dut noise from combined scope + dut noise ymmv.)



ps: i didnt notice 20MHz BW limit activated during the capture, it wasnt me! well you got the idea...

[Fungus]

dho900 can go down to 200uV/div...

No, it does not.
Best it can do is 1mV/div, and 500µV/div and 200µV/div is software magnification.

Yes, but those scales make it more visible to the eye.

PS: Hacked DHO800 can do it, too.

[Mechatrommer]

PS: Hacked DHO800 can do it, too.

all of the screen captures above...

[2N3055]

Ok, not quite a full division of peak-to-peak noise, but pretty close. And I guess you chose favorable conditions with the relatively fast timebase, which excludes most of the 1/f noise. Could you show this at a slower time base too, say 10 ms/div?

see attached 01,02,03 (peak, ultra, average 4) input 50 ohm terminated. i also captured my UTG962 AWG output 1mVpp but terminated with 12.5ohm (4x inline 50 ohm) because its not so low noise enough, 50Hz and 100KHz not so stable trigger due to noises (normal acq and average) ymmv. (ps: last image... how to estimate dut noise from combined scope + dut noise ymmv.)



ps: i didnt notice 20MHz BW limit activated during the capture, it wasnt me! well you got the idea...

So to make it clear about those images:

- 20MHz BW limit is applied by the scope automatically and you cannot switch it of.
- Scope samples at 6.25 MSps/s (which is good for 3MHz without aliasing, but for noise not important)

- In image 01_peak.png it shows 271µV P-P and 21µV RMS (12x ratio points to non Gausian noise)
- In image 02_ultra.png it shows 226µVP-P and 48µV RMS using same settings for vertical and timebase.

Which result is true ?

[Mechatrommer]

ps: i didnt notice 20MHz BW limit activated during the capture, it wasnt me! well you got the idea...

20MHz BW limit is applied by the scope automatically and you cannot switch it of.

something activated it. i think when i change to Ultra acquisition. i can deactivate it, i just didnt aware it was activated by not me.

Scope samples at 6.25 MSps/s (which is good for 3MHz without aliasing, but for noise not important)

eblaster asked me to do 10ms timescale, so surely sample rate will be reduced to that automatically.

- In image 01_peak.png it shows 271µV P-P and 21µV RMS (12x ratio points to non Gausian noise)
- In image 02_ultra.png it shows 226µVP-P and 48µV RMS using same settings for vertical and timebase.

Which result is true ?

i will look closely in more detail about this later cheers.

[shapirus]

20MHz BW limit is applied by the scope automatically and you cannot switch it of.

something activated it. i think when i change to Ultra acquisition. i can deactivate it, i just didnt aware it was activated by not me.

It is forced on when vertical scale is set to 200 uV/div, and it's not possible to turn it off for that setting. It's not enforced at 500 uV/div and above.

[2N3055]

eblaster asked me to do 10ms timescale, so surely sample rate will be reduced to that automatically.

You can set memory to longer...

[Mechatrommer]

20MHz BW limit is applied by the scope automatically and you cannot switch it of.

something activated it. i think when i change to Ultra acquisition. i can deactivate it, i just didnt aware it was activated by not me.

It is forced on when vertical scale is set to 200 uV/div, and it's not possible to turn it off for that setting. It's not enforced at 500 uV/div and above.

ok confirmed.

eblaster asked me to do 10ms timescale, so surely sample rate will be reduced to that automatically.

You can set memory to longer...

right! see attached... single capture, and while running.. good enough for 150MHz? what else did i miss? since its 20MHz BW limited, then its good enough to 20MHz without aliasing... correct?

[2N3055]

20MHz BW limit is applied by the scope automatically and you cannot switch it of.

something activated it. i think when i change to Ultra acquisition. i can deactivate it, i just didnt aware it was activated by not me.

It is forced on when vertical scale is set to 200 uV/div, and it's not possible to turn it off for that setting. It's not enforced at 500 uV/div and above.

ok confirmed.

eblaster asked me to do 10ms timescale, so surely sample rate will be reduced to that automatically.

You can set memory to longer...

right! see attached... single capture, and while running.. good enough for 150MHz? what else did i miss? since its 20MHz BW limited, then its good enough to 20MHz without aliasing... correct?

You know well what Nyquist is, I know that. I wasn't commenting your knowledge.
But we had show previously that DHO had some bugs with measurements and also some weird inconsistencies in how signal amplitude was shown with different settings(memory lengths for instance).
I believe this is what we see now.

On your last image values are now 273µV P-P and 60µV RMS.
Another different value.

Make note that RMS values of noise of 20MHz BW limited front end should not change, even when undersampled. Frequencies will just fold down but energy stays the same.

[Mechatrommer]

Make note that RMS values of noise of 20MHz BW limited front end should not change, even when undersampled. Frequencies will just fold down but energy stays the same.

i dont know. if we go back to 1st principle (link below)... lower RMS value means more samples are very near to best fit line vice versa, so i dont know if the noise is "non-random" resulting varying RMS, or implementation is wrong. i dont know which one.. but cases such as switching noise of smps specified in mVpp, we dont need any of RMS formula, just eyeballing down to noise floor say.. 300uVpp, not easy to achieve smps that clean. but for dut with white noise very close to dso noise floor, we need to make assumption, say dso noise is gaussian or white or blue and use appropriate formula to estimate dut noise, no need to rely on dso built-in measurement if suspected to be buggy, they are just bell and whistle. we have cursors that can help to measure Vpp as well... i dont know much details of those statistical measurements. but imho lower uV/div scale can make visual estimation easier for low noise dut. ymmv.

1st principle of ac rms?
https://support.waters.com/KB_Inf/Empower_Tips_of_the_Week/WKB244139_Baseline_Noise_Root_Mean_Squared_Noise_RMS_Calculations

dc and ac rms:
https://en.wikipedia.org/wiki/Root_mean_square

[gf]

RMS usually includes the DC offset, so better measure standard deviation instead.

[shapirus]

RMS usually includes the DC offset, so better measure standard deviation instead.

there's a special measurement function called "AC Vrms" or something like that.

[Mechatrommer]

i was refering to ac rms.. link above corrected.. its possible i was activating dc+ac rms in dho800... ymmv.
1st principle of ac rms: https://support.waters.com/KB_Inf/Empower_Tips_of_the_Week/WKB244139_Baseline_Noise_Root_Mean_Squared_Noise_RMS_Calculations
dc and ac rms wiki: https://en.wikipedia.org/wiki/Root_mean_square