Siglent SDS3000X HD and upgraded SDS1000X HD

[Martin72]

Hi egonotto,

What do we see there?
I'm guessing an FFT of the intrinsic noise, right?
What settings have been made to reproduce it?

[egonotto]

Hello,

Yes, I looked at my own noise with the input open.

Here are the pictures with the settings.
The amplitudes of the spikes are quite small. At 1 V/div only about 2 mVpp. At 1 mV/div it is about 6 uVpp. 

Best regards
egonotto

[tautech]

Hello,

Yes, I looked at my own noise with the input open.

Here are the pictures with the settings.
The amplitudes of the spikes are quite small. At 1 V/div only about 2 mVpp. At 1 mV/div it is about 6 uVpp. 

Best regards
egonotto

Change Mem depth to 10+ Mpts to eliminate spurious FFT results.
Please report findings.

[Martin72]

Hi egonotto,

Can reproduce it.
@Rob:
Even with more memory, nothing changes.

[egonotto]

Hello,

unfortunately, it hasn't gotten any better, as Martin72 confirms.

Best regards
egonotto

[Martin72]

The distances between the peaks are always the same, always 31.25Mhz.

[egonotto]

Hello,

what surprises me more than the spikes is the frequency response. I would have thought that the filter would be much clearer at 200 MHz or 20 MHz.

Best regards
egonotto

[Martin72]

As can be seen, these filters have no effect on the amplitudes of the peaks, even at higher frequencies, which suggests a non-natural origin.

[gf]

31.25 = 4000 / 128
My first guess would be interleaving spurs.
But does the ADC really interleave that many branches, running a such a low sample rate?

[egonotto]

Hello,

31,25 * 32 = 1000
31,25 * 128 = 4000

Perhaps this has something to do with the 4 GS/s.

Best regards
egonotto

[Martin72]

Just tried it, no.
At 2 GSa/s (2nd ADC channel active) nothing changes and even if you use fixed sample rate and select 1GSa/s, the 31.25Mhz distance between the peaks remains.

[egonotto]

Hello,

my suspicion is still that the peaks have something to do with the 4 GS/s. However, I have not yet found such stable quartz oscillators with 31.25 MHz.

Best regards
egonotto

[gf]

See https://www.analog.com/en/resources/analog-dialogue/articles/interleaving-adcs.html

The symptoms are typical, but I find it unusual if the ADC was 128-way interleaved, each branch running at only 31.25 MSa/s. But who knows.

[tautech]

Hi egonotto,

Can reproduce it.
@Rob:
Even with more memory, nothing changes.

Okay, in the Siglent forum we can see this being addressed in beta FW for another model.
Seems like it may be an issue in some models but it's been identified and Siglent should, one would expect, check all models and correct as/if necessary.

[egonotto]

Hello,

I believe that this is not a problem because the amplitude of the interference is so small.

As far as I can remember, you can also see similar interference with the PicoScope 5000.

It would be interesting to have a look at the Batronix Magnova, which also has a good FFT. I would be surprised if you couldn't see similar spikes there.

Best regards
egonotto

[Performa01]

Being an older LeCroy design, this particular 12 bit, 4/5 GSa/s ADC runs at a moderate sample rate per core, thus producing a dense spurious spectrum, albeit at rather low levels, as has been discussed in this forum before (reply #346 and following):

https://www.eevblog.com/forum/testgear/siglent-sds6000-pro-10-and-12-bit-dso-coming/msg5240349/#msg5240349

The bandwidth limiters for all models up to the SDS6000 are 1st order filters located in the analog frontend (part of the PGA). They only affect input signals (including analog noise) and not granular noise from the ADC. Consequently, we will not see any difference at low sensitivities like 1 V/div, because there is just no relevant analog noise. It will look very different at 1 mV/div though.

[AndyC_772]

I can confirm similar spikes on my SDS3034X HD.

Can't say I'm bothered personally, though. An instrument's noise floor will always be governed by something, and there's no reason to insist that "something" must always be white noise with no particular structure to its spectrum.

[Martin72]

Hi,

It would be interesting to have a look at the Batronix Magnova, which also has a good FFT.

There's not much to see...
On the other hand, with the LeCroy WavePro 12Bit, 2.5Ghz, you can see the same spectral distribution over the span, the distances are always the same at about 39Mhz instead of 31.25Mhz for the SDS3104X HD.
On the other hand, the amplitudes are significantly lower with the LeCroy.

[2N3055]

Spikes amplitude will be highly dependent on input settings (V/div).
If you setup 3000xHD to higher sensitivity where top of dynamic range will be -10dBm, its spikes will be less too.

As for Magnova, that is nice clean noise floor, similar to 800/1000xHD that have "similar" ADC.

[Martin72]

I had used the same vertical setting on the LeCroy: 1V/div, 1M.
I'll check tomorrow to make sure the attenuation is also set to 1:1, not that it's set to 10:1...

[2N3055]

I had used the same vertical setting on the LeCroy: 1V/div, 1M.
I'll check tomorrow to make sure the attenuation is also set to 1:1, not that it's set to 10:1...

I have no doubt what you have set.
Just make sure that SpecAn app didn't "optimize" something...

[mawyatt]

The SDS800X HD @ 2GSPS, 1GHz 2Mpt FFT @ 500uv/Div vertical scale Open Circuit, shows a couple spurs at ~216MHz ~-110dBv, next level is ~-117dBV at ~264MHz and ~-115dBv at 312MHz. Impressive for a ~$500 12Bit DSO IMO.

Does the SDS2000X HD utilize the same ADCs as the 800?

Edit updated image to show markers on spurs.

Best

[egonotto]

I had used the same vertical setting on the LeCroy: 1V/div, 1M.
I'll check tomorrow to make sure the attenuation is also set to 1:1, not that it's set to 10:1...

Hello,

it looks to me that the actual setting for the LeCroy is 100 mV/div. Then the spikes on the LeCroy are about 10 dB lower than on the Siglent.
5000 : 128 = 39.0625. Therefore I assume that the LeCroy has 5 GS/s.

Best regards
egonotto

[egonotto]

Hello,

with the SDS3000X HD you can no longer see so many spikes at 500 uV/div.

Best regards
egonotto

[Martin72]

I have no doubt what you have set.
Just make sure that SpecAn app didn't "optimize" something...

I wouldn't want to put my hand in the fire that it wouldn't be so.
Because with the “normal” FFT via the math function, you can't set almost anything; most things happen automatically.
And things like average, peak, markers, etc. are only available in the spectrum analyzer option, which you have to buy.