Siglent SDS2000X HD, 12bit, 2GSa/s high definition oscilloscope serie.

[2N3055]

Here are some Screenshots from my SDS2104X HD.

First Picture:      50 Ohm and 20 MHz Limit activated on all Channels.
Second Picture:  1MOhm and 20 MHz Limit activated on all Channels.

It seems that there is not much noise level difference between both settings.

Cheers Chris

1 ms/div please.

[rf-loop]

And also for noise, please use StDev instead of RMS (root mean square of data). 
(In some oscilloscopes this is named as AC RMS but Siglent use its correct name StDev... Standard Deviation of the data)

In practice StDev corresponds AC RMS

RMS include AC + "DC".  (there exist or may exist some "DC" offset residues even when well calibrated and this must not gategorize as noise)

[dekagon]

Hi,

1 ms/div please.

Sorry I didn't noticed that  Now the Timebase settings are done.

And also for noise, please use StDev instead of RMS (root mean square of data). 
(In some oscilloscopes this is named as AC RMS but Siglent use its correct name StDev... Standard Deviation of the data)

In practice StDev corresponds AC RMS

RMS include AC + "DC".  (there exist or may exist some "DC" offset residues even when well calibrated and this must not gategorize as noise)

Ah OK. I changed the settings and took new screenshots.


Cheers Chris

[Martin72]

If this is to be a comparative measurement compared to egonotto's measurements, then the same conditions should also prevail - so no average please...

[dekagon]

If this is to be a comparative measurement compared to egonotto's measurements, then the same conditions should also prevail - so no average please...

Martin you are right. I played with so much feature today that I did not switched it off. The scope is full of very interesting details which can be tried but as a new Siglent scope owner I do not have the overview 
The right shots are ecoming soon   

Cheers Chris

[egonotto]

Here are some Screenshots from my SDS2104X HD.

First Picture:      50 Ohm and 20 MHz Limit activated on all Channels.
Second Picture:  1MOhm and 20 MHz Limit activated on all Channels.

It seems that there is not much noise level difference between both settings.

Cheers Chris

Hello,

please use Stdev on at least 1 ms/dev and more memory. With RMS, the offset is still taken into account and is not suitable for noise measurements.

Best regards
egonotto
PS: A joke for Germans:
Psophometer is a measuring device for recording and quantitatively evaluating the noise of electrical or acoustic signals.
Bsoffen is also another word for intoxicated, i.e. drunk.
in German:
Psophometer ist ein ist ein Messgerät zur Erfassung und quantitativen Bewertung des Rauschens von elektrischen oder akustischen Signalen.
Bsoffen ist aber auch ein anderes Wort für berauscht also betrunken.

[dekagon]

PS: A joke for Germans:
Psophometer is a measuring device for recording and quantitatively evaluating the noise of electrical or acoustic signals.
Bsoffen is also another word for intoxicated, i.e. drunk.
in German:
Psophometer ist ein ist ein Messgerät zur Erfassung und quantitativen Bewertung des Rauschens von elektrischen oder akustischen Signalen.
Bsoffen ist aber auch ein anderes Wort für berauscht also betrunken.

  

[egonotto]

Hello,

I have just seen that dekagon uses 4-fold averaging.
Please do not use average or ERES.

Best regards
egonotto

[dekagon]

Hi egonotto,

now I have everything changed to your settings from your SDS3000...
with the exception that I cannot get your Samplingrate. I can only to 1 GSa/s 

The first pic is yours and the 2nd is the compare with mine SDS2000

Cheers Chris

[2N3055]

Hello,

please use Stdev on at least 1 ms/dev and more memory. With RMS, the offset is still taken into account and is not suitable for noise measurements.

Best regards
egonotto

No, not at least.
Exactly at 1ms/div.

[Performa01]

In the forum there were instructions for measuring the scope bandwidth. Does anyone know which thread it was in?

I don’t know the forum post you’re referring to, but for any serious measurement a levelled signal generator is required.

Yet in a pinch you can use a less accurate method that requires no further equipment:


SDS2504X_HD_Base_Noise

This method assumes that the base noise in the scope frontend is white - which isn’t quite true, particularly not for frequencies below 20 MHz. Taking this into account, you just need to use 20 MHz as the reference frequency (instead of 1 MHz).

Most of the settings can be plainly seen in the screenshot:
Input coupling DC with 50 Ω input impedance, 500 µV/div.
Timebase 1 µs/div and AC Line trigger.

An FFT with Flattop window over the full span from 0 – 1 GHz and 1024 times averaging. For a 100 MHz instrument, a narrower span like 500 MHz could be used.

Now we can use tracking cursors to find the frequency where the amplitude has dropped by 3 dB with regard to the reference frequency (20 MHz):

3.07 (ΔY) dB Bandwidth = X2 = 595 MHz;

This result isn’t too far off; the actual bandwidth is about 585 MHz, as measured in max hold mode with a properly swept and levelled signal source:


SDS2504X_HD_FR_2x600s_500uV

[2N3055]

Hello,

I have just seen that dekagon uses 4-fold averaging.
Please do not use average or ERES.

Best regards
egonotto

As sidenote, on Rigol DHO800, Rigol does use additional software averaging (equivalent to ERES3 on SDS2000xHD) when you enable 20MHz..
It is automatic and not user configurable.
So much for equal measurements.

On SDS2000xHD (and all Siglents really) 20MHz BW limit is a physical frontend of channel.

[egonotto]

Hello,

interesting is the difference in noise when using 1 ms/div or 10 ms/div.

Best regards
egonotto

[egonotto]

...
As sidenote, on Rigol DHO800, Rigol does use additional software averaging (equivalent to ERES3 on SDS2000xHD) when you enable 20MHz..
It is automatic and not user configurable.
So much for equal measurements.

On SDS2000xHD (and all Siglents really) 20MHz BW limit is a physical frontend of channel.

Hello,

I was close to buying an HDO4000 about a year ago until I realized that Rigol was processing the data.

I later fell for the very affordable DHO1074.

Best regards
egonotto

[dekagon]

Hi,

I am fiddling with the internal Signalgenerator right now and saw  that there is an Firmware Upgrade option for this.

My SDS2104X HD tells me that there is a SAG2021 inside.

And an upgrade path for a Firmware file.

But Siglent unfortunately offers in the meantime no longer Firmware updates for this generator 

Will these updates are also done with an update of the scope firmware?

Cheers Chris

[Martin72]

I once recorded the individual stages of the bandwidth "extension"  of an SDS800X HD and merged them using the REF traces.

[tv84]

Will these updates are also done with an update of the scope firmware?

Yes.

[2N3055]

Hello,

interesting is the difference in noise when using 1 ms/div or 10 ms/div.

Best regards
egonotto

Yes it is interesting, but you should only compare like for like.

Power noise density is not a constant across the BW. Once you reach 1/f corner it goes crazy.
So comparing at 1ms/div and 10ms/div is not good.
Compare 1ms/div with 1ms/div and 10ms/div with 10ms/div only.

[2N3055]

...
As sidenote, on Rigol DHO800, Rigol does use additional software averaging (equivalent to ERES3 on SDS2000xHD) when you enable 20MHz..
It is automatic and not user configurable.
So much for equal measurements.

On SDS2000xHD (and all Siglents really) 20MHz BW limit is a physical frontend of channel.

Hello,

I was close to buying an HDO4000 about a year ago until I realized that Rigol was processing the data.

I later fell for the very affordable DHO1074.

Best regards
egonotto

Such processing is not a bad thing per se if done right and is fully advertised in specifications.
On my Siglents I will also use that sometimes to squeeze every little bit.
But it disingenuous to publish that data as figures for comparison with standard 20MHz BW limit on other scopes that does not do that.

OTOH, depending on implementation,  it can lead to bad measurements. If someone did measurements PSU noise with standard scope with standard 20MHz BW limit, using Rigol can lead to more optimistic results, because while measuring it will LP filter data with more aggressive slope, compared to reference data, showing less noise that it should be.

So I prefer Siglent's way (and everybody else's) to do this.

[egonotto]

Hello,

If I understand it correctly, with the setting 100 Mpts, 1 GSa/s and 1 ms/div or 10 ms/div, only the proportion of memory used for the screen display is different.

Best regards
egonotto

[2N3055]

Hello,

If I understand it correctly, with the setting 100 Mpts, 1 GSa/s and 1 ms/div or 10 ms/div, only the proportion of memory used for the screen display is different.

Best regards
egonotto

Measurement will be performed on dataset that is defined by screen boundaries.
If you set 100MPts fixed, scope will take fixed 100Mpts capture , but it always analyses only displayed portion worth of data.

So despite both being taken the same, it will be 10 ms and 100ms of data being analysed. And that will make a difference in what lowest frequency will be included in Stdev calculation.

[Performa01]

If I understand it correctly, with the setting 100 Mpts, 1 GSa/s and 1 ms/div or 10 ms/div, only the proportion of memory used for the screen display is different.

Memory depth for measurements is limited to 10 Mpts on an SDS800X HD and data gets decimated if the record length exceeds that value. that means, the effective sample rate for the measurements would drop to 1/10 if the record length is 100 Mpts. This would not affect noise measurement though, because of aliasing this is more or less independant from the sample rate.

I'm not quite sure what the limit for the SDS2000X HD is, but it is higher for sure - probably the same as SDS3000X HD, where it is 40 Mpts.

Long story short: if it were only for the memory depth and sample rate, than 1 ms/div or 10 ms/div should not make a difference for noise measurements.

Yet there is another, much more important aspect - the lower bandwidth limit. if the input noise of an oscilloscope were white, this would not matter at all, but the input architecture of 99% of general purpose DSOs has a very pronoinced 1/f noise characteristics. Therefore it makes a big difference if you use a lower bandwith limit of 100 Hz (1 ms/div) or 10 Hz (10 ms/div).

To get rid of the bigest part of the strong LF-noise, a lower bandwidth limit of 100 kHz (1 µs/div) is the preferred setting.

[tautech]

Hi,

I am fiddling with the internal Signalgenerator right now and saw  that there is an Firmware Upgrade option for this.

My SDS2104X HD tells me that there is a SAG2021 inside.

And an upgrade path for a Firmware file.

But Siglent unfortunately offers in the meantime no longer Firmware updates for this generator 

Will these updates are also done with an update of the scope firmware?

Cheers Chris

Simply, if required, yes.

To date there have not been any specified updates for the internal AWG.
https://siglentna.com/wp-content/uploads/dlm_uploads/2024/07/SDS2000X-HD-Firmware-Revision-Record-and-Upgrade-Instructions.pdf

[dekagon]

Hi tautech,

Simply, if required, yes.

To date there have not been any specified updates for the internal AWG.
https://siglentna.com/wp-content/uploads/dlm_uploads/2024/07/SDS2000X-HD-Firmware-Revision-Record-and-Upgrade-Instructions.pdf

I only asked because I was a bit surprised that the generator is marked as obsolete by Siglent, but is built into the scopes.
I am very satisfied with the quality (despite the bandwidth limitation of 25 MHz). It is probably sufficient for most tests.
But I already know what my next purchase will be: a Siglent SDG2122X generator

The SDS2000X HD is a great scope. I am just sniffing the edges of his abilities right now 

Cheers Chris

[dekagon]

Hi Performa01,

SDS2504X_HD_Base_Noise

This method assumes that the base noise in the scope frontend is white - which isn’t quite true, particularly not for frequencies below 20 MHz. Taking this into account, you just need to use 20 MHz as the reference frequency (instead of 1 MHz).

Most of the settings can be plainly seen in the screenshot:
Input coupling DC with 50 Ω input impedance, 500 µV/div.
Timebase 1 µs/div and AC Line trigger.

An FFT with Flattop window over the full span from 0 – 1 GHz and 1024 times averaging. For a 100 MHz instrument, a narrower span like 500 MHz could be used.

Now we can use tracking cursors to find the frequency where the amplitude has dropped by 3 dB with regard to the reference frequency (20 MHz):

3.07 (ΔY) dB Bandwidth = X2 = 595 MHz;

This result isn’t too far off; the actual bandwidth is about 585 MHz, as measured in max hold mode with a properly swept and levelled signal source:

I finally managed to reproduce your example with the noise and you can see the result here in the image SDS2504X HD_PNG_5.png.
Fortunately it's pretty much the same as yours

Another try with a with noise source with 500mVpp was in image SDS2404X HD_7.png.
The full FFT curve is visible in image SDS2404X HD_8.png.
I do not know why there is this falldown at 100 MHz. After that it is going normally to the limit.

It seems that the noise source was also a possible measurement for the bandwidth but you are right. Exact measuremenst are only possible with a good sweep generator.

Unfortunately I don't have a good generator for such well-leveled sweep. This is on my wishlist

Cheers Chris