Author Topic: Rigol DHO800 memory depth / bandwidth upgrades and noise level!  (Read 18020 times)

0 Members and 2 Guests are viewing this topic.

Online ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #100 on: November 12, 2023, 09:08:47 pm »
That would be a tad more than 11 bits of effective resolution, right? Assuming 9 vertical divisions (with a bit of margin at the top and bottom), it's 9*50 mv / 0.2 mV = 2250 > 2048.
 

Online Fungus

  • Super Contributor
  • ***
  • Posts: 17519
  • Country: 00
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #101 on: November 12, 2023, 09:13:47 pm »
Can you try what I suggested and post a screenshot of your scope at 20MHz and Full BW, 1ms/div, open connectors but at 1V/div?  I'm curious what you'd get for ENOB when you get out of the lowest scales.

Second attempt, with the right vertical scale:


CH1+CH2 @20MHz, CH3+CH4 full.
 

Online ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #102 on: November 12, 2023, 09:18:00 pm »
Hmm -- pity that you deleted the screenshots taken at 50 mV/div. Those actually had a higher ENOB (yes, I know that's not quite it...) than at 1 V/div.

For the 1 V/div acquisitions, I get 9*1000 mV / 5 mV = 1800, so a bit less than 11 bits.
 

Online Fungus

  • Super Contributor
  • ***
  • Posts: 17519
  • Country: 00
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #103 on: November 12, 2023, 09:21:29 pm »
Hmm -- pity that you deleted the screenshots taken at 50 mV/div. Those actually had a higher ENOB (yes, I know that's not quite it...) than at 1 V/div.

I can do them again:

 
The following users thanked this post: ebastler

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 8175
  • Country: us
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #104 on: November 12, 2023, 09:30:36 pm »
For the 1 V/div acquisitions, I get 9*1000 mV / 5 mV = 1800, so a bit less than 11 bits.

9?  But yeah, I'm seeing 10+ ENOB, pretty good actually.
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Online ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #105 on: November 12, 2023, 09:33:20 pm »
For the 1 V/div acquisitions, I get 9*1000 mV / 5 mV = 1800, so a bit less than 11 bits.

9?  But yeah, I'm seeing 10+ ENOB, pretty good actually.

I don't have the DHO800 myself, but was assuming that the full ADC range is not mapped to exactly 8 vertical divisions, but that there is approx. half a division of "margin" at the top and bottom.

Edit: Dave shows it here for the MSO4000.
« Last Edit: November 12, 2023, 09:36:57 pm by ebastler »
 

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 8175
  • Country: us
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #106 on: November 12, 2023, 09:33:56 pm »
So here is the SDS1204X-E, the second one is expanded after capture to show that there's essentially 1LSB of flipping, so the noise probably can't get any better.  But I get an ENOB of a bit more than 8 bits--8.97 7.5 based only on noise?  IOW, this is as far as 8-bit tech can go here and the DHO800 clearly exceeds that at these scales.




« Last Edit: November 12, 2023, 09:42:27 pm by bdunham7 »
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Offline Martin72

  • Super Contributor
  • ***
  • Posts: 7019
  • Country: de
  • Testfield Technician
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #107 on: November 12, 2023, 09:34:36 pm »
Determining the ENOBs ( at 10Mhz) of a DHO4000 and R&SMX04:

https://electronicprojectsforfun.wordpress.com/to-enob-or-not-to-enob/

 
The following users thanked this post: egonotto

Online gf

  • Super Contributor
  • ***
  • Posts: 1425
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #108 on: November 12, 2023, 09:39:38 pm »
For the 1 V/div acquisitions, I get 9*1000 mV / 5 mV = 1800, so a bit less than 11 bits.

9?  But yeah, I'm seeing 10+ ENOB, pretty good actually.

But 9Vpp (provided that is the full scale range of the ADC) is only 3.182VRMS, for a full scale sine wave. Then the SNR is 20*log10(3.182/0.005) = 56.075 dBFS, or equivalently (56.075-1.76)/6.02 = 9.0224 ENOB.
 
The following users thanked this post: rf-loop, ebastler, 2N3055

Online bdunham7

  • Super Contributor
  • ***
  • Posts: 8175
  • Country: us
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #109 on: November 12, 2023, 09:41:31 pm »
But 9Vpp (provided that is the full scale range of the ADC) is only 3.182VRMS, for a full scale sine wave.

Doh!  I knew I had some error in there somewhere.
A 3.5 digit 4.5 digit 5 digit 5.5 digit 6.5 digit 7.5 digit DMM is good enough for most people.
 

Online ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #110 on: November 12, 2023, 09:55:03 pm »
But 9Vpp (provided that is the full scale range of the ADC) is only 3.182VRMS, for a full scale sine wave.

Doh!  I knew I had some error in there somewhere.

Same here... Thanks!
 

Online Fungus

  • Super Contributor
  • ***
  • Posts: 17519
  • Country: 00
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #111 on: November 13, 2023, 02:07:48 am »
I don't have the DHO800 myself, but was assuming that the full ADC range is not mapped to exactly 8 vertical divisions, but that there is approx. half a division of "margin" at the top and bottom.

I can grab a 2Vpp signal in the 200mV/div range. That means it has a whole division above/below the screen.  :)



9?

This one goes to 10!
 

Online Fungus

  • Super Contributor
  • ***
  • Posts: 17519
  • Country: 00
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #112 on: November 13, 2023, 02:17:23 am »
But 9Vpp (provided that is the full scale range of the ADC) is only 3.182VRMS, for a full scale sine wave. Then the SNR is 20*log10(3.182/0.005) = 56.075 dBFS, or equivalently (56.075-1.76)/6.02 = 9.0224 ENOB.

On the 1V range: 20*log10(3.535/0.005) = 56.998dB = 9.17 ENOB

Hmm -- pity that you deleted the screenshots taken at 50 mV/div. Those actually had a higher ENOB

On the 50mV range I get ... 64.95dB = 10.5 ENOB
« Last Edit: November 13, 2023, 03:02:41 am by Fungus »
 

Online 2N3055

  • Super Contributor
  • ***
  • Posts: 7466
  • Country: hr
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #113 on: November 13, 2023, 07:38:45 am »
But 9Vpp (provided that is the full scale range of the ADC) is only 3.182VRMS, for a full scale sine wave. Then the SNR is 20*log10(3.182/0.005) = 56.075 dBFS, or equivalently (56.075-1.76)/6.02 = 9.0224 ENOB.

On the 1V range: 20*log10(3.535/0.005) = 56.998dB = 9.17 ENOB

Hmm -- pity that you deleted the screenshots taken at 50 mV/div. Those actually had a higher ENOB

On the 50mV range I get ... 64.95dB = 10.5 ENOB


Martin posted a link to article about ENOB measurements. It seems not all read it, but here we are, magic numbers are coming out of the hat.
Measuring ENOB in scopes is made in accordance to IEEE Std 1057â„¢ standard "IEEE Standard for Digitizing Waveform Recorders".
There are many nice articles by various scope manufacturers on the topic.

ENOB is calculated from SINAD (Signal to noise and distortion) ratio. It is measured by inputting low distortion sinewave (better than scope measured) and sweeping across a frequency BW of scope. ENOB is measured as a curve, a graph, not a single number.
It will vary with with frequency.
On scopes it is measured with signal that is 90% of full screen.In this case it would be 7.2V P-P for DHO800 1V/div.
That curve than gets statistics treatment and some kind of average or worst case scenario number is picked by marketing. Siglent publishes worst case scenario ENOB number with some reserve included, for instance, making it very conservative number. They guarantee it will be better than number published.  It is always better than that. Most manufacturers are the same in that regard. Rarely manufacturers (any of them) publish full characterisation of ENOB curves directly in datasheet.

It is definitely not a ratio of scope AC RMS measurement of noise floor and theoretical RANGE of ADC.

Since really measuring ENOB is out of reach of many (no wide BW signal source of sufficient quality) we can compare and talk about just noise floor levels VS full screen scale. Which is definitely not the same as ENOB.In which case it is simply SNR (signal to noise ratio). Expressed in dB.
We could talk about bits equivalent of SNR but that is not usually made because it has no practical meaning.

But this is not ENOB. ENOB needs signal applied and  includes distortions (nonlinearities, sampling jitter...) . It is already a simplified metrics....

One another note: DHO800 has 8v P-P range on screen. That equates to 2,8369 V RMS sinewave.
That is your usable full dynamic range. The rest is reserve to prevent clipping.
"Just hard work is not enough - it must be applied sensibly."
Dr. Richard W. Hamming
 
The following users thanked this post: Performa01, egonotto, Martin72, hatte

Online Fungus

  • Super Contributor
  • ***
  • Posts: 17519
  • Country: 00
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #114 on: November 13, 2023, 09:14:05 am »
On scopes it is measured with signal that is 90% of full screen.In this case it would be 7.2V P-P for DHO800 1V/div.

That's when you're measuring THD, which we aren't. If you're measuring noise you should use the full range of the ADC for most accurate results.

It is definitely not a ratio of scope AC RMS measurement of noise floor and theoretical RANGE of ADC.

https://en.wikipedia.org/wiki/Effective_number_of_bits

The SNR tells you the ENOB when distortion is zero.

As noted: Calculating THD requires a lot of fancy gear that most people simply don't have.

OTOH it's a safe bet that THD is small, almost negligible. SNR will give a good approximation. It's the best number we have until more data comes along.

It's also a safe bet that using the same method on different 'scopes is a good way to compare them. We measured the "benchmark" Siglent using the same yardstick and found the new Rigol has two or three more bits more.

I think that's a fair assumption to make.

ie. Rigol's 12 bits are a big advance and this little 'scope has proved itself to be the new reference point.
 

Online ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #115 on: November 13, 2023, 09:44:02 am »
[ENOB] is definitely not a ratio of scope AC RMS measurement of noise floor and theoretical RANGE of ADC.

Yes, I know. That's why I initially talked about "effective resolution", and then added a disclaimer ("I know that's not quite it") when I mentioned ENOB later. Using the full ADC range was obvioulsy a brain fart, but when using full RMS signal amplitude instead -- how big is the difference between proper ENOB and that simplified "RMS full swing / RMS noise" in practice?

Quote
One another note: DHO800 has 8v P-P range on screen. That equates to 2,8369 V RMS sinewave.
That is your usable full dynamic range. The rest is reserve to prevent clipping.

Not sure I agree on this. The oscilloscope has the larger range captured, as witnessed by the ability to zoom out (vertically) on the stored data. You can also use the full range when you want to zoom in on small signal details which sit on top of a large signal swing. And I am pretty sure that the extended vertical range can also be used for measurements.
 
The following users thanked this post: egonotto, Fungus

Online gf

  • Super Contributor
  • ***
  • Posts: 1425
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #116 on: November 13, 2023, 09:55:11 am »
ENOB is calculated from SINAD (Signal to noise and distortion) ratio. It is measured by inputting low distortion sinewave (better than scope measured) and sweeping across a frequency BW of scope. ENOB is measured as a curve, a graph, not a single number.
It will vary with with frequency.

Yes, actually it should be calculated from SINAD, and it is a function of frequency.
Calculating it from SNR is just a first order approximation, assuming distortion is low.
At low frequencies it is not unlikely that this is granted, but at high frequencies I would not be so sure.

EDIT: DC non-linearity could also be measured during factor calibration, and compensated. But of course I don't know if the manufacturer does that.
« Last Edit: November 13, 2023, 10:13:39 am by gf »
 
The following users thanked this post: Performa01, egonotto

Online 2N3055

  • Super Contributor
  • ***
  • Posts: 7466
  • Country: hr
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #117 on: November 13, 2023, 10:22:19 am »
On scopes it is measured with signal that is 90% of full screen.In this case it would be 7.2V P-P for DHO800 1V/div.

That's when you're measuring THD, which we aren't. If you're measuring noise you should use the full range of the ADC for most accurate results.

It is definitely not a ratio of scope AC RMS measurement of noise floor and theoretical RANGE of ADC.

https://en.wikipedia.org/wiki/Effective_number_of_bits

The SNR tells you the ENOB when distortion is zero.

As noted: Calculating THD requires a lot of fancy gear that most people simply don't have.

OTOH it's a safe bet that THD is small, almost negligible. SNR will give a good approximation. It's the best number we have until more data comes along.

It's also a safe bet that using the same method on different 'scopes is a good way to compare them. We measured the "benchmark" Siglent using the same yardstick and found the new Rigol has two or three more bits more.

I think that's a fair assumption to make.

ie. Rigol's 12 bits are a big advance and this little 'scope has proved itself to be the new reference point.

Yes we all know you are smarter than IEEE combined...

SNR tells you what is SNR. Without signal. There is no "zero distortion" in real life. Why do you think ADCs with nominal 12 bit resolution barely break 10bit ENOB?
Nonlinearities. So no it is not negligible and no it is not even close to good approximation. Take a look at  IEEE Std 1057â„¢. It is 181 pages or so...
We measure and calculate. As I said there are excellent white papers by R&S, Keysight and Tek on the topic. Also few good ones on EDN.
No fancy anything needed for that. Just will to read and learn.

"An encyclopedia (American English) or encyclopædia (British English) is a reference work or compendium providing summaries of knowledge either general or special
to a particular field or discipline" . Quote from Wikipedia itself. 
Wikipedia is encyclopedia. It is just a summary to explain what word means. And to point you in the right direction where can knowledge of the topic be found. It is simply the index to give you starting point on a path to knowledge. It is not THE knowledge.

Fact that you cannot measure it means exactly that. YOU cannot measure it. Live with it. You have no clue what ENOB is on DHO800 and you never will unless either you acquire capability to measure it, or someone measure it for you. Stop inventing fantasy numbers. If you want to gamble go to bookies..

Comparing SNR is not wrong. I sad that. But you keep inventing "your special" ways of doing it. To be able to compare you need to compare at same conditions.
I told you how: full screen is considered full range. By Tektronix, Keysight LeCroy, Siglent ... For your scope that is 8 div vertical. 8 V P-P

At 1 V/div and full BW your scope has cca 55dB od SNR. At 20 MHz BW about 56-57 dB.

55dB at almost 200MHz BW is not bad. Not bad at all.



« Last Edit: November 13, 2023, 10:53:33 am by 2N3055 »
"Just hard work is not enough - it must be applied sensibly."
Dr. Richard W. Hamming
 
The following users thanked this post: Performa01, egonotto, Warhawk

Online 2N3055

  • Super Contributor
  • ***
  • Posts: 7466
  • Country: hr
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #118 on: November 13, 2023, 10:28:07 am »
ENOB is calculated from SINAD (Signal to noise and distortion) ratio. It is measured by inputting low distortion sinewave (better than scope measured) and sweeping across a frequency BW of scope. ENOB is measured as a curve, a graph, not a single number.
It will vary with with frequency.

Yes, actually it should be calculated from SINAD, and it is a function of frequency.
Calculating it from SNR is just a first order approximation, assuming distortion is low.
At low frequencies it is not unlikely that this is granted, but at high frequencies I would not be so sure.

EDIT: DC non-linearity could also be measured during factor calibration, and compensated. But of course I don't know if the manufacturer does that.

Yes, as I said we can compare SNR as it is a simple measurement and can be done in consistent manner and give good insight in large part of performance envelope.
If scope SNR is bad , ENOB won't be better.

As a rough estimate, if two scopes have same ranges and similar BW simply putting them in same range makes simple noise AC RMS comparisons meaningful.
If scope has 10 vertical divisions and another one has 8 at 1V/div they won't be the same but simple linear scaling works here.

Some scopes have nonlinearity compensation as a part of ADC calibration of masivelly interleaved ADCs (Keysight for instance. Maybe Rigol does something similar.in its ASIC ADC) Some have digital compensation after ADC, but not in this price range.
« Last Edit: November 13, 2023, 10:32:41 am by 2N3055 »
"Just hard work is not enough - it must be applied sensibly."
Dr. Richard W. Hamming
 
The following users thanked this post: egonotto

Online 2N3055

  • Super Contributor
  • ***
  • Posts: 7466
  • Country: hr
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #119 on: November 13, 2023, 10:44:20 am »
[ENOB] is definitely not a ratio of scope AC RMS measurement of noise floor and theoretical RANGE of ADC.

Yes, I know. That's why I initially talked about "effective resolution", and then added a disclaimer ("I know that's not quite it") when I mentioned ENOB later. Using the full ADC range was obvioulsy a brain fart, but when using full RMS signal amplitude instead -- how big is the difference between proper ENOB and that simplified "RMS full swing / RMS noise" in practice?

Quote
One another note: DHO800 has 8v P-P range on screen. That equates to 2,8369 V RMS sinewave.
That is your usable full dynamic range. The rest is reserve to prevent clipping.

Not sure I agree on this. The oscilloscope has the larger range captured, as witnessed by the ability to zoom out (vertically) on the stored data. You can also use the full range when you want to zoom in on small signal details which sit on top of a large signal swing. And I am pretty sure that the extended vertical range can also be used for measurements.


In practice difference will be exactly based on how much distortion is there. Make note that whole spectra of distortion artefacts from full BW will add up to RMS of baseline. Death by thousands cuts type of scenario.

As for vertical range used for calculation, I said what is industry practice. It stems from the fact that scope screen is visual reference and part that manufacturer guarantees specifications for. Not my invention.

By all means, if you use scope as sampler, and pull data somewhere else and postprocess, characterize it fully for full usable dynamic range (by characterizing distortion and clipping behaviour) and enjoy.
But on a scope you are left at mercy of manufacturer provided usage envelope.
Small part of dynamic range is reserve not mapped in normal use for user.
It is same for everyone. It is still fair if you compare like for like..
"Just hard work is not enough - it must be applied sensibly."
Dr. Richard W. Hamming
 
The following users thanked this post: egonotto

Offline Martin72

  • Super Contributor
  • ***
  • Posts: 7019
  • Country: de
  • Testfield Technician
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #120 on: November 13, 2023, 10:55:59 pm »
At the moment I am working on the ENOB measurement from the linked page as a kind of side project.
I have already finished the 10Mhz oscillator and successfully put it into operation, I have summarized the filters used for it and designed 2 circuit boards for it, I will soon know if this works.
And open a new thread for it, the test object will of course be my HD from siglent - but I also have the DHO804.
The website operator is a user from here, he has measured a DHO4000 there and an MX04.
At first I was amazed that he came up with such "high" results, over 10 bits after all.
Rigol had simply stated "over 8bit" in the data sheet, siglent is a bit more precise and declares values between 8.4 and 8.7.
When I later looked into the ENOB thing, it was clear why.
And therefore it is also clear why this thread comes to just under 11bit, because the frequency is left out or is much lower(10Mhz) than used by the manufacturers.
Very professional, by the way:
Rigol simply states "over 8 bit" and conceals the measuring circumstances.
Siglent also states the underlying frequency (50 or 100Mhz) in addition to the more precise values.
It was clear to me that ENOB depends on the frequency, among other things.
More on this soon, probably towards the end of the year when I don't have to work. ;)



 
The following users thanked this post: Performa01, egonotto, 2N3055, UK

Online gf

  • Super Contributor
  • ***
  • Posts: 1425
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #121 on: November 14, 2023, 08:40:02 am »
Determining the ENOBs ( at 10Mhz) of a DHO4000 and R&SMX04:

https://electronicprojectsforfun.wordpress.com/to-enob-or-not-to-enob/

Let me supplement this with a fairly comprehensive application note from LeCroy:
https://cdn.teledynelecroy.com/files/appnotes/computation_of_effective_no_bits.pdf
 
The following users thanked this post: Performa01, egonotto, 2N3055, Martin72, mawyatt

Online gf

  • Super Contributor
  • ***
  • Posts: 1425
  • Country: de
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #122 on: November 14, 2023, 03:55:48 pm »
https://electronicprojectsforfun.wordpress.com/to-enob-or-not-to-enob/

If the aim is to measure only at a single frequency, then I wonder, why not use a high-Q bandpass filter?

Google found for me e.g. these crystal filters:
https://www.amazon.de/Kristallfilter-Schmalbandfilter-Langlebig-spezielle-Instrumente/dp/B09MJP1Y14
They seems to be available from various sellers, and also for a couple of different frequencies, e.g. here
https://de.aliexpress.com/item/1005003627277580.html

They promise > 80dB stopband attenuation, and if they deliver what they promise, this should be more than enough for measuring a 12-bit ADC (-74dBFS quantization noise), in conjunction with an AWG like SDG2000X which achieves already -65dB harmonic distortion up to 10MHz and -70dB non-harmonic spurs out of the box (according to the datasheet).
 
The following users thanked this post: egonotto

Offline Martin72

  • Super Contributor
  • ***
  • Posts: 7019
  • Country: de
  • Testfield Technician
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #123 on: November 14, 2023, 10:50:50 pm »
Interesting comments, but I would like to take them up in a separate thread.
That would go beyond the scope of this article.
 
The following users thanked this post: egonotto

Offline JDW

  • Frequent Contributor
  • **
  • Posts: 336
  • Country: jp
Re: Rigol DHO800 memory depth / bandwidth upgrades and noise level!
« Reply #124 on: January 17, 2024, 04:25:57 am »


When I try to repeat the same test using the same settings, even when testing only a single channel (leaving the probe disconnected as Fungus apparently did), I get 60Hz noise that makes it impossible to detect the true noise floor.  Noise remained unchanged with or without the ground cable attached to wall socket Earth Ground.

Note that my Probe switch was set to 10x, and the "Probe" setting (in CH1) was also set to 10x.
« Last Edit: January 17, 2024, 06:00:10 am by JDW »
 
The following users thanked this post: egonotto


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf