RME ADI-2/4 Pro SE - Auto-measured with Multitone

[charleski]

Ok, I came across an oddity.
Although I hadn't set the zoom levels properly when saving results, I did notice that the Linearity result for my motherboard from the test plan seemed really good. So I repeated it as a single test (I set averages to 32, same as the test plan, if that means anything). This time I got a far worse (though still marginally acceptable) result:

I.e. linear (+/- 0.5dB) to -96dB: -alright but not great.
But I then repeated the linearity test in the test plan and it came out a lot better:

Now it's linear down below -104dB!

Levels were the same, and I couldn't see anything in the test plan settings that was different to those for the single test. What could be causing the difference?

 

[pkane]

Ok, I came across an oddity.
Although I hadn't set the zoom levels properly when saving results, I did notice that the Linearity result for my motherboard from the test plan seemed really good. So I repeated it as a single test (I set averages to 32, same as the test plan, if that means anything). This time I got a far worse (though still marginally acceptable) result:
View attachment 276233
I.e. linear (+/- 0.5dB) to -96dB: -alright but not great.
But I then repeated the linearity test in the test plan and it came out a lot better:
View attachment 276234
Now it's linear down below -104dB!

Levels were the same, and I couldn't see anything in the test plan settings that was different to those for the single test. What could be causing the difference?

I think the linearity measurement I added to the test script used Coherent averaging instead of standard, Amplitude average. Try changing it for a non-test-plan test to Coherent to see if that's what makes the difference. Coherent averaging removes more of the noise from the linearity calculation, producing a cleaner result.

 

[charleski]

I think the linearity measurement I added to the test script used Coherent averaging instead of standard, Amplitude average. Try changing it for a non-test-plan test to Coherent to see if that's what makes the difference. Coherent averaging removes more of the noise from the linearity calculation, producing a cleaner result.

Yeah, that's it. Switched to Coherent and I get the better result on single tests.

 

[pkane]

Yeah, that's it. Switched to Coherent and I get the better result on single tests.

That’s why I used coherent averaging, otherwise noise starts to dominate the linearity measurement at low levels, resulting in a much worse apparent result.

 

[Rja4000]

I'v updated some of the plots in this post above.

The crosstalk now looks better

The specs says "channel separation >130dB"...

 

[pkane]

@pkane
Another idea:
You export a screen snapshot, which is good.
Why not to export also the measurements in a txt or XML file ?
Then we could easily build a graph that includes several measurements, for comparison.
(As an example, in Excel)

Or we could share them for others to use them as reference.

And, obviously, being able to attach such a result file to the tool for a measurement would also allow you to plot the 2 on the same graph.

(Warning: I can be quite creative as for tool development idea)

Not an XML but an HTML table is now generated as part of the test plan execution to document all the results:

https://www.audiosciencereview.com/...loopback-analyzer-software.27844/post-1552755

And, this version also lets you group multiple results on the same plot (and in the same table with multiple columns) using groups.

 

[andy3d]

about the new automated measurement capability of the Multitone app

Dear pkane. Actual version of Multitone installer, after unzipping is removed by Norton Security as beeing "dangerous" (so called WS.Reputation1). Just that You know

 

[pkane]

Dear pkane. Actual version of Multitone installer, after unzipping is removed by Norton Security as beeing "dangerous" (so called WS.Reputation1). Just that You know

That’s what AV software flags any new software without too many downloads. It’s not that it found anything dangerous, just that it’s too new to have a ‘reputation’.

 

[andy3d]

Multitone is a brilliant software! First time used it today, and it's so simple. I can use two different ASIO devices as output and input. Deltawave is also great !

I have a kind request to someone who has RME ADI-2/4 Pro SE:
Could someone measure and report it's results (1kHz THD+N, multitone THD+N, frequency response, etc in frequency) but for sampling rates of 384 kHz and 768 kHz? But - here is a trick - not for the audible range, but using lower limit as 200Hz up upper limit as high as possible for each sampling rate ? We are looking for a something better than Fireface or Motu USB 192kHz cards, to use it for acoustic measurements in concert halls scale models. We normally build 1:10 up to 1:20 scale model of hall's interiors, so we need to record sound in frequencies 10x to 20x higher than audible range. For normal measurements (RT, EDT, C80, G) in 1:1 scale we use 50-5000Hz, so in scale model we would like to be able to measure within 500Hz-100kHz. This can be achieved with microphones (like this GRAS 46BF), but sampling range of 192kHz in present AD converters is limiting. Also, it seems most cards are not good when used for frequencies above audible range (above 22kHz), as they are mostly aimed at (obviously) audible range. If we want to do some auralizations, we would need to measure impulse responses as high as possible frequencies (up to 200kHz), however choice of microphones is limiting us to approx. 140kHz.

I would be grateful, if someone could put RME ADI-2/4 Pro SE to test for those two upper sampling frequencies and report it's results.
Andy

 

[morillon]

instead contact the people working on sonar technology
(the absorption coefficients of the different materials are so different that this approach does not become obsolete?)

 

[andy3d]

instead contact the people working on sonar technology
(the absorption coefficients of the different materials are so different that this approach does not become obsolete?)

Before exploring sonar tech, would like to see the performance of consumer products (like RME ADI-2/4 Pro SE). We have been using for many years EMU 0404, and later Motu or RME Fireface. So maybe the newer model would be an upgrade ? But to be able to do it, we need to look at higher frequencies, and SNR there.

We use scale models not so much for reverberation time predictions (we use geometric software like Odeon for that), but to test the influence of wave effects (diffusion, difraction, echo detection, etc), which still cannot be correctly simulated in computer model for such complex geometries like concert halls (still too complicated for modern computers ). Absorption can be predicted much easier or tested in the lab. One can test diffusion also, but it's much more fun to work and design with scale models than just to sit and measure samples in the lab

 

[pkane]

Before exploring sonar tech, would like to see the performance of consumer products (like RME ADI-2/4 Pro SE). We have been using for many years EMU 0404, and later Motu or RME Fireface. So maybe the newer model would be an upgrade ? But to be able to do it, we need to look at higher frequencies, and SNR there.

We use scale models not so much for reverberation time predictions (we use geometric software like Odeon for that), but to test the influence of wave effects (diffusion, difraction, echo detection, etc), which still cannot be correctly simulated in computer model for such complex geometries like concert halls (still too complicated for modern computers ). Absorption can be predicted much easier or tested in the lab. One can test diffusion also, but it's much more fun to work and design with scale models than just to sit and measure samples in the lab

Here you go... Maybe you can explain why you might need such a wide measurement bandwidth?


(Sharp filter DA and AD for all measurements)

384k sample rate Frequency Response:

768k Frequency Response:

384k 1kHz signal 10-22.4k bandwidth:

384k 1kHz signal 10-192k bandwidth:

768k 1kHz signal 10-192k bandwidth:

768k 1khz signal 10-384k bandwidth:

 

[morillon]

microphone input?
( in this projet, wonder what sound source ultrasonore is used?)
I maintain the interest in moving towards hf active sonar specialists

 

[NTK]

Here you go... Maybe you can explain why you might need such a wide measurement bandwidth?

@andy3d explained it in post #89. He is simulating concert halls using scaled models (scales of 1:10 and 1:20). Since everything is smaller, the wavelengths of the test tones need to scale accordingly, therefore the frequencies need to scale up by 10x or 20x.

 

[morillon]

you forget the published in your modeling, right?
;-)

 

[pkane]

@andy3d explained it in post #89. He is simulating concert halls using scaled models (scales of 1:10 and 1:20). Since everything is smaller, the wavelengths of the test tones need to scale accordingly, therefore the frequencies need to scale up by 10x or 20x.

That makes sense! Missed the need for shorter wave lengths.

 

[pkane]

you forget the published in your modeling, right?
;-)

Not sure what or who you’re asking?

 

[morillon]

Not sure what or who you’re asking?

;-)

 

[andy3d]

Here you go...

Thank You very much Paul ! The noise floor of that card is higher in high frequencies but still < -110dBFs. Frequency response is also quite ok (just -0,5dB drop at 200kHz). I think we will give it and try .

 

[weme]

Acoustic scale models
E.g. Acoustic model of the new Sinfonia Varsovia Orchestra's new concert hall:

3:03 ff
"The model includes viewers: 1827 people figurines"
"During acoustic testing they were dressed in felt vests that imitated the way the audience absorbs sound."

Subtitles: English, Polish, German

Model measurements under real conditions
(An extract (last page) from the German project description BTR-10_Akustikmodell_Sinfonia_Warschau.pdf has been translated into English for you)

"In 2020, Müller-BBM carried out detailed measurements in the model of the Warsaw Concert Hall. The measurements were carried out as in a real hall:

The sound source was an omnidirectional radiating mini-dodecahedron, which was positioned at different places on the podium. Miniature measurement microphones were positioned at selected audience locations. The measurements were carried out using frequency-transformed sweep measurement signals on a model scale.

The so-called room impulse response is calculated from the received signal, which can be understood as a kind of unmistakable "acoustic fingerprint" that contains all acoustic information about the fine structure of the reverberation and from which all acoustically relevant criteria such as reverberation time or clarity can be derived. In the further evaluation algorithm, a mathematical correction was made for the influence of air absorption.

The measurement results confirmed the forecasts from the computer simulations, i.e. the expected excellent concert hall acoustics in accordance with the client's requirements. In some areas, however, the measurement results also showed undesirable acoustic artefacts, which could be attributed to unfavourable sound reflections from curved parapet surfaces. ..."

@andy3d please report on the results of your tests with the RME ADI-2/4 PRO.