I've performed some Measurements on SPL Volume2 (Model 2602) volume control

[Rja4000]

Why didn't you do this at 384 kHz sample rate? Fully linear then...

Well, it seemed Virtins is limited to 192kHz.
I will try it again.
Also, I limited the sampling frequency because I use a FFR signal (similar to what HP-Works uses) and I couldn't find (or recreate) one above 96kHz.
FFR is very handy, since you immediately see the impact of any change.

BTW, I assume you use the mono trick at the input to get 3 dB more SNR? Are you aware you can do the same at the output? You can use a simple split cable, two XLR female to one XLR male. Let the generator send out the same signal on both channels. Internal resistors in the ADI cause an unchanged output level and THD, but 2.5 dB less noise. I use that quite often.

Thanks for the hint.
Is that any better than the balanced phone output I use sometimes ?
Well, at least, I could then use any output range, not only +13dBu and +28dBu.
Usefull to get better performance @+4Vrms, as an example.
I'll give it a try.

One of the problem I'm facing is the following: I have 2 ADI-2 Pro fs, but I still don't know how to have them working at different sampling frequencies.
Due to a limitation in Virtins MI, I have to use ASIO4ALL for one of them.
This is because both interfaces come under the same ASIO driver and MI doesn't make the difference and consider them as one single interface.
But even with ASIO4ALL, it seems most of the time to force sampling frequency on both devices to the same.
I have to figure out a way to automatically set the DAC to, say, 44.1kHz and the ADC to 192kHz.

 

[MC_RME]

That's the way it is, and the only way around it is by using two separate systems. I personally have all my test signals on iPad and iPhone and use Neutron to play them at any sample rate. Analysis is then done on the computer with HpW, APFlex and MI. This has become even more useful as the latest Neutron allows to convert PCM in real-time into DSD64 and DSD128, so I can use my existing PCM samples (that are all short but loop perfectly) even to test DSD.

And Virtins is not limited to 192 kHz, it goes up to 768.

 

[Rja4000]

Thanks.

I don't know why I thought MI was limited to 192kHz. Sorry for saying that. I'll try again.

As for the 2 systems: what I do is automate tests with MI through a .Net app, storing results in a SQL database. I can of course use 2 systems, but that would make it much more complex to program.
I will give it a try with some VM on the same PC.

I had a huge amount of work recently, mostly abroad, so no time for experimenting. I'll use some free time during this holiday period to tune this procedure.

I want to revisit the graphs. They don't look readable enough here for my taste.
I have a Telerik library available and that could provide more flexibility. I'll have to rewrite plot code, though.

I also would like to include some Amir-s-dasboard-like view - which means rebuilding the FFT graph myself. Not too hard to achieve, since MI allows to save all FFT data, but, still, time consuming...
That's the easiest way to combine several separate FFT measurements on same plot anyway.
That's also probably consuming less HDD space than saving each value plot as image. (EDIT: It's not.)

 

[MC_RME]

I don't know why I thought MI was limited to 192kHz. Sorry for saying that.

You might have had the ADI in Multichannel mode.

As for the 2 systems: what I do is automate tests with MI through a .Net app, storing results in a SQL database. I can of course use 2 systems, but that would make it much more complex to program. I will give it a try with some VM on the same PC.

Can't help with that, sorry. But the limitation of having a unit that supports only one sample rate is the same everywhere unless you buy into dedicated audio measurement systems (which are at least dual inside). And ASIO allows only one sample rate. Of course, my workaround, which works tremendously well in most cases, is a fully manual one - no way to automate anything.

 

[Rja4000]

You might have had the ADI in Multichannel mode.

Indeed !

By the way, would you know where I could find FFR test signal for 192kHz response ?

 

[MC_RME]

Frequency response? In MI...just record/save/capture/sample the generator output, noise or multitone. But that would be a question for the MI forum, it seems.

 

[Rja4000]

Frequency response? In MI...just record/save/capture/sample the generator output, noise or multitone. But that would be a question for the MI forum, it seems.

I'm looking for something like in this link (but for 192kHz - 512k FFT window size)
https://www.beis.de/Elektronik/FreqResp/InstFreqRespMeasure.html
That's to generate a sweep signal allowing almost instant frequency response measurement.
Like HP-Works is using for its Fast Frequency Response measurements.

 

[MC_RME]

I am not aware of any software that has this method implemented other than HpW. You can use the files provided by Beis and make the MI FFT look like an FR (IMHO MI can do that, but I am not a specialist for MI, mainly using the Oscilloscope of the package).

 

[Rja4000]

I am not aware of any software that has this method implemented other than HpW. You can use the files provided by Beis and make the MI FFT look like an FR (IMHO MI can do that, but I am not a specialist for MI, mainly using the Oscilloscope of the package).

That's what I do.
I use Beis' 96kHz file with MI and you see the result in the Frequency response plot.
But they don't publish a 192kHz file, unfortunately.
I asked them, but they never answered.

Thanks for your time, anyway.
And Merry XMas

 

[LTig]

I'm looking for something like in this link (but for 192kHz - 512k FFT window size)
https://www.beis.de/Elektronik/FreqResp/InstFreqRespMeasure.html
That's to generate a sweep signal allowing almost instant frequency response measurement.
Like HP-Works is using for its Fast Frequency Response measurements.

Interesting Link. But 512k is probably too much - you need 512×1024 added sinus signals which means that the singular sinus has very low amplitude leading to problematic SNR.

 

[LTig]

That's what I do.
I use Beis' 96kHz file with MI and you see the result in the Frequency response plot.
But they don't publish a 192kHz file, unfortunately.
I asked them, but they never answered.

I wrote a program for creating test files in WAV format (Linux only for the moment). I had a short look into the link and think I can add support for such a signal and provide them for you. Have to find out how to dermine the phase of the individual sinus signals.

 

[LTig]

I think is not balanced indeed. But I'm not an electronic expert.

IMV it makes more sense to convert the incoming balanced signal to unbalanced, thereby getting rid of all unwanted common mode signals.

 

[Rja4000]

Have to find out how to dermine the phase of the individual sinus signals

That's actually where the problem usually is.
Same for the multi tone test.

It's all about crest factor, alignment to FFT bin, according to sampling frequency and desired (low frequency) resolution.

For 32 tones, you'll need { >4 x <Sampling frequency>} as the FFT window size if you want to be able to resolve the low frequency on the graph (ie see some dip between peaks at lowest frequencies)
For this FFR, this is not needed to go that far, probably.

As you said, it's a matter of SNR, so crest factor is the critical topic.

I've found some articles for crest factor optimization methods (by computing phases), but they are mainly for linearly-distributed frequencies, not for log distributed frequencies as we like to see them for 32 tones.
They may be applicable for FFR though, since FFR seems to look for each FFT bin to be present, and the bins are linearly distributed.

Random phases method, although better than same phase for all, doesn't look like optimal, from what I read.

 

[Rja4000]

I can add support for such a signal and provide them for you.

That would be awesome !

 

[MC_RME]

Interesting Link. But 512k is probably too much - you need 512×1024 added sinus signals which means that the singular sinus has very low amplitude leading to problematic SNR.

Agree. I usually use below 65k, even at 192 kHz. Typical stepped FR measurements use 200 points for the whole audio range, so using 4096 bins in a FFT is already lightyears ahead.

Edit: As much as I like this kind of real-time measurement, one should be aware of its limitations: generator and analyzer must be in coherent sync, and have the same sample rate. Using the FFR signal on an iPad with an unsynced DAC will produce deviations/errors in the analyzer. And aliasing can produce strange deviations in the stopband region. Finally, as mentioned above, crest factor prevents this method to do CMRR and crosstalk sweeps as the signal drowns into the noise floor.

 

[SIY]

Edit: As much as I like this kind of real-time measurement, one should be aware of its limitations: generator and analyzer must be in coherent sync, and have the same sample rate. Using the FFR signal on an iPad with an unsynced DAC will produce deviations/errors in the analyzer.

There may be a workaround if you use a log f sweep (Farina chirp). In their transfer function measurement, AP uses a correlation function to take out the delay, and I think you can do this (albeit in a two step process) with the Virtins.

 

[Rja4000]

I think the whole topic is that FFT uses linearly distributed frequency bins and we all want to see Frequency-related measurements on a log graph.

So if you lower the number of FFT bins, you loose low frequency resolution. If you increase it, you loose SNR, especially if you want a signal to combine all frequencies.

If you measure FR with a step method, you'll distribute your steps logarithmicaly.
That may be giving more accurate data, but that's much slower, for sure.

 

[Rja4000]

I think you can do this (albeit in a two step process) with the Virtins.

How ?

 

[Rja4000]

By the way, googling a bit, I found this, from our local university
http://www.montefiore.ulg.ac.be/~stan/ArticleJAES.pdf
Not directly related, though.

 

[Rja4000]

generator and analyzer must be in coherent sync,

Indeed. If not, we see strange artifacts.
Like here
https://www.audiosciencereview.com/forum/index.php?attachments/minidsp-dirac-ffr-profile2-jpg.33545/