ADC behavour and phase shifting ???

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

zamproject

Well-known member
Joined
May 11, 2010
Messages
1,502
Hello

I just purchase a motu PCIe card to update my motu PCI ring (24io), in short I have to replace my G5 which start to have some problem with a mac pro (on his way)
I have a good trade for a 2408mk3 coming with the PCIe card so I take it too

Now the point:
I hook up everything to my analogue setup (28 chn desk), and decide to compare 24io to 2408 when printing my mix master 2T
I split at patchbay and send master out to 2in at both converter.
Then I compare difference in audio files at daw with phase flip for on, instantly It was wrong !!! cancellation was a disaster (files are sample accurate...). Quick A-B test don't reveal drastic difference as the phase cancellation test might suppose ...
So I perform loop back directly for both converter, phase cancellation is excellent from source (by tweaking 0.01 or 0.02 dB) for each converter.
Then I perform two loopback 24io to 24io and 24io(same out as before, same cable) to 2408 and then I test by phase cancellation... disaster !!! for this last test I use 5Hz-20kHz 30 sec sweep and I notice the result of phase cancellation test is a perfect cancellation at low freq and NO cancellation at 20kHz. The cancellation "error" is linear (in log freq scale)
My conclusion is that one of the two converter continuously shift the phase from 0° to 180° in the 20Hz/20kHz freq range.

Is that due to how the AD chips and digital filter is handled ?
I just try to search and have some read about this, but I'm not expert in digital signal...
Can this be relater to the fact that one ADC have 128x oversampling and the other 64x

Any input and science (basics) welcome  :)

Best
Zam
 
The two different ADCs will have different throughput delays so you can never hope for a cancelation with a polarity flip.  This delay will also not likely be a specific number of samples, as you say due to filtering on on chip processing.

On another note a polarity flip test is not useful in this case.  Even if you did get good cancelation and a residual, what is the residual from?  Converter A, B or a mixture of both?

 
  The problem you are describing (linear phase shift with freq) sounds very much like a delay. 1 sample of a 44.1kS/s is 23µs.
That's    0.16º @ 20Hz          1.6º @200Hz              16º @ 2kHz      160º@ 20kHz. Of phase error , given 1 sample delay at 44.1kS/s.
That's -108dB@ 20Hz      -68dB @200Hz      -27dB @ 2kHz      -6dB@ 20kHz. Of cancellation, given 1 sample delay at 44.1kS/s.

  Try with smaart, you should detect the delay between them and then it will tell you the freq response (should be pretty flat, except the ends) and the phase difference between both, so now you have the timing difference and the phase error. The time difference should be easy enough to fix, as each channel should have a time correction trim. The remaining phase error would be harder to deal with if you still want to correct it, but will be much smaller. That error would be due to the filtering and analog path.

ruairioflaherty said:
On another note a polarity flip test is not useful in this case.  Even if you did get good cancelation and a residual, what is the residual from?  Converter A, B or a mixture of both?

  For most applications it doesn't really matter as long as they are consistent with each other, unless some hype around phase way over the quality of the converters, or measuring absolute phase values with a known uncertainty. For recording music I wouldn't bother too much, given consistency between channels and not huge phase errors because the filtering. That last part can be compared with a loopback compared with a directly digital signal, but it will have the errors of the ADC and the DAC. There might be some workarounds with multiple converters and a lot of math but also you can compare it with a much better DAC and discriminate the error sources.

JS
 
Thanks to all of you

I re open the project to give the asked value and respond to questions
BUT i'm little more messy now... the first re-recoded test I do yesterday (two AD device receiving same first DA ) give me no signal over -80dBfs in null test @ the whole freq range today !!!
I mean the null test is between loopback AtoA audio file and loopback AtoB audiofile

Is that possible that I have a clock issue yesterday, or a "sample accuracy" DAW compensation that mess with the two converter ? as Joaquins calculation look like what I have (but not a full sample as I can't null when moving 1 sample.

I will redo thing and check what happen if I can reproduce it.
I'll update

Best
Zam
 
I've heard reports that certain newer Motu interfaces will have varying latency for no apparent reason,  this latency changes on each power up, even when all settings remain the same.

Motu are aware of the issue but I don't think there is a solution.
 
ruairioflaherty said:
I've heard reports that certain newer Motu interfaces will have varying latency for no apparent reason,  this latency changes on each power up, even when all settings remain the same.

Motu are aware of the issue but I don't think there is a solution.

Hi
I'm using old PCI Motu not new AVB, but maybe it's the same king of problem (if there is one)
The point is just to know, to avoid mistake like multi-mic drum for example send to different interface (in OTB mix situation)
Best
Zam
 
Ok...

I made a mistake, -80dBfs null test is comparing convertor A loopback accros convertor B loopback (no cross loopback)
So my first post statement still correct and there is no inconsistency with clock or micro delay/filtering

I just take more time to understand and perform more test, I think I get it, I won't go further as it will be endless fore some details that I finally don't care

As stated I won't be able to test original vs loopback without specific tools.

I have a sub sample delay not the same with the two converters... that's it, and somehow the DAW (or the DSP) compensate it fine for each converter.

Thank you all  :)

Best
Zam

 
I know this is old but...
I have several MOTU's using pci (hd192 x 2 & 2408 mklll-which I use solely to connect other adda) i record live drums all the time and my phase is never through the converters but through the mics and poor placement. Once adjusted it works perfectly. I also switched and used apogee as the master and that helped.
 

Latest posts

Back
Top