REW measurement timing instability

[dima1stg]

I recently had a discussion on Audiolense forum regarding using UMIK-1, and USB mics in general, for mesurements. As I understand, AL mostly doesn't work well, or doesn't work at all, with USB mics. When I mentioned that I never had problems with REW, I got this from Bernt (AL creator): "Mitch (author of Acourate) and myself have examined a bunch of REW measurements in the past. Several takes sometimes give significantly different results with regards to timing and shape of the IR's, which is a clear sign that the problems appear there too. There's no way to know whether a REW mesurement is 100% OK or not when all of them are a bit different. It's more a matter of figuring out, based on various additional information, that this measuremen is probably good enogh, but that isn't".
Could you please comment on this?

 

[John Mulcahy]

Could you please comment on this?

Not without seeing the measurements, no. I get very consistent results with UMIK-1 and UMIK-2, but we have seen instances on the forum of people having issues which have in some cases been tracked down to long USB cables or (unrelated to UMIK) long HDMI cables.

 

[Mitchco]

Perhaps there is a bit of confusion with the OP. First, I am not the author of Acourate, that would be Dr. Ulrich Brüggemann @UliBru here on the REW forum. 2nd there is no issue with UMIK-1 or 2 using the "Measuring with file playback" method as described in John's REW Making Measurements help file.

The issue with the USB mic and a separate DAC is that there are two clocks involved. And it is not so much about clock drift, as both REW and Acourate can account for that, as it is about the variable sync times between when the sweep starts and the mic picks up on a channel by channel basis. @UliBru has a very cool "mic align" feature in Acourate in which one can compare an analog measurement mic using one clock as the input and output which is facilitated by a single audio interface versus a USB mic (with its own clock) and DAC which is another clock. Once you see it, you will understand.

 

[John Mulcahy]

See this post for some relevant data, USB mic designs are not all alike...

 

[dima1stg]

@Mitchco: my apologies for confusion.

My problem is that I don't have an analog mic and it's unlikely I will ever have it. So I'd like to keep it purely theoretical, just to have some level of understanding of a degree of errors that could be produced by using USB devices for both input and output.

When talking about clocking, is it about USB clocking or ADC-DAC sampling rate?
USB specs state that it's clock variation is <0.25%. Is this a such a big deal? Agree that error might be higher than that with analog path, but would it be higher to a such a degree that correction result would be really way off? And to a degree that AL would just plainly reject UMIK-1 and USB output combination?
Now I'm really confused, John mentioned that aside from long cables or another unrelated issues, he "gets very consistent results with UMIK-1 and UMIK-2". Doesn't this contradict to above mentioned posts? And to quote from Bernt: "Mitch and myself have examined a bunch of REW measurements in the past. Several takes sometimes give significantly different results with regards to timing and shape of the IR's, which is a clear sign that the problems appear there too. There's no way to know whether a REW mesurement is 100% OK or not when all of them are a bit different."

What I'm trying to achieve is to create correction filters for miniDSP OpenDRC 2x8 for by-amped 3-way mains and 2 or 3 subs and figure good XO frequencies and slopping. Will REW + MSO work well for me or I'll go nowhere without AL?

 

[Mitchco]

@dima1stg there still appears to be some confusion... There is nothing wrong with REW. There is no timing instability. Period. And nothing to do with sample rate.

The issue is 2 clocks (derived or otherwise) between the USB mic and audio interface. And as I said in the post above, it is not about clock drift. The issue is specifically the syncing of the start time of when the mic is recording and the start time of when interface plays back the test signal. There is variability there. And it is DAC dependent and the number of channels used. For stereo, usually not a big issue, especially if using REW's file playback capability which provides a start and stop "chirp" which removes any timing variability.

However, as mentioned above, the variability does worsen as the number of channels increase (and the DAC being used). I have seen 3 sets of back to back measurements where the timing can be off by over 3 milliseconds for the same channel. Sounds travels roughly 1ft per millisecond, so it is off by 3 feet. So if you are trying to time align a driver with another driver down to the sample, this could be a significant issue.

While the UMIK-2 mic appears to be more stable, I have seen measurements where the polarity is reversed using this mic (and compared side by side using the UMIK-1). Don't know if it is a manufacturing glitch or... With the UMM-6 I found that some mic cal files are way off...

The only real way to ensure timing integrity is to use a calibrated analog measurement mic and a single audio MCH interface like a Motu mk5 for example. This will provide consistently repeatable measurement timing results down to +-0.02ms tolerance for each channel when taking 3 sets of back to back measurements, regardless of how many channels are used. Good luck!

 

[dima1stg]

3ms is surely a lot. Thank you very much Mitch, it's clear now. Really appreciate your thorough explanation and especially your time spent on it. Cheers.