Propper use of Clock adjustment

[JoachimStrobel]

I am probably getting carried away a bit here - but for the sake of it:
I discovered the task "adjust clock" in the impulse window and played with it.
I played a 48 Khz sweep and looked at the impulse and then changed the clock by - 5 ppm to make the wavelet a bit more symmetric. This does change the phase a little bit.

The I did the same with a 192 Khz sweep and applied a -5ppm offset. There is of course more conrol with the 192 Khz version which allows a bit finer tweaking. The result a slightly changed phase again.

I could not find any help file on this feature: What is the correct method here? Is it worh recording everything with 192 kHz and finetuning the impulse for a later use in rePhase or else?
Am I overcorrecting if I set "clock-adjust with timing signal" in preferences to "on" too?
(These are full speaker responses recorded at the LP with an Umik1 from a Windows PC´s soundcard output)


And, btw, this is the 48 Khz pulse played as file through Roon and a Raspi4 with a Hifiberry HD2DAC and Volumio ... time to think.

 

[John Mulcahy]

The UMIK-1 only works at 48 kHz. If "Adjust clock with acoustic ref" is selected in the Analysis preferences REW will automatically correct any clock rate difference between the source and the mic.

 

[JoachimStrobel]

The UMIK-1 only works at 48 kHz. If "Adjust clock with acoustic ref" is selected in the Analysis preferences REW will automatically correct any clock rate difference between the source and the mic.

Yes, but finetuning the clock will still create a more symmetric impulse response regardless if the acwar switch is on or off- is that something to aim for? And Oversampling the UMIK will then just create an illusion of a smoother adjustment process.

 

[John Mulcahy]

No, you have no reference to know whether the adjustment you are making is correct. If the automatic adjustment has already been applied any change will be incorrect.

 

[JoachimStrobel]

I confess that I probably do not understand this fully.

I assume that we correlate the input sweep to the recorded one. Time alignment will tell us about the time delay between the two signals. Will Rew then still finetune the match? Or take that delay as it is and stretch the time axes with the final time signal? If I do not have a time signal, how does Rew guess the correct time alignment?
My understanding would be, that one could change the needed fine adjustment time a bit and see which cross correlation function gives the best results and that is it. On the other hand, the room gets in the way. The results would be a kind of symmetric zero-phase wavelet that we see at T0, of course starting a bit before T0. Or is the assumption that the impulse is shown by a symmetric zero phase wavelet wrong? As you suggested in your comment, that one does no know how it would look like and hence cannot try to recreate it by fiddling with the time axes. A solution could be to see the correlation coefficient while changing the alignment time to decide what is best. Or am I on the wrong track and that kind auf autocorrelation would destroy what we want to measure?

 

[John Mulcahy]

Clock adjustment uses the timing signals that are placed at the start and end of the sweep. REW knows how many samples there should be between the timing signals and if there are more or fewer it knows the clocks were different and can resample the data accordingly.

is the assumption that the impulse is shown by a symmetric zero phase wavelet wrong?

It is wrong, the IR is only symmetric for a system that is flat to Nyquist. Acoustic responses are causal, the response is asymmetric and mainly to the right of the peak.