Phase in loopback calibration

[smantila]

Hi, I'm using REW 5,19 linux with Asus 970 motherboard internal soundcard and loopback wire connected headphone / line-out to line-in.

When I do soundcard loopback calibration phase graph looks odd:

Phase goes from -177 to 80 degrees (20-20000Hz). Is this normal? Is this unlinear phase affect of my PC maybe delay or something like that and will this cause issues when measuring with setup like this?

 

[John Mulcahy]

It is showing that the overall path is inverting. You can adjust for it by selecting "Invert" in the Input options of the Soundcard preferences.

 

[smantila]

Ok, thank you. Now phase graph looks better:

So I would assume that Asus line-out or line-in inverts the signal. There is still difference from 17 to -140 would this be considered normal?

Ps. Thans for the great software!

 

[John Mulcahy]

Yes, that's normal. The response is a little bit noisy, since it is only down by 0.5 dB at 10 Hz it would be fine to continue without using the cal file, to avoid adding that bit of noise into your measurements.

 

[smantila]

Ok yes. I ran the calibration without thinking if I actually need it.

But now I'm more confused. I ran loopback measure without calibration (no mic, no soundcard) and phase response is almost flat:

And then I ran the calibration and with that calibration I measured loopback response again:


I'm most likely doing something wrong here. But it would be fairly easy for me to just make calibration file manually as phase is allmost flat (not many datapoints needed).

Why I'm so interested for phase data is that. I plan to do impedance and phase measurements for crossover design.

Attached is also the calibration measurement mdat file.

 

[smantila]

This seems to have something to do with minimum phase. If I click the create minimum phase for measurement that was done without calibration it creates the phase plot as shown on later measurement. I need to go do some reading of this minimum phase

 

[John Mulcahy]

The phase response is dependent on where t=0 is assigned for the impulse response. If you make loopback measurements without a timing reference t=0 is placed at the peak of the IR. That is where you are seeing flatter phase at high frequencies. When the soundcard calibration is made REW chooses the t=0 point based on an estimate of any time delay in the measurement, produced by cross correlation of the measurement with a minimum phase equivalent. That produces more phase shift at high frequencies as the t=0 point is before the peak (by about 1 sample usually). How correct that is can depend on the type of reconstruction filter used in the D/A conversion and the type of antialias filter used in the A/D conversion. It isn't straightforward to determine the "true" t=0 point and the corresponding phase plot, other than to say if the phase is rising at high frequencies the t=0 point is too late.

 

[Mick W]

Mild thread dig here but I had the same problem as smantila the OP.
Every setup tutorial I read or watched glossed over a few things, one of them was how to deal with those whacky looking phase shift graphs that the OP posted. My graphs are the same on a Presonus Audiobox USB.
Reading John Mulcahy's comments prompted me to look in the Analysis tab of the preferences window.
There is a tick box to "Adjust clock with loopback"
So, apart from the oft mentioned method of plugging a cable from your interface's audio output to one of it's inputs to calibrate the internal soundcard's frequency response there is another "output to input" loop that can be done to keep the phase in sync ?

 

[John Mulcahy]

Adjusting clock rate using a loopback connection is only relevant when the input and output are on different devices and so do not share a clock. There is nothing whacky about the phase plots posted above.

 

[Mick W]

Adjusting clock rate using a loopback connection is only relevant when the input and output are on different devices and so do not share a clock. There is nothing whacky about the phase plots posted above.

Thank you John, I should have realized that. And I'm unfazed about the plots now