Minimum phase and excess phase problem

[carlo chiarelli]

Hi John,

sorry to bother you but i have some problems on phase calculation.
When i load a calibration curve that include also phase, phase of measure is correctly compensated, but also minimum phase calculation is affected.
Obviously, due to this, also excess phase is wrong.
To check i create a file with an ALL PASS filter at 1KHz and i save it as impulse response and also as TXT.
I load impulse as measure and TXT as calibration file.
According to theory resulting file is a Dirac ( flat phase, flat amplitude).
But if i calculate minimum phase version , it don't take account of phase compensation but only of amplitude compensation.

Something wrong on my thinking?

PS i have some difficult also with wav filter response: it's only one sample before impulse and too short after: if i reload it give some problem on REW for low frequency convolution ( A*B )

 

[John Mulcahy]

Regarding the filter impulse response, there are two changes in the next beta to help with that:

• Exported filter impulse responses now start at sample zero rather than sample 2 (so phase is correct when using the option to set t=0 at first sample for imported impulse responses)
• Imported impulse responses are padded with additional zero samples before and after the response if the peak is near the start of the response (without changing the t=0 location)

Cal file phase corrections should not be applied to the min phase response, I have fixed that for the next beta.

The situation is a little more complicated of course, since the minimum phase response is calculated from the impulse response, which doesn't include the effect of cal files. Ideally the min phase response would be calculated from an impulse response that included cal file effects and would be recalculated if the cal file changed, but including cal files in impulse responses has more extensive implications that are on the list to consider at some point in the future.

 

[carlo chiarelli]

Ho John,

Many thanks for clear explanation!
For second issue i solve with a A/B with B impulse response on calibration.
However a Hilbert transform inside REW world be very usefull.
Many thanks for REW.

 

[entrelaslineas]

- Привет! Но почему фазовые поправки Cal file не должны применяться к минимальной фазовой характеристике?
Я хочу построить минимальную фазу на идеально плоской частотной характеристике, но вместо плоской линии получаю кривую линию.
Например, посмотрите на картинку. Верхняя линия-это частотная характеристика. Ниже приведен график плоской фазы (красный) и минимальная фаза (серый). Почему минимальная фаза является такой кривой?
Спасибо! Я прошу прощения за свой английский...

 

[John Mulcahy]

- Привет! Но почему фазовые поправки Cal file не должны применяться к минимальной фазовой характеристике?
Я хочу построить минимальную фазу на идеально плоской частотной характеристике, но вместо плоской линии получаю кривую линию.
Например, посмотрите на картинку. Верхняя линия-это частотная характеристика. Ниже приведен график плоской фазы (красный) и минимальная фаза (серый). Почему минимальная фаза является такой кривой?
Спасибо! Я прошу прощения за свой английский...

You have limited the bandwidth of your measurement so it is missing the lowest frequencies. Measure from 0 to half the sample rate. Here is a perfect impulse.

 

[entrelaslineas]

Вы ограничили полосу пропускания вашего измерения, поэтому он пропускает самые низкие частоты. Измерьте от 0 до половины частоты дискретизации. Here is a perfect impulse.

But, with all due respect, I think that's not the point. I'm attaching a new dimension. It has a full band from 0 to 24000 Hz. But it can be seen that the minimum phase is still not minimal. The important thing is that my sound card in the laptop does not have a flat phase characteristic. I calibrate it to get a linear phase when measuring, but instead I can't plot the minimum phase.

 

[John Mulcahy]

I'm not sure what your point is. That is the minimum phase response of that soundcard. The phase deviations are a direct consequence of the magnitude deviations in the soundcard's response. Why do you care what the minimum phase response is? For measurement the calibration file removes the card's magnitude and phase deviations so they don't influence the measured results.

 

[entrelaslineas]

I think I'm beginning to understand... I thought the minimum phase graph was calculated from the frequency response graph. But, if you confirm, now I understand that it is calculated from the impulse response. Right? I apologize if these questions are stupid.

 

[John Mulcahy]

Yes, the minimum phase response is calculated from the impulse response. Cal files are used when generating the graphs of frequency and phase response but their effect is not included in the measured impulse response.

 

[entrelaslineas]

Thanks! You've been very helpful.