Problem with calculation of phase response (time zero)

[pawelpl]

Let me explain how do i calculate phase response.
First step is to playback sine sweep signal and record response. This step i perform in Audacity software. Then i export stimulus and recorded response to Matlab software.
The second step is to found delay between stimulus and response, using cross correlation, and remove this delay from response. So, after this step i should have two time zero signals.
The third step is to calculate Frequency/Phase response by H = FFT(Recorded response) / FFT(Stimulus).

The problem is that calculating delay by cross correlation (step two) not always return correct value and it is affect on phase response.
Can somebody suggest me how to properly calculate time zero to get correct phase response ?

[JohnPM]

In the absence of prior knowledge of the nature of the system (e.g. physical distances, air temperature, electrical and processing delays) it can be very difficult to come up with a "true" zero time for the measurement. For full range measurements cross-correlation with a minimum phase version of the response gives a pretty good figure, that is the method REW uses in its "Estimate IR Delay" function.

[pawelpl]

John, thanks for advice, i've did like you describe and it works in my case (passive RLC circuits).