Phase trace appears to be showing false information

[(Nat)]

Hi, I'm trying to measure the TF of a comb filter as a demonstration for someone, so I set up a simple test system into Logic and back into REW via the ProTools Audio Bridge. I got the result below, which all looked good until noticed that it was wrapping at 90 degrees instead of 180, so it's telling me that my first cancellation is at 90 and first peak at 180. That's wrong (right!?).

Signal flow is:
Output from REW is protools audio bridge output 1
Input to REW from Logic is protools audio bridge input 3. The numbers are arbitrary. I have signal in and out with no feedback loops.
In logic, I have the output from REW routed into two separate mono channels (both at -6dB to compensate), one with a 48 sample delay on it (I'm working at 48kHz (I tried 44.1 too to see if that fixed the issue but it made no difference)
The two channels are then routed into the same mono output back into REW.

No amount of playing around with polarity, IR settings, time references, loopbacks, wrap/unwrap phase (it doesn't actually let me unwrap) or sample rates has given me any different results (except for worse ones).

Firstly, why is it wrapping at 90, and how can I tell it to stop?!

Secondly, am I wrong, or is REW right and the first peak of a comb filter is at a 180 degree phase shift?

Or maybe I'm just being stupid. I don't know. Even Chat GPT couldn't figure out what was up!

Any help would be much appreciated, ta.

**Edit: I'm now seriously doubting my understanding of comb filters and TF measurements. Is there a good reason for it displaying 90 degrees instead of 180 at the first dip? It seems counter-intuitive.

 

[John Mulcahy]

Perhaps you are confusing the phase response of the overall system (i.e. the sum of the two paths, so an IR with two peaks separated by the delay time) with the phase difference between the direct and delayed paths at the nulls and peaks.

 

[(Nat)]

Perhaps you are confusing the phase response of the overall system (i.e. the sum of the two paths, so an IR with two peaks separated by the delay time) with the phase difference between the direct and delayed paths at the nulls and peaks.

Hi John,

Thanks for replying so fast. I think this is the case as I just ran a measurement through the undelayed channel to find T=0, and then ran a second measurement with that time reference through the delayed channel and the phase response (below) showed what I was expecting.

That makes sense now, though I need to think through the maths to make sure I understand how they sum to apparently make half the phase shift.

best

 

[John Mulcahy]

For clarity, individual and combined phase responses on log and linear frequency axes:

 

[(Nat)]

That's helpful. Is the sum simply an average of the direct and delayed?