Why does my measured phase plot slope downward?

mojozoom

New Member
Thread Starter
Joined
Jun 10, 2017
Messages
7
Here's a question I'm trying to get a handle on: What makes the phase curve of the mids in my car slope continuously downward when measured at the listening position?

I dug up an old test I did near field (maybe 3" away from the cone) with the car door open and the phase curve was very flat across the range, with an upward slope occurring below about 80 Hz, which I assumed was due to the high pass filter. But other tests I have done from the driver's seat show the phase sloping down continuously over the passband of the driver.

Maybe I didn't have the TA turned on at the time I did the near field test, and the listening position phase slope is a result of the time alignment being applied?

Or maybe the phase slope is due to the car itself? Perhaps the small volume of the car allows low frequencies to be sensed by the mic earlier because they are pressurizing the room, and the higher frequencies aren't?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Time delay is the main contributor - you can have REW estimate and remove time delays using the controls in the SPL and phase graph.
 

mojozoom

New Member
Thread Starter
Joined
Jun 10, 2017
Messages
7
What causes the time delay? Is it something that is truly occurring or something we see in the measurement due to the way we take measurements?

If I took a series of measurements on a driver starting up close and kept increasing the distance would I see the phase curve start flat and slope steeper and steeper the further away the mic was located?

Thanks in advance for helping me understand this - sometimes I feel like there's just those critical little pieces of comprehension that get lost in shuffle of "doing" and not knowing exactly why the results are as they are.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
There are buffering delays in the operating system, processing delays in pieces of equipment and transit delays in the sound travelling from the source to the mic. Buffering delays can be highly variable from measurement to measurement. The overall result is that measurement software has no way to distinguish the various delays within a measurement and know the 'true' delay that should be included in the phase. Any time delay affects the phase response, shifting it by the number of degrees that time delay corresponds to at each frequency - for example, a 1 ms delay is a 36 degree shift at 100 Hz, a 72 degree shift at 200 Hz, a 360 degree shift at 1 kHz and a 3,600 degree shift at 10 kHz. There are a few options for the software to handle the delays:
  • Pick some feature of the measurement result and designate that as the reference for time, i.e. make it the point where time is considered to be zero. Most common are the peak of the impulse response, or the first point the impulse response exceeds 10% of its peak value. The effect is to throw away delays, but the chosen reference point is often not the 'true' start of the response, so includes some time offset which affects phase.
  • Make some assumptions about the nature of the system, usually that it is minimum phase, and use that assumption to estimate the delay in the measurement and remove it. REW does that in the Estimate IR Delay function. It is better than picking a single reference, but is not fully accurate as determining the minimum phase response requires response information that extends above and below the typical bandwidth of measurements.
  • Use a timing reference of some sort, such as a loopback connection on another soundcard channel or a signal from another source that remains a constant distance from the mic (REW's acoustic timing reference feature). That should leave only constant delays, since the variable buffering delay is removed, and measurements with a timing reference are directly comparable for delay - in your example of measuring at different distances, that would be apparent in measurements with a timing reference but not in those that pick a feature or remove the delay. There is still a delay with a timing reference, but at least it is fixed. That still leaves you with the task of deciding what that fixed delay is and removing it from measurements (using the t=0 offset controls on the IR graph) if you want to see the phase as it was at the source. That may require some knowledge of the nature of the system you are measuring to know what the correct phase or delay is at some part of the response. In most full range measurements there shouldn't be any group delay at high frequencies, so looking at the group delay plot can reveal whether there is some constant delay offset that should removed (group delay is determined from the slope of the phase response). Alternatively, the wavelet spectrogram peak energy trace provides much the same information.
 
Top Bottom