Group Delay Questions and Analysis

Status
Not open for further replies.
I'd say it would be better to include your mic cal to the correction. Exported impulses from REW don't include the mic correction, so you'd have to do that within DRC.

I didn't know that. Good point.

mic cal.jpg

This could make a difference. (despite what it says, this graph is not smoothed)
 
Last edited:
There is the moving mic technique where you collect a bunch of sweeps in succession and average them. While collecting hold the mic in your hand and move it about the approximate "sweet spot" to get a good representation (on average) of where you might listen.

Some listening furniture guides you to sit in the same place/position more than others. With a couch, the variance could be much greater than with a stand alone chair.

I am not sure how much this matters given any system that images half way decent will have a strong center. Vocalists in particular are often located dead center. So your head ends up in pretty much the same place every time if your listening for the location of the vocalist.

With classical music, and instrumental music in general, this may not be so strait forward though.
 
Last edited:
Last edited:
Lets go with 3 versions, a 2 cycle, 3 cycle and 4 cycle. One of each. For Fr curve, make a strait tilt that more or less follows the mdats I am giving you (about a 9db tilt from 20-20k, or 1db/octave).

I'd include at least one correction with a Normal template (10 cycles!) just to get a glimpse on what correcting time behaviour could do. I don't expect it to sound superb, probably fake. But it might give you an idea on what's possible.

It's just one more song to listen to. :cheers:
 
GD is based on derivative of frequency response

GD is the - derivative of the phase response with respect to frequency, -dTheta/df.
This has the correct dimensions of time.
A linear phase response, which is a constant time delay for all frequencies, gives zero GD,
since the derivative of a constant is zero.
 
Last edited:
GD is the - derivative of the phase response with respect to frequency, -dTheta/df.
This has the correct dimensions of time.
A linear phase response, which is a constant time delay for all frequencies, gives zero GD,
since the derivative of a constant is zero.

Thanks. That is why I have been trying to get my phase to be flatter on the FAST box. 🙂
 
I wanted to go back to something mentioned before. That is, looking at GD in terms of cycles instead of absolute time.

A full cycle for these frequencies is as follows:

50hz - 20ms
40hz - 25ms
30hz - 33ms

Some say 1 cycle or even 2 is inaudible. Having heard my system with 60ms, 25ms, and 15ms GD at 40hz, I can say this. 60ms to 25ms was definitely audible. 25ms to 15ms was noticeable, but barely.

Based on my listening, like many things, the idea of diminishing returns is at play. Somewhere around 2/3 or 1/2 cycle seems a good goal.

Additionally, some frequency regions are more important than others. What happens below 40hz is less important than what happens above it, IMO.
 
Last edited:
Status
Not open for further replies.