Home Theater Forum and Systems banner

Generate minimum phase incorrect?

2K views 2 replies 2 participants last post by  chChrLam 
#1 · (Edited)
My tests of the generate minimum phase(MP) button is giving some strange results

The input and output device used is a bit-perfect miniStreamer with an SPDIF cable looping the input and output.
It provides as expected a perfectly flat frequency response with zero phase.

I performed 5 sweeps centered on 640 Hz with start and stops X octaves below and above 640 Hz (X is 1-5)

20 - 20480 MP 130 to -50 degrees
40 - 10240 MP 180 to -120 degrees
80 - 5120 MP gets worse and wraps from here on
160 - 2560
320 - 1280

My issue here is that I was about to use the generated MPs in a crossover designer application(ACD), but it looks to me like generate MP is not working correctly(*)

How does generate MP work? Does it just look at only the frequency response and do a hilbert bode transform(HBT)?
AFAIK the HBT will give good results only if you generate tails for the frequency response from 1 Hz all the way to 100 KHz.
 

Attachments

See less See more
2
#2 ·
The minimum phase response is calculated using the real cepstrum, but it is equivalent to using a Hilbert transform. That's not the main source of your problems though. The MP response is calculated on the windowed response, the shape and extent of the windows have an influence. To minimise that preserve the full IR (see Analysis settings) and expand the windows to include the full IR and use Rectangular window. That would typically keep the excess phase extremely low down to pretty much DC for your digital loopback, but there will still be increasing errors as Nyquist is approached. Those can be reduced over the range of interest by using a suitably high sample rate. The biggest problem is using a sweep that doesn't include the full bandwidth, however. If you sweep over a limited bandwidth your system is effectively being combined with a steep-sided bandpass filter which will have a very large influence on the results.
 
#3 ·
Thanks John for the answer.

I think I will just stick with my relative phase with timing reference measurements for designing my digital crossover. Minimum phase with acoustic centre offsets is nice to have and allows off-axis simulation but the relative phase with timing reference works perfectly fine for on-axis simulation. My baffle cannot be changed (Veneered 3 way box will be fully activated) anyway. And also I like to eliminate all sources of error.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top