Phase difference from previous REW version in offline measurement

yas

Registered
Thread Starter
Joined
Apr 22, 2020
Messages
2
Hello.

I have WAV files that was previously measured offline.
There is measurement data imported at that time with Import-> Sweep Recordings. (20200725_phase_issue.mdat > L-1, L-2)
In that measurement data, the phase rotates around 2kHz, and the crossover of my speaker is 2kHz, so I think it is the correct result.
Since the measurement data was created on July 25, 2020, his version of REW used is V5.20 beta 56 or earlier. (I don't remember the exact version I used at that time)
I use V5.20 RC7 currently, but importing the same offline data produces different phase results than before.
The phase rotation around 2kHz disappears, and the phase does not fluctuate from around -180 degrees after 3kHz. (20200725_phase_issue.mdat > L-1 by V5.20 RC7, L-2 by V5.20 RC7)
The calculation result of Delay is also different.

I created filters that corrected phases to the minimum phase from these measurements by rePhase, convolved them, and compared their sounds.
The filter created based on the phase generated by V5.20 RC7 clearly makes the high frequency sound worse.

In particular, the effect to string instruments of classical sources is remarkable, the tone is extremely sharp, suggestiveness of sound and smoothness are lost, and the sound becomes very false-sounding.
With the filter created from data measured on July 25, 2020, the sound of the strings instrument is very natural and has no false-sounding.

Aside from the final sound quality results, why does the REW V5.20 RC7 produce a different phase than previous versions?
Which phase result is correct?
Is it possible to download and try V5.20 beta 56 or earlier?

Attach the mdat file that imported measurements file and each files used for offline measurements.
- mdat file: 20200725_phase_issue.mdat
- offline measuremts files: 4MMeasSweep_20_to_20000_48k_PCM16_L_refL_20200725_1939_cut.wav (L-1), 4MMeasSweep_20_to_20000_48k_PCM16_L_refL_20200725_1950_cut.wav (L-2)
- sweep stimulus file: 4MMeasSweep_20_to_20000_48k_PCM16_L_refL.wav

41053
41054

41056
41055
 

Attachments

  • 20200725_phase_issue.mdat
    3.5 MB · Views: 6
  • offline measuremts files.zip
    14.3 MB · Views: 3
  • sweep stimulus file.zip
    8.1 MB · Views: 3

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Previous versions are on the History tab for each OS build, e.g. MacOS previous versions. The problem with the files is the sweep length, prior to V5.20 RC6 4M sweep files could not be imported without truncation - a message is shown saying the import was truncated at 90.1 s, losing the end of the stimulus and the responses. V5.20 RC6 onwards can import 4M sweep files and shows the correct response.
 

yas

Registered
Thread Starter
Joined
Apr 22, 2020
Messages
2
I installed and tried the previous version, but I couldn't get the result of the phase rotating at the same 2kHz as before.
Earlier versions have the same results as V5.20 RC7 too.
It's very strange, but I think I was doing something different at the time. I can't remember that, so I won't think about it anymore. I think the current measurement result is the correct phase.
Thank you.

If you know the following, please let me know.
Currently I measure directly with REW using a USB audio interface. The result is that the phase rotates around 3kHz.
Generate Excess phase by REW, and the rotation around 3kHz is linearized by Filters Linearization of rePhase.
However, if you follow the general procedure for phase correction using such REW and rePhase, the high frequencies will sound very unnatural, as I mentioned earlier.
Using the same measurement data, the timing offset function of REW is used to make fine adjustments so that the frequency at which the phase rotates is the same as the crossover (2kHz) on the speaker specifications, and then generate excess phase.
If I linearize that excess phase with rePhase and use it as a correction file, the high frequencies will sound very natural.

The result of the correction by linearization of these two different approaches is the minimum phase, and there is almost no difference in the characteristics on the graph.
However, there is a clear difference in sound. I'm not familiar with phase, so please let me know why this difference is made.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Sorry, can't help with your phase question, but the frequency at which phase passes through 180/-180 doesn't have any special significance.
 
  • Like
Reactions: yas
Top Bottom