Can you try to describe what you're hearing? I'm curious if it's the same thing.
Idk how to desscribe it, it's pretty much similar sound as on 'diff' button, also diff or not, it's always about 80% on right channel.
Found it. It took me almost a full day starting at it and debugging it before I realized that there was a typo in the code. If the phase difference was very big, due to that typo we were looking at the left channel instead of the right channel (so it said "L" instead of "R"), which affected the audio in the right channel. I think there's another improvement possible (it may do too much in some cases), but the main issue that you heard seems to be solved.
This wasn't a new issue by the way, but changes in the new version made it far more noticeable.
I figured that this image might be interesting.
The "Phase limiter" is there to avoid extreme phase differences between left and right, of which complete anti-phase is the most extreme version. For lower differences we want to reduce the maximum phase difference so not all sounds disappear when you listen in mono; extreme differences often indicate problems with the source material and will typically sound better if you just remove the phase for such instruments (yes, this works on a per-instrument basis) completely.
The image below shows the mono output of 2 tones that starts in complete anti-phase, then move to complete-in-phase, in versions 9.83, 9.84 previous beta's, and the new version, with the maximum phase difference set to 135 degrees.
1. Up to version 9.83, setting the value above 90 degrees caused complete anti-phase to stay there. So it didn't repair it at all.
2. Previous beta's fixed that, but did WAY too much.
3. This is what it's doing now, and I think this is what it should do.
Aside from doing way too much, the previous beta's (2) also caused weird artifacts on the right channel. That's a separate issue, which isn't really visible in graphs like this.
90 degrees will cause a 3 dB level drop in the output, 135 degrees a 6 dB level drop, so 135 degrees is probably too much if you expect a lot of people who listen in mono. Also, very high values tend to sound annoying.
If you use "Limit maximum phase offset" under "AZIMUTH", you can control both thresholds (the maximum number of degrees difference and the distance above which it removes any phase differences) separately. All other places where this type of filters are present will try to choose a "sane value" based for the 2nd based on the 1st.
phase.png [ 35.97 KiB | Viewed 1757 times ]