static / popping audible when doing quick seeks through track

[Owen Williams]

I have a knob on my controller set up to act like a strip-search input -- if I twist it, it seeks ahead by a certain amount so I can quickly preview a track. On current master, when I do this, I hear some static and popping -- it sounds like the crossfade over the sync isn't working right. I don't hear any pops when doing something like a loop

[Owen Williams (ywwg)]

crossfade over the seek, I mean. This is with master sync and quantize on, so each seek is quantized.

[Owen Williams (ywwg)]

I do hear the static even without quantize on, though.

[Daniel Schürmann (daschuer)]

This is probably the known issue, produced by cache misses.
The old stream is faded out, but there are no samples for fade in. Once the cache is in place the samples start without fading.
This should be solved in a common solution with Bug #1435170, to avoid introducing a third place for ramping code.

[Owen Williams (ywwg)]

Yeah, there's a fade out (crossfade to empty samples) and then the next buffer is full volume. In case of a cache miss we should probably not fade, then crossfade the next buffer from the last sample value. Or we could keep the fade, and then zero out the new buffer until a zero crossing.

[Owen Williams (ywwg)]

Actually in case of a cache miss we could write the buffer all the way to the last zero crossing and then write zeros to the end of the buffer. It would be pretty easy to write testable sample_util functions like "ClearAfterLastCrossing" and "ClearUntilFirstCrossing"

[Owen Williams (ywwg)]

Hm, currently enginebuffer has no way of knowing that we had a cache miss. CachingReader has been hacked to never report failure, though that's easy to fix. ReadaheadManager should also report that upstream. Then, finally, the scalers should report a cache miss by way of an output parameter. We can still handle the error case as we do now by zeroing out the missing samples, but we shouldn't be hiding the error like we are.

Once the scalers report the cache miss, we can decide how to handle it in enginebuffer.

[Daniel Schürmann (daschuer)]

It dos not make much difference if the stream stops at a zero sample or at any other sample value.
You can check it using Audacity Spectrum Analysis.
Once you stop without fading, you get a click due to the rectangular gain change.
The only solution is to fade out the last stream, wait until the samples are cached and fade in the new stream.

By the way: we can improve by using a Hanning window instead of our triangular ramp.

[Daniel Schürmann (daschuer)]

This has been fixed.

[Swiftb0y (swiftb0y)]

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/7943