Waveform or Grid drifting after zoom

[kek001]

 Mixxx 2.3 beta, windows 10

After using zoom, other decks are out of "sync"
Waveform or grid is drfting.

https://mixxx.discourse.group/t/strange-grid-behaviour-2-3-beta/21035

[kek001 (kek001)]

This is an other problematic grid behaviour

https://mixxx.discourse.group/t/mixxx-2-3-beta-grid-wobble/21117

[Daniel Schürmann (daschuer)]

I can reproduce the issue by disabling zoom sync in the waveform preferences.
Now I have hard times to sync the waveform Zoom manually.

We may improve the situation by Snapping the zoom into sync when coming close.

However, is the non synced state a valid use case for parallel Waveforms? Is the issue solved for you by enabling the zoom sync?

If yes, I like to close this bug.

[kek001 (kek001)]

Using the zoom sync helps for the first problem.

But it wont help second problem. Where grid is drifting without zooming.
https://mixxx.discourse.group/t/mixxx-2-3-beta-grid-wobble/21117

It happends when you load track, it looks fine first minute, but end of track or 2/3
can seen grid is not placed correctly.
I have verified 2.2.3 is working well, but 2.3beta the grid is not right positions.
So it can't be there is something about file or the track is not constant tempo.
The second one force me to stop using 2.3 beta.

First problem its strange the zoom method cant restore orginal state before using zoom, and messing with scale.

So enabling zoom for all tracks sync will not solve the error.

[kek001 (kek001)]

for me second problem it looks some point the grid or waveform shrink or expands.

[Daniel Schürmann (daschuer)]

Do you use constant beat grids (default)?

Does the issue change without changing the beat grid or is the order the same whenever you seek to that off beat position?

I think we have a rounding bug in the constant beat grid calculation. Maybe you are seeing this. Is the issue gone when you reset the best grid and reanalyze it with the conts flag removed (beat detection preferences)?

Can you point me to a significant track that suffers the issue? I can use it to test the rounding code fix.

[kek001 (kek001)]

Yes I do use Constant by default.

Q:"Does the issue change without changing the beat grid or is the order the same whenever you seek to that off beat position?"

A: Sorry I dont understand.

I was thinking it can be rounding or declaration or type.
I try to check PD track, what can use for testing.

[Daniel Schürmann (daschuer)]

The question is if we can assume that it is an issue of the analysis date that persists as long as the beat-grid is not adjusted manually, of if it is a visual issue that is floating around depending on waveformnams skin settings.

[kek001 (kek001)]

Ok. Thank you.

For first problem, zooming in out. I think its visual.
My default zoom is 25% If I zoom deck 1 in out and deck 2 is as loaded without zoom.
it looks like deck 1 is coming behind and chasing deck 2, and they are correct sync in middle (cursor line).

Second problem, i will triple check and create 8 min constant beat mp3 track, so i can be sure
there is no flavour or mistake from artist. At Wednesday i should have time to do it.

[kek001 (kek001)]

I have tried find logic for the second problem.

I check about 20 tracks and i often found the grid is 0.01 - 0.02 off from beat, even i tried to assign it again. Now my head is jelly, will continue tomo ....

[kek001 (kek001)]

I created long mp3 files using joint stereo , stereo and created by lame 3.99r and Sony Acid. mpeg III
different BPM.

The mp3 are very simple repeating Kick Drum, i dont know is it too simple and should I add more artifacts, which can cause differencies while lame packs mp3 file.

I can't reproduce 0.01- 0.02 off from grid.

It makes me scratch my head, Because yesterday, i pick up randomly 20 tracks from my set, and saw allmost all had shift from grid. and the ammount were same, and the track should be constant beat,

For the first problem enabling sync zoom with all decks, is somekind of solution, but it will leave "rounding" scale restore issuer..

I have to think thank second problem, so if you can give me time end of this week, before closing this topic.
Thank you.

[Daniel Schürmann (daschuer)]

Another cause of drifting waveforms are discontinues due to decoding issues.
In this case direct seeks to a position will land to different samples than playing to it. In addition our decoder has the issue that gaps are closed shifting that track instead of adding silence.

When this happens, the sound is out of sync, while the beat-grof is still on the waveforms.

[kek001 (kek001)]

I was testing some tracks in DJuced and Mixxx 2.2.3, my traktor machine is not near me now.

Mixxx 2.2.3 is working well and DJuced is working well and grid positions are right place.

This is not only visual thing, i test it. Sometimes kick or hihat is missing or its playing part of Kick while i am jumping to hotcue.

About month ago i was thinking whats going on and have to record one set five times because of this.
adjusting hotcues without quantization.

 Thinking whats going on, cant be so bad after half year pause, because there was odd sounds while jumping hotcues.

Now I am wondering why simple 14 min Kick.mp3 (lame 3.99) and an other one created by Sony Acid.
are working well and grids are correct places in mixxx 2.3 beta.

I try to check and think more at saturday.

[kek001 (kek001)]

While trying to find free track,
I saw sometimes verion 2.3beta track length is different than 2.2.3 or remaining time.
This one track it was 0.02, i dont know is it coincidence or related to the second problem.

Check attachment image

and here is track.
https://ektoplazm.com/free-music/step-high-celestial-colors

[kek001 (kek001)]

From same album you can check a track Green Path. Grids goes well 2.2.3 but not 2.3beta.

[Daniel Schürmann (daschuer)]

Cool Thank you very much. Which encoding did you test? Is one of the other encoding also affected?

[kek001 (kek001)]

I am not sure about, but i think it doesnt matter what encoder, or is joint or stereo. what I tested for my own plain kick. I will test and check more.

[kek001 (kek001)]

I was using mp3 file from ektoplazm. I will download flac and check how that will behave.

[kek001 (kek001)]

 Green Path flac file has correct Grid placement in Mixxx 2.3 beta !

[Daniel Schürmann (daschuer)]

Thank you for confirming.
This means we can close this bug?
The shifting issue with the mp3 file is a known issue when upgrade Mixxx for some files. Please file s new bug if this issue still persists after reamalze the file with Mixxx 2.3 beta.

By the way, I have a pending PR that aims to remove the static offset of the detected beat grid. Do you have interests to test this? https://github.com/mixxxdj/mixxx/pull/3626

[kek001 (kek001)]

Hi

Thank you for reply. Sure I would like to test it.
I lack of knowledge how to get it. So if you are kind and let me know how to download it.

[kek001 (kek001)]

Can I use this for test it.

Index of /builds/2.3/Windows

Mixxx-2.3-r8012-b8bb1bb8ae.msi 2021-02-13 16:02 94M

[kek001 (kek001)]

I meant the flac version of the track is working well, but not mp3, so it can't be the track Green Path has variable tempo. So this issue still has problem and is active.

[Daniel Schürmann (daschuer)]

This is the link to the windows installer of the version with improved offset correction:
https://github.com/mixxxdj/mixxx/suites/2039375465/artifacts/41020746

I think you need to be logged in on GitHub to make the link work.

Let's move the rest of our conversation to GitHub to not clutter this bug more.

[Swiftb0y (swiftb0y)]

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/10300