Possible bug in 5.1 encoding

[bmcfee]

Hey folks, I've been banging my head against a transcoding issue with 5.1 multichannel recordings, and I'm reasonably sure that it's down to an error in either oggenc or libvorbis. TLDR is that the last channel (surround-right) is corrupted.

I'm working with the following recording: https://freesound.org/people/Bokir26/sounds/383922/ but other recordings from different sources (eg https://freesound.org/people/ToXXuS/sounds/591235/ ) demonstrate the same behavior.

I first noticed the erroneous behavior when using sox to convert:

sox 383922__bokir26__commuter-train-03-traffic-jakarta-indonesia.wav 383922__bokir26__commuter-train-03-traffic-jakarta-indonesia.ogg

and the behavior replicates when using oggenc directly, both with version 1.4.0 (provided by ubuntu) and 1.4.2 (built locally from source):

oggenc 383922__bokir26__commuter-train-03-traffic-jakarta-indonesia.wav -o 383922__bokir26__commuter-train-03-traffic-jakarta-indonesia.ogg

I've also tried working directly from raw format and explicitly setting the channels/rate/bit depth, but no luck there as well.

For comparison purposes, here are spectrogram plots of the original channel 5 and that produced by oggenc:

And as a baseline, channel 0 looks as follows:

Channels 0-4 (front left/right/center + LFE + surround-left) all seem unaffected, so it seems unlikely to be some kind of byte-ordering or interleaving bug.

Flac encoding via sox does work fine (all channels preserved), so I don't believe there are any problems with the input audio.

Additionally, if I discard the last channel when transcoding (using sox [infile] [outfile] remix 1 2 3 4 5) everything works as expected. Similarly, if I skip a channel (eg discard the LFE channel 4: sox [infile] [outfile] remix 1 2 3 5 6), there is no corruption. It seems to be a fundamental issue with generating more than 5 channels: replicating channel 1 6 times (remix 1 1 1 1 1 1) triggers the error, but only on the last channel: the first five are preserved.

(Yes, I know this is not the sox repository: I'm only using sox to easily generate test cases that break things in controlled ways.)

Anyone have a hunch about what's going on here?

[bmcfee]

Quick follow-up: it also works fine if I add an empty seventh channel (?!) remix 1 2 3 4 5 6 0. (Or replace 0 by 1-6 to mix all channels as a seventh output). There seems to be something magically wrong with 6-channel output.

[xiphmont]

Currently on the road, but wanted to let you know that I saw the message and will be looking into it quickly. Of you don't hear back from me by tomorrow, feel free to pester incessantly.

Monty

On Wed, Dec 1, 2021, 13:16 Brian McFee @.***> wrote:

Quick follow-up: it also works fine if I add an empty seventh channel (?!) remix 1 2 3 4 5 6 0. (Or replace 0 by 1-6 to mix all channels as a seventh output). There seems to be something magically wrong with 6-channel output.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/xiph/vorbis-tools/issues/30#issuecomment-983930173, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACWMHEGGZC3ZMJOZ474URYDUOZQ65ANCNFSM5JFEAX4A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[bmcfee]

I don't at all want to pester incessantly :grin: but am checking in to see if there's any update / confirmation of this one?

[bmcfee]

Hey, just checking back to see if there's been any progress on confirming this?

[bmcfee]

Checking in again to see if anyone's been able to confirm or update this one?