QOA Decoding: A tragic truth about speaker sound quality

[martandrMC]

I have been developing my own virtual file system library and built a music player ontop of it to stream DFPWM audio files from the web, here's the source code if you'd like to take a look: https://github.com/martandrMC/computercraft-cdn. However I was not satisfied with the sound quality of the codec so I went into writing a QOA decoder to get improved quality (but at a 3.2 times size increase).

Long story short, under my development environment, using CraftOS-PC, everything seemed correct and working, until I went to test it ingame. The quality was nowhere near what I expected from my tests. I went into the source code of the mod only to discover the speaker peripheral encodes the samples with DFPWM, the very codec that I wanted to supersede, before sending them over to the client.

[TheRedXD]

I don't know what was the goal in trying to encode audio client-side with DFPWM, but it honestly does not make sense to me whatsoever. We should be able to play high-quality audio. Forcing it to be DFPWM client-side does literally nothing other than make the audio quality worse (it won't even reduce the network load because the high quality audio is already being sent through, or at least should be getting sent through).

I would say we should be able to directly send .wav if we wanted to without any reduction in audio quality, at the expense of file size of course.

I'm hoping I'm not misunderstanding this, but it seems to me that server-side has nothing to do with this (?)

[Wojbie]

Allow me to point your attention at #1574 PR, you will find adding a config option has already been suggested and denied from main mod, but if you are dedicated to see this on your own server, you should be able to merge and build custom CC:T with that PR to add configuration option you want for yourself.

[TheRedXD]

the crunch of DFPWM is an intentional aesthetic choice

it's a stupid aesthetic choice, not gonna lie, especially not even having a config option for it

@Wojbie as for making a custom version, the sole reason this discussion exists is to implement this into the main mod. If we didn't care, we would just fork CC:T.

I apologize if I may be slightly harsh here, but time and time again, when devs do not provide a config option for an obvious thing that is subjective rather than objective, people get mad and it's literally never ever fixed. CC:T has had this "aesthetic choice" for ages, and despite having the computational power and network resources to do this, we literally can't without modifying the mod ourselves and redistributing it to players.

[martandrMC]

@Wojbie That is unfortunate, though SquidDev seemingly did not rule out the possibility of using another codec.

I'm definitely open to changing the audio encoding used for transport to something else

If you look at my QOA code you will find that it is quite trivial to decode it (about 160 lines of Lua), and it would still be better than raw samples, at 3.2 bits per sample compression. There is a case to be made here, though my post's original point was partly to show off my code and partly to show you all the folly of flying too close to the sun.

[martandrMC]

the sole reason this discussion exists is to implement this into the main mod

@TheRedXD I don't think so. My point was what I wrote above. I do not mean to force this change onto SquidDev, because if I did then I'd have made an issue/pull request instead of a discussion post. Though, again, I would like such a change. Besides, if not for anything else, this was a fun exercise and I am happy that QOA files even play in the first place.

[TheRedXD]

makes sense. my bad.

I would say that this definitely has to be changed though. We're left with no other option other than to modify the code ourselves and redistribute the mod.