For LegacyClonk we want to update from using SDL_mixer (version 1) to SDL2_mixer.
It seems to mostly work fine, however in Linux the sound quality is very bad.
It seems like high frequencies are missing.
But in Windows the quality seems to be ok, according to my testing.
The parameters passed to Mix_OpenAudio are equivalent to those of the SDL_mixer version.
Also, according to pactl list sink-inputs, the sample format, frequency and channel count are set correctly.
Does anyone have an idea what is going on here?
I would have edited the original post, but I can’t.
Update:
According to further testing, it seems that actually all platforms are affected.
Currently, I am guessing that this change roots in the “new” resampler in SDL2.
See e.g. http://hg.libsdl.org/SDL/rev/a8382e3d0b54.
At least the effect is very obvious for sounds which have to be upsampled from 11025 and 22050 Hz to 44100 Hz.
Now, when playing back these sounds with Audacity they actually sound very similar to SDL2_mixer.
SDL_mixer in contrast has significantly stronger high frequencies, which seems odd, but gives it better seeming quality.