Hi everyone. I'm working on a homebrew music app for the Nintendo Wii (16bit, 48000Hz output) and I would like to implement dither and gain options. A rough from-file-to-hardware process is shown below - At what point should I apply these steps?

Decode samples (16 or 24 bit) and convert to float format in the range -1.0 to +1.0
Apply gain here?
Resample (libsamplerate) the audio to 48000Hz
Convert samples from float to signed short - Apply dither during this step?
Send samples to hardware

Dither has been discussed here at length, I want to make sure I'm on the right path though...

Thanks!

"Convert samples from float to signed short"
This is when you add dither, before the conversion.

If you need amplification, it doesn't matter very much, but doing it while in float format would be my choice. The quantization errors are smallest here.

Thanks! Much appreciated.