Emulator implementation.

[audetto]

Hi,

I would like to try to implement a more correct speaker behaviour in AppleWin.

For this I have a couple of open questions, which you might be able to help with

General questions:

1. inv_scale: with a brute force attack I checked that the minimum scale to ensure [-1, 1] is just below 3300. The current value of 1738 is a bit short of this. How did you select it? You say you used the equilibrium value (I get 1755), but this risks a lot of clipping.
2. sub question. is encoding volume-dependent?
3. initialisation: do the initial values of y1 y2 v1 v2 matter? probably very very little as I can see from the simulation

Emulation:

1. the hardest part is coding librosa.resample in c++ without making it too complicated. Do you have any suggestion on how to implement a very simple but acceptable downsample? Currently AppleWin does an average of the values every 23 samples.
2. I was thinking of coding a low-pass filter at 22050Hz and then decimate. What do you think?