@Busnoise said:
Take a 16-bit signal, divide by 512, multiply by 512 -> only 8 bits of resolution preserved, right? (analogously to shrinking and re-enlarging a pixel image)
This won't work as pd's floating point magic will leave the signal (almost) unchanged.
(also in the image analogy, shrinking would be more like downsampling, and 8bit-sound would compare to reduced color depth)
you'll need something like this to actually "clip the bits"
[*~ BD]
| \
| [wrap~]
| |
[-~]
|
[/~ BD]
from discussion here
http://lists.puredata.info/pipermail/pd-list/2008-12/066923.html
searching for bitcrusher here and on the list will give you lots of options