Erik Hofman wrote:
[EMAIL PROTECTED] wrote:
Features 192kHz @ 24bit inputs and outputs (48 bit fixed point
internal)
After thinking about it some more I think you can get away by using
32 bit fixed point (16.16) arithmetic.
Probably true, but would a 48 bit DSP cost much more?
When using integers to mix data it is best practice to add all
channels together and divide them by the number of channels (to avoid
errors) but when using 16.16 data one could divide the sample by the
number of samples first, before adding the channels together.
e.g., instead of:
r = (s[0] + s[1] + ... s[n-1])/n
you could use:
r = s[0]/n + s[1]/n + ... s[n-1]/n
Your formula is correct, but do not actually do division. Have a ROM
with the reciprocals and do multiplies.
Also, a DSP is designed to do FIR filters and, therefore, they usually
have some overflow bits in the accumulator. Probably not enough if we
have a lot of channels.
--
JRT
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
