Marcel Birthelmer wrote:
> Hmm, from a strictly theoretical perspective - the 8-point "curve" genera=
ted
> by 8 bandpass coefficients is the frequency response (sort of), so if you
> take the FFT of the input signal and multiply it with the bandpass values=
,
> you have basically convolved the two signals together. An IFFT would then
> give you the adjusted signal. Whether this is doable on a dsPic, though,
> I don't know.

That would be a really bad way to implement a real-time filter.

However, what you *can* do is take the IFFT of the bandpass coefficients in
order to create the impulse response of the desired filter, and then use th=
is
sequence as the coefficients of a FIR filter. That actually works quite wel=
l,
assuming you have the CPU horsepower for a reasonably long (high-order) FIR
filter. A 512-point filter at 48 kHz sample rate is only 24.576 MMAC/sec, b=
ut
I don't know offhand if the DSPIC has the memory and data paths to support =
a
filter this large.

However, I'm not sure that this is what the OP was asking for. His descript=
ion
included the mention of 8 DAC channels, which makes me think that perhaps h=
e
wants 8 separate outputs, one from each of his filters. (Fancy color organ,
perhaps?) It wasn't clear whether or not the filters would all be using the
same input signal.

-- Dave Tweed
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.