[sdiy] Looking for cheap lowpass filter algorithm

[Seb Francis]

Antti Huovilainen wrote:
> On Wed, 16 Aug 2006, Seb Francis wrote:
>
>> Ah, 32 multiplies per sample is more expensive that I wanted.  I was 
>> hoping to have say <=4 multiplies per sample.
>
> Not going to happen. You might be able to get something semi-usable 
> with 16 multiplies but you'd really have to go to 24 to get decent 
> results.
>
> What are you doing where you can't spare extra 20-30 cycles?
>

I could spare enough cycles for 32 multiplies per sample per channel, 
but this will be cutting it rather fine and since the processing is all 
realtime (no buffers at all), and there isn't a fix amount of processing 
taking place .. it all depends on what parameters are being changed in 
real time.

So to be comfortable I think 16 multiplies per sample is what I'm going 
to try.  Magnus pointed out it only needs half this number of 
multiplications as they are symmetrical about the middle, but I don't 
think this will help me since it takes 1 CPU cycle to do a 
multiple&accumulate and also 1 CPU cycle to add a stored result to the 
accumulator.

Could you point me in the direction of where I can find how to calculate 
the optimal coefficients for a 16 tap FIR filter?

Thanks,
Seb