[tech] AVR

[Ian McKellar]

On Mon, Mar 27, 2000 at 10:43:39AM +0800, John West McKenna wrote:
> Yakk writes:
> 
> >I think Leigh's work was based on wavelets, I don't want to implements 
> >wavelets procesing code on a microcontroller. Of course this is _exactly_
> >what we should be using beige's DSP for.
> 
> So what's wrong with sampling the audio on the AVR, sending it to Random
> Host by RS232, then over to beige by ethernet, doing a little processing
> and then sending the results back the same way? :-)

Umm - what sort of sampling rate were you thinking of? 115200 serial should
just be enough for 8bit 11hz...
> 
> Who said that DSP stuff must be done on a DSP?  If Christopher can write
> Wolf-3D on a Toad, then an AVR (somewhere around 30 times faster) can do
> this.  Don't let the absence of a multiply instruction put you off.  The
> Toad doesn't have one either.  (Here's a challenge for you: write a pair
> of fast 8*8 and 16*16 multiply routines for the AVR.  You can download a
> datasheet for the AT90S4433 from http://www.atmel.com)

Hmm - I guess we can look for a set of FFT routines which don't need 
multiply. Or write them...

> OK, maybe not the wavelets.  But these are unusually quick little micros
> and I'm sure they're capable of some interesting things.

I'm sure we can do something, but low pass filters, etc to do our FFT in
hardware might be the right way to go.

Ian

-- 
  ___________________________________________________________________________
 /\                yakk at yakk.net.au  http://www.yakk.net.au/                /\
<  >          finger yakk at student.uwa.edu.au for more information          <  >
 \/_____Disclaimer: These comments are those of my fingers, not my mind_____\/
                   Peer Pressure: Thats what friends are for.