Re: [avr-chat] will an avr be fast enough...

[Reza Naima]

Thanks for all the info that I've recieved so far.  To clarify, I need
16x20khz sampling (20khz per channel).  I've looked at the datasheet for
the MAX1031 part 

http://www.maxim-ic.com/quick_view2.cfm/qv_pk/3872

and it's SPI protocol seems quite odd compared to all the other devices
I used.  It looks like you can initate sampling, wait some time, then
read from all the channels sequentially (mode 00).  Assuming I can get
the data into the AVR via SPI at the required speed, I am then open to
implement any type of IR communications, and I'm not forced to stick
with a 38khz modulated IR signal.  There are receives that work at
higher frequencies, and will investigate if I can simply do something
like manchester encoding and transmit the data directly (adding a
checksum along the way).  The receiver will be in close proximity to the
transmitter, and I'm hoping that it can be configured such that it will
not recieve any ambient light.

If samples are corrupt, it's not a big deal as long as the checksum
indicates it as being corrupt.  And the signals are changing quite
rapidly so simple compression schemes (like only sending data when the
signals change) is not a viable option.  However, the data of interest
is will only last a few hundred ms, so one option is to add ram, cache
the data, and then transmit at a slower rate.

The tricky part is that the entire circuit needs to be about the size of
a quarter, and as thick as two quarters.  This includes an amplifer, the
A2D, an AVR or some sort of CPLD/FPGA, the IR mechanism, as well as
whatever power conditioners required (power is transmitterd inductivly
to a coil that will wrap around the circuit).

Thanks!
Reza