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

[Curtis Maloney]

Reza Naima wrote:
> So, I'm working on a project where I'll be taking data from an MAX1031
> 16-channel 10-bit SPI based A2D and need to stream the data out via an
> infrared link.  The max SPI clock of the MAX1031 is 10mhz, and I need a
> sampling rate of around 20khz/channel.  Will any of the AVRs (I know
> there's a new small 20mhz one out) be fast enough to read the data, and
> a checksum byte, and somehow modulate an IR LED?  The reciever will be
> in close proximity, but going through a layer of opaque material.  It
> can be configured that minimal outside light reaches the receiver.  

FYI - There's also a few Mega series AVRs at 20MHz now.
(I wonder what happened to the 24MHz parts that showed up briefly on the Atmel 
web site :)

Assuming it takes you 3bytes to issue a "read all channels" request, plus 2bytes 
per channel, your bandwidth on the SPI is looking like:

20kHz * ( 16channels * 2B reply + 3B request + 1B chk ) = 720KB/s

So, nothing too scary there, so long as you don't need to do masses of 
processing on the data.

Depending on the modulation scheme you want for your IR, the timers can 
alleviate a lot of CPU load.  If you're wanting to use IRDA, there are cheap 
external chips that can do the modulation for you, straight from your UART, from 
people like Microchip, Zilog, and Vishay.

If you haven't decided on your IR scheme, yet, you might want to look at a 
remote control demodulator on the receiver (like a Sharp IS1U60) which gives a 
TTL output for the presence/absence of a 38kHz IR signal.  Then you can OOK 
modulate your UART output with a timer pin set to 38kHz.. I don't remember the 
chip number, but there are some 8ping chips with and gates with driver 
transistors... ideal for this use.

Of course, my calculations have been wrong before on this list.. :P

--
Curtis Maloney