AVR-Chat

I would use async serial over RS422/485 with simple packets
containing a CRC.

One of the big problems with TWI in this situation is that
the addressed device has to ack in a particular bit time.
In a long wire system, this means that you have to slow
down the clock. Further, the transceiver is not built to
drive hi-C loads which you get with long wires. 

Also, the master is supposed to provide the clock. If you
have multiple "masters" (how else would you have
"collisions"), then you would have multiple devices trying
to inject clocks onto a single clock line. This all sounds
to me like a prescription for big problems.

Jim


On Fri, 21 Jul 2006 07:42:59 -0000
 "gary skinner" <g.skinner@elec-solutions.com> wrote:
> I need to have up to 100 AVRs talk to each other over
> unknown topology 
> network with long wires. collisions expected.
> TWI can handle this nicely other than the long wire spec.
> philips makes a I2C extender part that basically gives
> SCK & SDA a 
> boost to allow low pullup R's to work. the driver has
> ~30mA drive and 
> allows a pullup of 330 ohms - now if I reduce the speed,
> things "could" 
> work well.
> the mega8 data sheet is not totally clear if the low
> drive when TWI is 
> turned on is a standard port pin capability, or a much
> reduced (3mA) 
> drive, it does say it is slew limited. the other thing
> that is not 
> obvious is if I can use the internal port pin pullup as
> the I2C pullup 
> between the AVR and the extender chip as that "network"
> is actually 
> point-point over no real distance. (the bad network is on
> the other 
> side of the driver)
> 
> what I am really asking is if I really need the driver?
> 
> anyone experimented?
> 
> 
> 

---------------------------------------------------------------
The Think Different Store
http://www.thinkdifferentstore.com/
For All Your Mac Gear
---------------------------------------------------------------