> > Tell us how you went about doing collision detection.

> I did say it was simple, all I do is check to see if there is traffic
already
> being transmitted before I try to transmit.  I check by just sampling the
> received RF data for a few bits worth of time.  If I can receive some data
> (variations in the bit stream), I just wait for the max packet length (all
my
> packets are the same size) and try again, checking for data before each
send
> attempt.  That with a CRC16 is what I mean by simple.

Good first start.

A significant improvement is obtained by backing off for a random period
AFTER the end of the packet being sent and then sending if still free. This
allows multiple would be senders to compete randomly for the free channel.
If you don't do this and their are several stations waiting to send they
will ALL see the channel free at once and ALL try and send at once and ALL
collide and then ALL retry again in due course. Worst case, if response
times are closely matched you could spend most of your time in collisions.
This basic scheme is historically referred to as "Aloha". Sending on fixed
method length boundaries with random backoff is modified Aloha. (Name, not
surprisingly, from early Uni of Hawaii radio net).

Random backoff to allow multi station competition is a standard feature of
most collision detection schemes. Note that it should be as truly random as
you can reasonably get to prevent some stations hogging the circuit. If you
all drop into lockstep and compete together repeatedly you can also be in
trouble.



        Russell McMahon

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics