I have made a "random" routine for a CSMA/CD network in a
16C84. I used bit0-3 from TMR0, it gave me a 16 step
"random" number. With 208 uSec delay pr. step, it gave
a delay from the last collision, between 0 and 3.12 mSec. (to
that, I added 5 mSec). It works very well, due to the fact, that
all the 4 MHz x-tal on the network, are not synchronized.

mvh
Arne Hedevang
Denmark


> I need a routine that returns a random number to
implement something similar
> to CSDMA/CD. I need it to make a random wait time after
a collision, i've
> seem some routines but how do i get different seeds to
inicialize the
> routine ?

Hmm... depending upon the number of nodes in your system
and how they're
powered you /may/ have a problem.  While using a timer to
seed your RNG is
a nice concept, it won't do much good if several PICs
happened to choose
the same seed (as would be quite possible if all the
machines were powered
on simultaneously).  If the seeds happen to match, it doesn't
matter how
many times you cycle the RNG; all the "random" numbers
are going to match
as well.

In many network topologies, a node's address is used to
seed the
random-delay routine (or else simply sets a fixed delay
duration).  Would
such an approach be feasible for you?  If every node has a
unique address,
that would provide a good way of producing the "random"
delays without the
"bad" cases you could get with a timer-seeded RNG.