On Mon, 5 Jan 1998 10:40:33 -0500 Tom Sgouros
<tomfool@CIRCUS.GSO.URI.EDU> writes:

>Ok then, how about a hardware way to make the seed?  I would like a
>slightly higher quality random number generator than I can get by
>putting the seed in ROM.

If the project has asyncronous external inputs such as a user pressing
buttons or data from an external source, the exact time they arrive will
be pretty much random.  If not, the value in RAM at power-on is somewhat
random from chip to chip, but probably very consistent for a particular
chip.  The exact time the WDT expires is not synchronized with the PIC
clock.  Set the WDT for maximum duration and increment a counter as fast
as possible until it is stopped by the WDT expiring.  The value in the
low few bits of the counter should be rather random.

Using the WDT, it can take several seconds to collect the 20 or 30 bits
to make a good seed mumber.  Maybe this could be the primary random
generator in the random interval generator though.  Using the WDT as the
primary timebase, the distribution of interval times will shift with
temperature, voltage, etc.  Use all the bits of the counter to correct
for this by computing the number of WDT timeouts per minute, multiply by
the (random) number of minutes desired, and count down.


>-tom
>
>----------------------------------------------------------------------
>tomss at ids.net - 401-861-2831 - 42 Forge Rd, Potowomut, RI 02818 USA
>