> Or you could use a high-freq astable, maybe a RC+inverter

The white noise generator is fairly slow compared to the PIC
but its waveform is quite unpredictable. Ashley mentioned
using digital gates as the amplifier and then one to square up
the output. It seems instinctive to have a clean signal into the
micro, but I think it would be better left dirty. That is, irregular
rise / fall times as well as irregular timing

> As usual (grin) you are being vague as to the exact needs
> of your circuit. What's the most important? Time? Randomness?

Randomness. The system I'm using now is to load Tmr1L into a
counter, count that number of white noise transitions, then sample
Timer1L again as the random number. It seems to produce what
I'm looking for. The table size is variable (set by external command),
and can be anywhere from 1000 to 100000 bits in size. Picking
random numbers for a small table takes longer, because so many
Timer1L values are rejected for being too big for the high byte, and
it can take a second or two to complete the full number but as long
as it's as close to random as possible it's not too important. If the
table size was set to 30,000d (7530h) for example then it would take
a short time to find an acceptable high byte (00 - 75) but longer to
find numbers less than 0x31 if the high byte is 0x75

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads