Douglas Butler wrote:
>
> That is hard to do since Microchip tries so hard to make their chips
> repeatable...
>
> Maybe use a few LSBs from a very high resolution temperature
> measurement?  Can you use a human element, such as a high resolution
> measurement of how long they held the START button down?  Good noise
> diode circuits are usually complex, as are Gieger tube circuits.
>
> It is hard to find something random to start with.


Not at all! Think hardware. What about a R and C
attached to a PIC pin, with maybe 0.5sec before it
turns the pin high after powerup. Just loop the
code in a fast count loop until the pin goes high
and you will get a pretty good random number
from 0-255 with every powerup.

Electro caps in this size range (10uF to 100uF)
vary in value maybe 15% from one cap to the next,
and the resistors will be different by maybe 1%
to each other. Electros are hugely affected by
heat too. You could maximise instability by
using a smaller C, and very high R (like 3.3Meg)
the time to turn the pin high will be affected
by many factors. :o)
-Roman

PS. Generating good random numbers FAST is harder.

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.