Michael Rigby-Jones wrote:
>> Some time last week I was thinking about randomness and such
>> and I came up with an idea. I haven't had time to research it,
>> but I do find it intriguing. What if you took two separate
>> oscillators (possibly drifty RC oscillators with widely
>> different tempcos) and fed them into a shift register -one as
>> data the other as clock. Since they're not directly related to
>> each other, their relative phase will continually drift -
>> producing a continuous stream of uncertainty on the shift
>> register.
>
> The relative frequencies of the two oscillators will have a constant
> frequency difference in the short term so the output bit stream will be
> very predictable and most unrandom IMHO.

Short term is OK if it's used as an external perterbation of a digital
random number generator.  Imagine the reverse.  What would it take to create
a 1/0 digital stream using analog oscillators that is accurate enough to
cause even an 8 bit CRC to repeat consistently?  What about a 16, 24, 32, or
more bit CRC?  You couldn't design such a system even with atomic clocks,
and a CRC is pretty simple to compute in a PIC and only requires 4-8 bytes
of state.

The more state and therefore longer repeat sequence (assuming a reasonable
algorithm) of the digital random number generator, the less often it needs
an external random input to make the output indistinguishable from a truly
random 1/0 stream.

I'd be more worried that the two analog oscillators might have a little
coupling between them and sync up at a sub-harmonic.  That makes them no
longer two independent events.  Of course, if the PIC oscillator is separate
and not coupled, that can be the second independent clock.


*****************************************************************
Embed Inc, embedded system specialists in Littleton Massachusetts
(978) 742-9014, http://www.embedinc.com
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist