Subject: Hardware random number generator
Summary: Very high frequency oscillator sampled by a low frequency oscillator
From: battle@cs.utk.edu (David Battle @ University of Tennessee CS Department)
I had the following idea for a hardware random number generator.  It works
something like this:

        +-------------------------------------+
        | High frequency oscillator (~ 10 MHz)>------+
        +-------------------------------------+      |
                                                     |
    +------------------------------------------------+
    |
    |   +----------------------------+
    +---> DATA    Serial in parallel |
        |         out shift register |
    +---> SHIFT       OUTPUTS        |
    |   +---------v-v-v-v-v-v-v-v----+
    |                outputs go to LED bank or
    |                parallel port input
    |
    |
    |                low frequency oscillator could be replaced
    |                with output from parallel port for easy
    |                synchronization
    +------------------------------------------+
                                               |
        +-----------------------------------+  |
        |Low frequency oscillator (~ 10 KHz)>--+
        +-----------------------------------+

Presumably the randomness comes from slight fluctuations in the frequency
of both the high and low frequency oscillators.  Obviously the low frequency
oscillator must shift 8 times for a whole byte's worth of bits to be
generated.

I built one of these and although I haven't conducted any statistical
tests, the the output bits *seem* random.  I used a crystal oscillator
for the high frequency oscillator and a 555 timer for the low frequency one.

Does anyone have any comments about the usefulness/practicality/reliability
of such a random number generator?

						-David L. Battle
						battle@cs.utk.edu