From: "James A. Donald" <jamesd@echeque.com>
Date: Fri, 19 Jan 1996 00:52:07 +0800
To: "Timothy L. Nali" <tn0s+@andrew.cmu.edu>
Subject: Re: Random Number Generators
Message-ID: <199601181557.PAA06221@mailx.best.com>
MIME-Version: 1.0
Content-Type: text/plain


"Timothy L. Nali" writes:
> The most promising design I've seen so far (that I can actually
> do) is based on clocking a D flip-flop in the following way:
> 
> [Shifts the output of a clock that he hopes will be sloppy into
> a shift register]

If you want noise, your circuit needs a known noise source, with 
known good properties

Your circuit has no known noise source, you are just hoping 
that there will be noise in it somewhere.

Johnson noise is amplified thermal noise thus it is known to 
be good:  Amplify johnson noise to signal levels, and then 
shift this random analog output into a long shift register.  
(You will need a long shift register to suppress metastable 
states.)  You should set up your low frequency analog 
feedback to get near equality of ones and zeros, and you 
should have digital feedback (similar to a CRC
generator) to get perfect equality of ones and zeros.

 ---------------------------------------------------------------------
              				|  
We have the right to defend ourselves	|   http://www.jim.com/jamesd/
and our property, because of the kind	|  
of animals that we are. True law	|   James A. Donald
derives from this right, not from the	|  
arbitrary power of the state.		|   jamesd@echeque.com