From: "Kurt Buff (Volt Comp)" <a-kurtb@microsoft.com>
Date: Thu, 18 Jan 1996 10:08:57 +0800
To: "Timothy L. Nali" <perry@piermont.com>
Subject: RE: Random Number Generators
Message-ID: <c=US%a=_%p=msft%l=RED-06-MSG960117144111CF009D00@red-02-imc.itg.microsoft.com>
MIME-Version: 1.0
Content-Type: text/plain


If you're going to work with hardware to get really random numbers, why not 
go to the back of any of several PC-type magazines, and order the radiation 
detector board that someone is hawking? Can't really do any better than 
that, can you? Counting cosmic ray hits and noting their time differentials 
should be just what the doctor ordered, right?

Kurt

----------
From: 	Perry E. Metzger[SMTP:perry@piermont.com]
Sent: 	Wednesday, January 17, 1996 11:00
To: 	Timothy L. Nali
Cc: 	cypherpunks@toad.com
Subject: 	Re: Random Number Generators 


"Timothy L. Nali" writes:
> 	For a class project, I will be designing a VLSI cmos chip to generate
> truly random numbers (The chip will be fabricated).  I'm limited to a
> 2-micron standard cmos technology

> 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:

I'd say that the design you have picked has a couple of problems with
it. The first is that you are, from what I can tell, building a
synchronizer, which means that you may have metastability
problems. (Your diagram wasn't completely clear so I can't
tell).

Also, you are depending on a sloppy clock and a not sloppy clock
actually having the stated properties, which means you aren't really
generating randomness so much as hoping you can detect and exploit
it. As it is very hard to determine if a stream is really random, this
makes your life difficult.  Far better to try to use some analog
tricks in the circuit itself to generate the random numbers for
you. Of course, some of these end up producing metastability problems
of their own...

Can anyone point this guy at good texts on all of this? I've never
found one...

Perry