From: Jim choate <ravage@bga.com>
Date: Thu, 21 Jul 94 12:23:05 PDT
To: cypherpunks@toad.com
Subject: (fwd) Noise diodes
Message-ID: <199407211922.OAA23855@vern.bga.com>
MIME-Version: 1.0
Content-Type: text/plain


Path: bga.com!news.sprintlink.net!news.onramp.net!convex!cs.utexas.edu!swrinde!ihnp4.ucsd.edu!agate!msuinfo!harbinger.cc.monash.edu.au!aggedor.rmit.EDU.AU!goanna.cs.rmit.oz.au!not-for-mail
From: ok@goanna.cs.rmit.oz.au (Richard A. O'Keefe)
Newsgroups: sci.electronics,sci.math.consult
Subject: Noise diodes
Date: 21 Jul 1994 18:03:24 +1000
Organization: Comp Sci, RMIT, Melbourne, Australia
Lines: 35
Message-ID: <30la4c$bng@goanna.cs.rmit.oz.au>
NNTP-Posting-Host: goanna.cs.rmit.oz.au
NNTP-Posting-User: ok
Keywords: rng

Some people I'm working with have built a machine to generate "real"
random numbers, using a BC546 transistor as an avalanche mode noise
diode (12V Vcc).  The noise output is supposedly 100mV peak.  That's
then fed into an LM311 comparator, to generate 0/1 signals.  This is
then fed to a divide-by-2 counter.  When their CPU wants a random
number, it samples the output of the divide-by-2 counter eight times
at 6.25kbit/sec.

They did collect a bunch of samples from this, and claim that successive
samples did seem to be uncorrelated, but there seemed to be a slight bias
in favour of 0 bits.  However, they say the test results have been lost.
I don't really understand how the output of a divide-by-two counter can
be biassed this way
	(free-running biassed random 0s and 1s) ->
	(divide by 2) ->
	(sample at regular intervals) ->
	(take 8 consecutive samples as one random number)

They don't need to produce random numbers at a very high rate (a couple
of hundred a second is more than enough for their application).

I have a faint memory that there are several problems with generating
random numbers from noise diodes, but I can't remember what any of them
are.  The requirement is for
	- independent
	- equidistributed
	- random 0..255 integers
	- which remain so throughout a 0 to 40 degree Celsius range
If there is a standard way to get something like this, I'd like to hear
about it.  If there is a standard set of problems I should know about
and check for, that'd be great.

-- 
30 million of Australia's 140 million sheep
suffer from some form of baldness.  -- Weekly Times.