Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!odi!ed
From: ed@odi.com (Ed Schwalenberg)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Re: Hardware random bit generator needed
Message-ID: <1990Oct5.142304.24704@odi.com>
Date: 5 Oct 90 14:23:04 GMT
References: <1234@manta.NOSC.MIL>
Distribution: na
Organization: Object Design, Inc.
Lines: 12
In-Reply-To: psm@manta.NOSC.MIL's message of 4 Oct 90 19:57:49 GMT

To produce truly random bits, you need to rely on truly random events,
such as radioactive decay.  Sampling the state of a high-frequency
digital oscillator every time you detect a radioactive particle will
give you truly random bits.

Producing pseudo-random bits is easier.  The canonical method is to
use a shift register, the input to which is the XOR of, say, the 19th,
22nd and 35th bits in the shift register.  The length of the register
and the number of inputs to the XOR gate can be varied at will.  I don't
know how to prove, for a given configuration, when the pattern repeats;
certainly if you need 2^32 non-repeating bits one of the inputs to the XOR
function must come from bit 32 or higher of the shift register.