# Re: Random Numbers

From: Bill Unruh (unruh@string.physics.ubc.ca)
Date: 11/08/02

```From: unruh@string.physics.ubc.ca (Bill Unruh)
Date: 8 Nov 2002 17:42:21 GMT

```

Jem Berkes <jb2002-delete-this-AND-users@users.pc9.org> writes:

]> The thermal noise of a (hot) resistor or a zenor diode with current
]> flowing through.
]> To get the numbers out of the noise you need an A/D converter - and
]> here we go with determinism. Missing bits, effective resolution a.s.o.
]> For most app's this will work, though.

]Thermal noise is pretty good for randomness. I've heard the best way is to
]take a reverse biased zener diode and observe the fluctuations of current
]through the device. The zener breaks down and conducts in the reverse
]direction due to a quantum mechanical tunneling effect. Find any
]correlation in that, and physicists will love you for years to come :)

It is very very easy to put correlations into that. Just hook up a
capacitor across the zener, and there you have correlations in the
noise. Most physical systems have correlations over time introduced by
all sorts of dynamics. That the underlying fundamental process may not
is irrelevant, since you only see the fundamental process through a maze
of other connections, which have things like stray capacitances,
inductances, etc. Or A/D converters which like to put out even number
rather than odd. Almost all physical process have 1/f noise, which is
correlated (that is what 1/f means-- a spectrum which is not flat over
long time scales-- Ie, long term correlations).

]--
]Jem Berkes
]http://www.pc-tools.net/
]Windows, Linux & UNIX software

## Relevant Pages

• Re: Random Numbers
... ]"The thermal noise of a resistor or a zenor diode with current flowing ... but correlations are harder to detect and remove. ... One possibility would be to just use the output of your sound card, ...
(comp.security.misc)
• Re: Does this have a flaw in de-biasing an entropy stream?
... When I am talking about sampling the output I mean with a nominally constant frequency clock - I'll refer to that clock as the sample clock which will have a related sample rate or sample frequency. ... The random noise that clocks the flip-flop I'll call the noise clock with noise rate or noise frequency etc. ... As the rate of sampling the output increases, correlations become stronger and eventually unacepatble. ...
(sci.crypt)
• Re: Toaster to Generate Random Numbers
... ]>]should agree that thermal noise can be considered random ... ]> correlations, especially at long times (1/f noise). ...
(comp.security.misc)
• Re: Toaster to Generate Random Numbers
... Most (sane) people ... >]should agree that thermal noise can be considered random ... > correlations, especially at long times (1/f noise). ...
(comp.security.misc)
• Re: Toaster to Generate Random Numbers
... ]Is it possible to use thermal noise to generate random numbers. ... ]Get a toaster, and measure the noise generated by it. ... They all have biases, correlations, etc. ...
(comp.security.misc)