Re: non-pseudo random number generators

From: Mitch Harris (harrisq_at_tcs.inf.tu-dresden.de)
Date: 07/13/04 

Date: Tue, 13 Jul 2004 10:10:33 +0200

Barb Knox wrote:
> Mitch Harris <harrisq@tcs.inf.tu-dresden.de> wrote:
>>Barb Knox wrote:
>>> Mitch Harris <harrisq@tcs.inf.tu-dresden.de> wrote:
>>>>Peter Olcott wrote:
>>>>
>>>>>There are now hardware based random number generators that are not
>>>>>limited to pseudo random.
>>>>
>>>>Really? What are these? Geiger counters?
>>>
>>>That would work, but would be bulky. Practical chips currently use
>>>white-noise diodes, where a random bit is generated depending on the
>>>instantaneous level of the diode's thermal noise waveform. This varies
>>>rapidly and essentially randomly over time. I'm not sure whether quantum
>>>randomness comes into this in a big way (like it does for Geiger counters),
>>>but the classical (statistical mechanics) thermal noise is sufficiently
>>>random anyway.
>>
>>And these well calibrated to deal with change in physical
>>properties? temperature? material deterioration?
>
> Yes. I expect that somewhere on the Web is an article describing various
> tests of randomness applied to such a device. As for deterioration, it's all
> solid-state (oops, "condensed matter"). Temperature does affect the noise
> waveform, but it's easy to compensate for any variation. For example, one
> algorithm to generate a random bit might be: take an 8-bit ADC sample of the
> instantaneous voltage; take another sample; if the first is less than the
> second then the bit is 0, if greater or equal then the bit is 1. This way,
> the PP voltage of the waveform doesn't matter (as long as it's in the useful
> range of the ADC).

- cool
- which is provably more trustworthily "random"? the pseudo
random generator or this physical device?

To me there seems to be more control over the randomness in
a pseudo random generator. 

-- 
Mitch Harris
(remove q to reply)