> So, it seems that a randomness generator could be done 
> with the silonix
> part and a bit of uranium ore for a few dollars If a few 
> counts per sec
> is OK. For instance, one could count just the decimal part 
> of the
> seconds between pules to get 6 random digits (1mhz clock), 
> do that as
> many times as needed to get a longer number. I wonder how 
> big a random
> number has to be to be generally useful? I wonder if there 
> is a product
> in this concept?

There's a potential product in everything :-).

As even a truly random generator can only generate R^N 
states (radix R, N places) then a "universal" generator has 
to "simply" be big enough to generate enough states that the 
tasks that use it get swallowed up in its immensity by an 
adequate amount. (Some highly techo terms there :-) ).

So "how big is enough' is going to depend a lot on the 
target. Randomly selecting from people on earth wants, if 
done in one go, something suitably larger than 6 billion 
that when used to generate the selection it doesn't add bias 
due to its size. I'm sure that mathematicians would be 
horrified by that intuitive interpretation but it seems to 
make first cut engineering sense.

Stars in the universe or sand grains on the beaches of 
earth, which are about the same magnitude (as far as we can 
tell and as Abraham knew) needs something enough larger than 
about 10^23. If you are doing that by counting between 
events then I think your counter speed is going to be 
"rather high" :-). I'd guess that in such a case a multistep 
approach is going to be almost mandatory.

But for many purposes I suspect that something in the 
millions to (US) trillions range is going to be quite 
useful.


        Russell
 

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist