Re: Sorting Clock Generator?



"David Eather" wrote in message news:E_mdnV5OvJ0vxN7TnZ2dnUVZ_rOdnZ2d@xxxxxxxxxxxxxxx
"mike" wrote in message
I'm assuming any statistical problems would be lessened if the output
of the generators was broadened. So maybe instead of 0 to 1048576,
maybe the individual generators should output 0 to 4294967296.
More?

On 11/08/2011 11:11 AM, ashwood@xxxxxxx wrote:
[some apparently pretty stupid shit - edited by Joe]

He is talking about increasing the size of the output of the PRNG. He is not talking about increasing the number of PRNG.

My mistake. Then it would improve it somewhat (increasing length in the equations I cut out helps), wouldn't eliminate the problem though. Bringing the problem down below 2^-x probability requires more than x bits, so the equivalent of 80-bits of resistance to this would require more than 80 bits in each pull (I don't feel like breaking it down to a function). Clearly you'd be better off just XORing the two streams together.
Joe

.