On Thu, Mar 13, 2003 at 08:14:57PM -0500, Jason Healy wrote: > At 1047600013s since epoch (03/13/03 19:00:13 -0500 UTC), sean finney wrote: > > On Thu, Mar 13, 2003 at 11:43:54PM +0100, martin f krafft wrote: > > > 1. is there already a package that enables this? > > > > not that i know of. > > While there isn't a package, there is a guy who is doing this on a computer: > > http://www.random.org/ > > Perhaps he'd share his source code? > > He also has links to this site: > > http://www.cs.berkeley.edu/~daw/rnd/ > > Which had the following interesting-looking links about audio: > > ftp://ftp.dnai.com/users/shipley/audio_rand.tar.gz > http://www.cs.berkeley.edu/~daw/rnd/audio-entropyd > > None of those are white-noise RNG (like the one you described), but > they may still fit the bill for your needs. >
"White noise" is a theoretical ideal that is never realized in practice. White noise is a signal that has the same power level at every frequency. In any real analog circuit, there is an upper half power frequency, and in most there is also a lower half power frequency. In a digital sampler there is also Nyquist folding frequency, which is half the sampling frequency. You should be knowledgeable about all this theory before you embark on this project. But the randomness that is derived from computer OS data logging is also surely not truly random either. The package that captures entropy has some fancy footwork to extract good randomness from the raw data. It should be possible to speed the creation of randomness by feeding noise data from an analog amplifier into this process, but be sure you feed it in at the right place. Where is that place? Ask an expert on the entropy pool stuff in the kernel. Randomness is a tricky business! And if you do produce a system that is "theoretically perfect", be sure to test its output quite thoroughly for randomness before you make the claim. There have been many embarising failures in this line of work, and they are not well documented! Good luck! -- Paul E Condon [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
