Path: utzoo!attcan!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!hplabs!hp-pcd!hplsla!tomb From: tomb@hplsla.HP.COM (Tom Bruhns) Newsgroups: comp.sys.amiga.tech Subject: Re: Random #s (was Re: ARexx pseudo-random number generator) Message-ID: <170008@hplsla.HP.COM> Date: 30 Sep 89 00:26:08 GMT References: <194@teslab.lab.OZ> Organization: HP Lake Stevens, WA Lines: 27 andrew@teslab.lab.OZ (Andrew Phillips 289 8712) writes: >. >. >. >Another interesting idea that I saw in sci.crypt recently is to test a PRNG >by drawing pixels on a screen. A good one will make something that looks >like snow on a TV. The eye can very easily detect some patterns which are >caused by a poor PRNG. ... A way to make this extremely sensitive for at least some schemes: Method A: Plot two sequential points as x and y on the screen. But only plot them if the first of the two is within a range which is very narrow compared with the range of the generated numbers. Obviously, spread its axis so that narrow range fills the screen. Method B: Draw numbers till one falls in a narrow range, as in method A. Draw two more, and plot them as X-Y. Method A will display problems in congruential generators; method B is nice to see how you have done on a binary shift register. Expect to see a pattern like a sawtooth wave for A and a pair of interleaved triangle waves for B, on the suggested generator types... >-- >Andrew Phillips (andrew@teslab.lab.oz{.au}) Ph. +61 (Aust) 2 (Sydney) 289 8712 >----------