From: Colin Hales (colin@versalog.com.au)
Date: Sun Sep 01 2002 - 18:08:52 MDT
I spent a couple of years (straight out of Uni, 1979) working in England
analysing the garbage collected by an airborne sea-search radar.
Unbelievable masses of data tapes. Fun to collect, though! Buzzing russian
trawlers off the Irish see at 300 feet and drinking really bad packet soup.
Eeeeeehaaaah! That was the cold war, for me, anyway. What a joke.
The data is called called clutter. Wicked stuff, clutter. Looks white at
first glance, but --nope-- all sorts of deterministic phenomena are buried
in it. I wrote a whole bunch of software to do autocorrelation functions and
power spectral densities. A 'Nova' minicomputer. wow. Fortran. eeek. them
wuz the daze!
Thing is, you've really got to watch out for behaviour on different
timescales. Deterministic behaviour can poke its wicked little head into
things in all sorts of ways (which is how you see a periscope/snort sticking
out of the waves in a force 8 gale from 50 miles).
I am very suspicious of 'random number generators'. White noise is a
mathematical construct only. The autocorrelation function is a mathematical
infinite spike (Dirac Delta function), the power spectral density uniform
from zero to infinity Hz. Real noise isn't that.
My advice:
Buyer beware again. Buy to suit the application and know what will be wrong
if the numbers aren't random in your particular context. Maybe run it
through a Fourier analysis for a while and see what it looks like over the
operational timescales in your project.
cheers,
Colin
*sorry about the reminiscence*
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:16:37 MST