I'm puzzled by the implication that thermal noise or avalanche or Zener noise is somehow inferior to noise from radioactive sources. It's not. Take as an example Johnson noise, the voltage noise from a resistance. It's the result of the interaction of vast numbers of electrons. It is unpredictable in the same way that individual radioactive decay events are unpredictable, and they are both results of friendly quantum mechanics. Small biases/interferences that come from nonideally sampling such voltages do not matter, since the entropy is still 1-epsilon bits per sample (more for nonbinary samples). There is no "chink in the armor" available for cryptanalysis---the user need only acquire N*(1+epsilon) random bits rather than N bits. The video-snow-noise described by Tim May is mostly Johnson noise in the low-noise-amplifier electonics, not atmospheric or ionospheric noise, at least above 50 MHz or so. Cheers, Peter Monta pmonta@qualcomm.com Qualcomm, Inc./Globalstar