<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<div class="moz-cite-prefix">On 9/8/25 2:40 PM, Tony Patti wrote:<br>
</div>
<blockquote type="cite"
cite="mid:027c01dc2109$310fe5e0$932fb1a0$@glassblower.info">
<pre class="moz-quote-pre" wrap="">XORing 4 of the crystal oscillators (in hardware) was VERY GOOD,
XORing 16 of the crystal oscillators was EXCELLENT.
</pre>
</blockquote>
<p>I like it. </p>
<p>As I think about it a little I worry about RF coupling between
the oscillators, but these aren't easily persuadable RC
oscillators, these each have a rather stubborn crystal at the
heart, keeping them very independent. Fun.<br>
</p>
<p>It reminds me of my favorite entropy source in this era: The
multi-GHz system clock counter deep inside a modern CPU, beating
against the servicing of an interrupt source, <i>any</i>
interrupt source, the more interrupt sources the merrier. For
those relative timings to be known precisely off-chip seems at
minimum hard. And to know those relative timings at any distance
seems impossible. It feels like some sort of Heisenberg-like
effect should make it <i>theoretically</i> impossible, but I'm
satisfied with how practically impossible it seems. </p>
<p>Now xor it with the built-in——even if not honestly built——CPU
random number generator, and things get even better. To sleep
extra soundly quietly slide the phase of between these different
sources now and then by some unadvertised amount, and even the
crooked ones are very useful.<br>
</p>
<p>-kb</p>
<p>P.S. Isn't there someone here who now should start using the word
"squish"? I think that was it.<br>
</p>
</body>
</html>