[Cryptography] Entropy of a diode

[Peter Todd]

On Sun, Jul 24, 2016 at 10:51:58AM -0700, Ray Dillinger wrote:
> But yeah, if you're generating noise by reverse breakdowns it
> won't take all that long for your diode to turn into a friode.
> And friodes just aren't all that chaotic.

Not at all: a zener voltage regulator is simply a diode designed to be operated
in reverse breakdown. They're perfectly reliable and are very widely used so
long as you operate them within the allowed current and temperature limits
specified in the datasheet.

But of course they're designed to do so - a non-zener diode isn't. For
instance, Vishay's datasheet(1) for their version of the (very common) 1N4148's
specifies a maximum reverse current of 50uA at 20V/150degC, and all they're
really saying is that the diode doesn't leak more than that amount of current,
in that extreme situation. That's not to say you should use it like that.

Meanwhile, the common BZX79 zener does(2) indirectly list allowable reverse
current in the form of a total power dissipation and reverse voltage... but
doesn't have any specs at all for noise. So who knows what you'll get? You
could buy a batch that appeared to be sufficiently noisy for your RNG, then
find out the next batch wasn't when you went into production. Or even worse:
you might find that the noise actually decreases over time, causing your RNG's
to fail in the field.

tl;dr: Don't build noise sources with diodes operating in reverse breakdown
unless you can find a diode that actually specifies a minimum noise output in
reverse breakdown over the temperature range you're expecting to operate at.

1) http://www.vishay.com/docs/81857/1n4148.pdf
2) http://www.nxp.com/documents/data_sheet/BZX79.pdf

-- 
https://petertodd.org 'peter'[:-1]@petertodd.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: Digital signature
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20160725/cedce277/attachment.sig>