[Cryptech Tech] User auditable hardware entropy source/random number generator

[Benedikt Stockebrand]

Hi Bernd and list,

Bernd Paysan <bernd at net2o.de> writes:

> Indeed. Looking up e.g. NXP's low-voltage avalanche diodes (PLVA6xxA) reveals 
> that they have been replaced with zener diodes (1/10 of the original noise! 
> Announced as feature of course...), and the original ones aren't available 
> anymore...

believe it or not: one of the diodes I received with the last batch came
with a notice in the manifesto that they (Farnell in this case) won't
restock on them...

> [...]
> http://pdf.datasheetcatalog.com/datasheet/good-ark/BZX55C30.pdf

That's exactly the type (well, the BZX55C6V8, that is) that I've got
from that set as well as individually, with both behaving drastically
different.  Argh...

> page 3.  The good ones for noise are the ones with a very steep plot.

...until they silently start to sell some "improved" models with
"better" characteristics or whatever...

But thanks for the parameters to look out for.  I guess I should really
focus on sorting this out---especially so because all the diodes from
the latest batch consistently failed one of the faster tests I've got.

I think it's only prudent to design the circuitry around those beasties
in such a way that I can effectively drop in whatever I can get hold of
in the future.  It's no use designing some kind of circuit where I
eventually can't get the parts for.

> With an FPGA, I would do the following: [...]

Since I haven't got any hands-on experience with FPGAs and right now
simply can't afford the time to start working on them as well, I
probably can't be of much help with them.  I'll stick to struggling with
the available components and amplification circuitry, extraction and
testing methods and generally things I can do with a microcontroller for
now.  Once that works, I'll start on FPGAs.

Anyway, I've tried something kind of similar on the ATtiny; it should in
theory be able to handle a minimum amplitude of IIRC .5V peak-to-peak,
but the test results that way weren't convincing.  My conclusion was
that I'd rather not depend on the properties of a specific
microcontroller, but rather feed it something close to an (unclocked)
TTL level near-digital input and that way get a chance to replace the
microcontroller in the future---especially so because there are some
DIP-14 PICs which have USB onboard.


Cheers,

    Benedikt

-- 
Benedikt Stockebrand,                   Stepladder IT Training+Consulting
Dipl.-Inform.                           http://www.stepladder-it.com/

          Business Grade IPv6 --- Consulting, Training, Projects

BIVBlog---Benedikt's IT Video Blog: http://www.stepladder-it.com/bivblog/