Re: Realtime Statistical Analysis of RNG Output

From: Andrew Swallow (am.swallow@eatspam.btinternet.com)
Date: 03/04/03 

From: "Andrew Swallow" <am.swallow@eatspam.btinternet.com>
Date: Tue, 4 Mar 2003 22:36:20 +0000 (UTC)

"Jay Miller" <jnmiller@@cryptofreak.org> wrote in message
news:cj0a6vck7nvbn3hk53rqphrnamqb1mh2i9@4ax.com...
> Words by "Andrew Swallow" <am.swallow@eatspam.btinternet.com>:
> >Are you, or your customers, required to keep written records
> >to show that the equipment was in calibration?
>
> Too early to say. Why?
>

The easiest way of testing calibration is to run the suite
of tests.

The company procedure ISO9000 standard requires all
hardware (including tape measures) to be in calabration
and records to be kept. I am not certain what the security
standards require but they are unlikely to ban calibration
checks.

A means has to be provided for the calibration tests
to be run and their results recorded. The tests could
be run automatically on power up, constantly during
normal operation, as a speical function or a special
data collected performed with the data analysed off line.
The easiest way for the calibration to be recorded is
for the results to be printed out, however that can be
a little difficult if the machine is a hand held device
without a printer.

When recording raw data in a file do not forget to
insert the time, date and machine name/serial number.

> >Even if you fully test off line, quick tests like ensuring that
> >you do not get all zeros or all ones are worth while.
>
> Indeed - this is the sort of thing I've been coming back to. As long
> as the marketing literature says 'real time statistical analysis', I
> don't expect what actually happens is going to be an issue.
>
> However, I would also like to do it right. Is checking for all zeroes
> enough? How likely is it that the RNG goes bad and only fails, say, a
> FFT spectral test?
>

An all the same check will detect the more common faults
like a broken wire, particularly if performed before (and
after) the hashing. To get all the faults you will have to
run the full sets of tests. You will have to ask your
engineers what the estimated mean time between failure is.
(For a single circuit board it may be 10 years.)

This all comes down to how important is it that faults in
the RNG be detected immediately? For a lot of purposes
an all the same plus a monthly calibration run is sufficient.
When the user only uses the equipment once a year a
full calibration run followed by the production will
probably be needed.

Questions I cannot help you with are:
Is there room for the calibration software in the device?
Would the software that writes the offline data be smaller
than the calibration package?
Does your computer have sufficient free time to perform
the tests on all runtime data?
Is your customer so paranoid that even with full online
checks he still requires an offline calibration run to check
that the equipment has not been tampered with? (If so
they probably go as far as keeping the two machines in
different safes in different military bases.)

Andrew Swallow

Relevant Pages

  • Re: Calibration equipment design
    ... I guess the equipment manufacturers did the same a long time ago to have ... secondary standards are pricey, ... An above standard precision is ok. ... Making electronic equipment or calibration standards do not differ much, ...
    (sci.electronics.equipment)
  • Re: Attn: Fish Punsters
    ... Yes, I have a PAT box, but I can't use it in anger as it's a non-legit ... Cost of calibration is enough to make it not worth ones while unless ... I used to use the Clare tester; ... equipment that I used to check was tested on a three yearly basis, ...
    (uk.media.radio.archers)
  • Re: Realtime Statistical Analysis of RNG Output
    ... >Are you, or your customers, required to keep written records ... >to show that the equipment was in calibration? ...
    (sci.crypt)
  • Re: New portable DVD player....
    ... Electronic calibration in the panic before an inspection is gonna cost ... I remember one time the Canopus cal lab got overloaded and there was ... and my shop filled up with some sub's equipment within an hour! ...
    (rec.boats.cruising)
  • Re: New portable DVD player....
    ... > Electronic calibration in the panic before an inspection is gonna cost ... > I remember one time the Canopus cal lab got overloaded and there was ... > and my shop filled up with some sub's equipment within an hour! ...
    (rec.boats.cruising)
Loading