Maker Pro
Maker Pro

True random nunber generator

B

BarNash

Jan 1, 1970
0
Hi all

Is there a way to implement a true random number generator with logic gates
,nand nor FF etc without using white noise from a zener diode for example?

Thanks
BarNash
 
B

Bob Eld

Jan 1, 1970
0
BarNash said:
Hi all

Is there a way to implement a true random number generator with logic gates
,nand nor FF etc without using white noise from a zener diode for example?

Thanks
BarNash

I made a pseudorandom noise generator for audio testing with a
microprocessor based 32 bit two tap shift register. The processor is a
PIC16F676.

The generator operates on a 4 MHz internal clock with a 77kHz chip rate. It
produces a white noise sequence that is pseudorandom, not actually true
random which is not possible in this kind of digital system.

Connected to the processor output is an RC network to convert the white
noise sequence into "pink" noise for audio work.

If you are interested, I can give you the code and other details. Email me
for details: nsmontassocatyahoodotcom. Remove the ns from the address start.
 
F

Frank Buss

Jan 1, 1970
0
BarNash said:
Is there a way to implement a true random number generator with logic gates
,nand nor FF etc without using white noise from a zener diode for example?

An interesting idea is to use the rule 30 cellular automaton (rule
classification by Wolfram). This produces a high quality white noise with
long cycle:

http://www.frank-buss.de/SignalGenerator/vhdl/RandomGenerator.vhd

A client for which I'm working uses this algorithm in a DSP for
multichannel decoherent white noise generators (multiple generator with
different seeds) and with some additional software, switchable to pink
noise. The device works very well.

Of course, it is not a true random number generator, because this is not
possible with deterministic logic without some analog tricks, or by getting
some entropy in complex systems with multiple interfaces to the analog
world, like the Linux kernel does.
 
M

Mike

Jan 1, 1970
0
Sorry, that's not true. Repeats are necessary for any RNG that has to be
taken seriously in simulations - and there are at least a dozen
algorithms that do it. The ability to produce repeats is also a
prerequisite of any RNG that is intended for cryptography - and there
are hundred if not thousands of algorithms to do that.

I will suggest that there is an infinite number of RNGs that repeat
the values.

Actually, it's infinite minus one. I'm about to write a Wikipedia article
whose existence will prove it.

-- Mike --
 
N

Nobody

Jan 1, 1970
0
A trick that can be very useful is to combine a little bit of truly
random stuff with a lot of pseudo-random so that the pseudo-random
isn't easy to break out. The XOR operation is good for this because
it is very like a multiply in its action. If you think about it in
the frequency domain, it adds and subtracts the frequencies. This
means that the repeat of the pseudo-random is not given away by some
frequencies being missing.

If you're talking about XOR-ing random with pseudo-random, you need as
much random as pseudo-random. It's more productive to use the true RNG to
periodically re-seed the PRNG.

XOR-ing independent streams has the advantage that a flaw in one stream
can't make the output any worse than just using the other stream alone.
 
Top