hi: I'm busying with a True Random Number Generator. Would you please help to give some suggestion? Thanks! --- Best regards. Jun Zhang Phone: 010-88354422-114 Mailto: zhang_jun at tsinghua.org.cn _______________________________________________ http://www.opencores.org/mailman/listinfo/cores

I would definitely start with Knuth - "Seminumerical Algorithms" (chapter 3) its a classic and a surprisingly fun read. -its in any library and on many good programmers bookshelf. "Numerical Recipes In C" has a very good section on this issue also as I recall. I'm not sure how "True" you need - but those will get you pretty close without too much work. bj Porcella http://pages.sbcglobal.net/bporcella/ ----- Original Message ----- From: "Jun Zhang" zhang_jun at tsinghua.org.cn> To: cores at opencores.org> Sent: Tuesday, March 09, 2004 5:18 PM Subject: [oc] Any suggestion about Random Number Generator. _______________________________________________ http://www.opencores.org/mailman/listinfo/cores

Thanks a lot!
In fact, I'd like to implement a TRNG in FPGA ( such as Xillinx
Virtex-II series ), then many ways such as detecting nuclear decay
become infeasible.
So maybe the only way is to sample two oscillators' frequence deviation.

Anyone has such experience?

At 18:04 2004-3-9 -0800, you wrote:

--- Best regards. Jun Zhang Phone: 010-88354422-114 Mailto: zhang_jun at tsinghua.org.cn _______________________________________________ http://www.opencores.org/mailman/listinfo/cores

hi: I'm busying with a True Random Number Generator. Would you please help to give some suggestion? Thanks! --- Best regards. Jun Zhang Phone: 010-88354422-114 Mailto: zhang_jun at t...

Hi.

A few years ago I build a random number generator by xor-ing together
amplified zener noise samples. I used a 40Mhz 8-bit A/D, and sampled
many times, xoring each sample onto a shift register that was rotating
the bits. I'd read it out a something like a megabye every few
seconds. One important trick was getting the amplifier right.

The results were great. I generated a full CD of random data, and
passed the Die-Hard tests. In fact, I found a bug in the DOS port of
one of the tests. So far As I know, this was the first hardware
random number generator to pass these tests.

The important thing to note is that randomness increases rapidly with
xored values. If there is even a small percentage of "true"
randomness in the input signal, you just need to xor enough input bits
together until you get the desired precision. For example, if xoring
10 bits gives you a random value of .5 +/- about .1, then if you xor
20 bits, you'll get .5 +/- about .01. Do another 10 bits, and you get
another decimal place reduction in the non-random noise.

With a little carefull work designing parallel zener noise amplifers
and A/D converters, you should be able to spit out many megabytes per
second of noise so random that only God could tell the difference. I
still have the board if you need tips on the design. I think I lost
my hand-drawn schematics, but now that I've got gschem, I could
recreate them.

Do you think there's a market for such a chip? I think I could create
one fairly easily.

Bill

Hi.

A few years ago I build a random number generator by xor-ing together
amplified zener noise samples. I used a 40Mhz 8-bit A/D, and sampled many
times, xoring each sample onto a shift register that was rotating the
bits. I'd read it out a something like a megabye every few seconds. One
important trick was getting the amplifier right.

The results were great. I generated a full CD of random data, and passed
the Die-Hard tests. In fact, I found a bug in the DOS port of one of the
tests. So far As I know, this was the first hardware random number
generator to pass these tests.

The important thing to note is that randomness increases rapidly with
xored values. If there is even a small percentage of "true" randomness in
the input signal, you just need to xor enough input bits together until
you get the desired precision. For example, if xoring 10 bits gives you a
random value of .5 +/- about .1, then if you xor 20 bits, you'll get .5
+/- about .01. Do another 10 bits, and you get another decimal place
reduction in the non-random noise.

With a little carefull work designing parallel zener noise amplifers and
A/D converters, you should be able to spit out many megabytes per second
of noise so random that only God could tell the difference. I still have
the board if you need tips on the design. I think I lost my hand-drawn
schematics, but now that I've got gschem, I could recreate them.

Do you think there's a market for such a chip? I think I could create one
fairly easily.

Bill

Jun Zhang wrote:

_______________________________________________ http://www.opencores.org/mailman/listinfo/cores