[sdiy] Uniformly distributed noise generator?

Martin Klang mars at pingdynasty.com
Mon Jun 10 16:55:34 CEST 2013


On 10 Jun 2013, at 15:06, cheater00 . wrote:

> On Mon, Jun 10, 2013 at 3:15 PM, Scott Gravenhorst <music.maker at gte.net> wrote:
>> 
>> It has been sooo long since the implementation of the chuff chuff LFSR with only a
>> few bits and low clock rate that it is truly a ridiculous comparison to analog noise
> 
> That's not really good enough in my application. This is for a
> measurement fixture. I need true, never-repeating randomness.

Never-repeating, for realz?

If you don't need more than 1/ms for the next 4 billion years then a Mersenne Twister PRNG [1] will do just fine.

There are some nice implementations out there [2]. I developed a solution for the STM32F mcu's which will be published at some point.
Of course the STM32F2 and F4 have built-in true RNG, making the twister a bit superfluous.

If you are very, very particular about the randomness of your values then verify with the diehard (or dieharder) test suites [3].

hth,

/m

[1] http://en.wikipedia.org/wiki/Mersenne_twister
[2] http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/emt.html
[3] http://code.google.com/p/dieharder/



More information about the Synth-diy mailing list