Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Random number generator

  1. Mar 29, 2010 #1
    I've been experimenting with a hardware random number generator. My circuit amplifies transistor noise and converts it to a digital output. My program (.Net) reads the output into an array of double precision values. My voltages range from -7 to 7 (well within the range of my ADC) and have 16 bit precision. The array of voltages distributes in a Gaussian fashion centered over zero. I want a distribution that has equal probability of every value’s occurrence.
    One method I am toying multiplies each value by 1000 then drops all the digits that are greater than 1, eg 10.7823432... becomes 0.23432....
    Using this method I get a distribution that appears to be random with equal probability for each value. Two questions:

    1) What is the technical term for a distribution of random values where every value has an equal probability of occurring?

    2) Is my method for obtaining this set of numbers reasonable?
  2. jcsd
  3. Mar 30, 2010 #2


    User Avatar
    Science Advisor

    1) Uniform distribution. You didn't say what you did for - numbers.

    2) Pretty reasonable, since Gaussian would be close to flat in a narrow interval around 0.
  4. Mar 30, 2010 #3


    User Avatar

    Staff: Mentor

    I wonder if you can't invert Box–Muller transform. Not that I have an idea how...
  5. Mar 30, 2010 #4
    I should have said I multiply by 1000 then form a new number (retaining the sign) by using only the decimal point and the numerical characters to the right of the decimal point.

    -10.7823432... becomes -0.23432
  6. Mar 30, 2010 #5


    User Avatar
    Science Advisor
    Homework Helper
    Gold Member

    If you have two independent Gaussian random variables [itex]x[/itex] and [itex]y[/itex] and you define a complex Gaussian r.v. [itex]z[/itex] such that

    [tex]z = x + iy[/tex]

    then the phase of [itex]z[/itex] is uniformly distributed over [itex](-\pi,\pi)[/itex]. This is because the pdf of [itex]z[/itex] is circularly symmetric (depends only on [itex]|z|[/itex]).


    [tex]\theta = \arctan(y,x)[/tex]

    is uniform over [itex](-\pi,\pi)[/itex], assuming you use the 4-quadrant arctangent
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook