Re: Creating noise in the frequency domain?

"(Random-0.5)*3.4626 (just do it for all real&imag values) should bring a
standard deviation of 1, some more or less. I got this number by try."

You should do this analytically. The variance is the definite integral of
x^2 from -1/2 to +1/2 = 1/12. The proper coefficient is sqrt(12) =
3.46410162.


.