Computer white noise generator, which software?

[Bricolo]

I'm trying to make some bandwidth measurements, but the white noise generator software I'm using is far from optimal.

For the moment, it's spectralab build in signal generator, but the spectrum is far from a straight line

Does anyone have a good white noise generator? The attached picture is what I get from spectralab, directly recorded from the wave device (so 100% software, it's not line in shorted to line out) with a 2x averaging

[darkfenriz]

search for a random numers generator.
'twister mercena ' or sowething like this is believed to be best now

[Bricolo]

is this just a random numbers generator or a "true" white noise generator that uses your soundcard output?

[darkfenriz]

MERSENNE TWISTER is just an algorythm.
if you have matlab look here
you can run the function, normalize values to +-1 16-bit and save a vector of values as a wave ( command: wavsave() )

regards

P.S. there doesn't exist 'true' digital white noise generator, all are based on pseudo-random numers generators

[Mr Evil]

I use Cool Edit to generate noise (n.b. Adobe bought it and changed the name). The white noise it makes is about as white as it gets, as per the attached pic. It also does other colours of noise.

[Davey]

If you use the dual-channel capability of Spectralab and make a transfer function measurement it won't matter if the noise source is not absolutely flat. You can even use pink noise.

Cheers,

Davey.

[Bricolo]

can you give us more details?

[darkfenriz]

hi bricolo
what is the length of your sample of WN?
the fft could be not straight because of short wave length
it is stochastic proces- you cannot require flat fft on any length

cool edit is here (oldish version)

regards

[Davey]

What kind of details to you want? A basic setup configuration for testing line-level gear?

Use either the left or right line output from your soundcard and "Y" it....one leg back to one of the line inputs and the other leg to your DUT. The output of your DUT goes to the other line input. Select "real transfer function" from the dual-channel processing box in the options menu and activate the white or pink noise generator. Increase the averaging setting so you get nice smooth curve results.

A similar setup will work for close-miked speaker measurements if you incorporate a microphone and preamp into setup.

Cheers,

Davey.

[paologatto]

Originally posted by Bricolo
I'm trying to make some bandwidth measurements, but the white noise generator software I'm using is far from optimal.

For the moment, it's spectralab build in signal generator, but the spectrum is far from a straight line

Does anyone have a good white noise generator? The attached picture is what I get from spectralab, directly recorded from the wave device (so 100% software, it's not line in shorted to line out) with a 2x averaging

Another way to obtain quite accurate and flat spectra in bandwidth measurements is measuring the impulse response of the system (using MLS signal, that is similar to white noise) and taking its FFT.

This is an example of the FFT of a loopback impulse response of a normal soundcard:






taken from http://purebits.com/appnote6.html


Bye,
Paolo