Noise PSD

Altair Forum User
Altair Forum User
Altair Employee
edited October 2020 in Community Q&A

Submitted by John Haine on Tue, 03/15/2011 - 15:14 

The Gaussian Noise block allows one to set the mean and standard deviation of the noise, but it isn't clear what the power spectral density should be. Can one assume for example that if the iteration time is 10 microseconds, the effective sampling rate is 100 kHz, and the noise bandwidth would be 50 kHz? Should one assume a flat spectrum up to that limit or a sinc spectrum?

Thanks, John.

Tagged:

Answers

  • Altair Forum User
    Altair Forum User
    Altair Employee
    edited June 2016

    Submitted by Anders89 on Tue, 03/15/2011 - 20:55.

    John, the Gaussian block gives a flat spectrum. You can check this simply by plotting the gaussian and then checking the 'Frequency Domain' option on the plot.

  • Altair Forum User
    Altair Forum User
    Altair Employee
    edited June 2016

    Submitted by John Haine on Tue, 03/15/2011 - 21:57.

    Thanks. So should I assume flat to 50 kHz in my example? So the power spectral density for an rms value of 1, i.e. varaiance 1 as well, would be 1/5e-4 = 2e-5/Hz?

    John Haine

  • Altair Forum User
    Altair Forum User
    Altair Employee
    edited June 2016

    Submitted by Anders89 on Wed, 03/16/2011 - 04:05.

    Hi John, if you integrate the square of the gaussian block with mean=0, stdDev=1 at 100Khz for 1 sec you get .9999, pretty close to your rms value of 1, and the FFT gives a noise band centered roughly on 2e-5. Looks like it is working as you expect.

  • Altair Forum User
    Altair Forum User
    Altair Employee
    edited June 2016

    Submitted by John Haine on Wed, 03/16/2011 - 14:34.

    Anders, excellent, thank you.

    By the way, I downloaded the 6.0D update last night and it fixes the iteration count issue, so thanks again for your help - I can now run simulations for as long as I like (or my PC gives up).

    John.

    John Haine

Categories