Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

FFT

Submitted by ljmayes on Mon, 10/18/2010 - 16:34

I am puzzled by the implementation of the FFT in VisSim. Normally the ouput of the discrete Fourier transform is a vector of complex numbers (excepting the dc and max frequency term which are always real). The exception is the special case when the input function is odd (when all the tranformed terms are real). In VisSim the transform is unconditionally wholly real and (to my surprise) the IFFT will accept this and reconstruct the original waveform.

Additionally, there is no 'reflection' in the frequency domain.

The 'help' doesn't help.

What is going on?

Find more posts tagged with

Altair Embed

English

Accepted answers

All comments

Altair Forum User

Submitted by Anders89 on Fri, 10/22/2010 - 05:25.

Good question. The fft block is returning a 1 dimensional vector of alternating real and imaginary elements since VisSim 7 does not support complex matrices. You can, however, reshape the vector to a 2xN matrix and feed that to the mag/Phase block which will produce 1xN vectors of magnitude and phase of the fft result of the sum of a 10 hz and 13 hz sin wave. The 128 buffer of the signal is sent to the fft block, reshaped to a 2x64, resulting in a 1x64 magnitude and 1x64 phase plot.
Here is a sample screen shot of the result:

Altair Forum User

Submitted by ljmayes on Mon, 11/08/2010 - 19:38.

This is a reply to Anders89.

Thanks (I didn't get an auto notification that there was a reply, although I had asked for one, so I just bashed away at the problem and have only just seen your comment). I worked out how the FFT block orders the frequency domain data by running though a series of time domain functions and I’ve devised a strategy (albeit a messy one) for sorting out the output vector.

Using the example you give: as there are 128 samples in the time domain then there will need to be 65 components in the frequency domain – the dc component (index 0) and the max frequency component (index 64) are real (i.e. their imaginary parts are always zero). The dc and max f components from the FFT block are the first two numbers in the vector. In my model I have converted the vector to scalars and then applied 'convert' followed by 'magPhase' blocks (as appropriate) and then converted the magnitudes back to a vector (I’m not interested in the phase).

The 'reshape' block almost sorts things out – but the first row is wrong – as it makes a complex number from the dc and max frequency term which need to be kept separate.

In my version of VisSim (7.0A22) the 'magPhase' block won’t accept a vector input anyway.

Lawrence

Altair Forum User

Submitted by Anders89 on Fri, 10/22/2010 - 05:25.

Good question. The fft block is returning a 1 dimensional vector of alternating real and imaginary elements since VisSim 7 does not support complex matrices. You can, however, reshape the vector to a 2xN matrix and feed that to the mag/Phase block which will produce 1xN vectors of magnitude and phase of the fft result of the sum of a 10 hz and 13 hz sin wave. The 128 buffer of the signal is sent to the fft block, reshaped to a 2x64, resulting in a 1x64 magnitude and 1x64 phase plot.
Here is a sample screen shot of the result:

<?xml version="1.0" encoding="UTF-8"?>

Hello ,
I have a single displacement Vs. frequency signal derived by performing certain operations.
How do I convert this back to time domain. If its iift what's the function and curve parameter to enter in Hyergraph (x and y) .
If I have to extract mag and phase out of the single curve how do I do it in hypergraph(Altair)..
Kindly help.
Thanks
Regards
Subbu