Real-Time frequency analysis of acceleration data

Hi,

I am using Simulink Desktop Real-Time environment for reading a 3-axis acceleration sensor. The data is sampled in 2kHz and contains the magnitude of the acceleration. For each of the axis I would like to perfom some kind of FFT with the last 4 received values and find the dominant frequencies.

One criterium is that the frequencies in each axis should be comparable such that I can find "the" frequency of the measurements over all axis.

Does anyone have an idea on how to achieve this?

Thanks in advance,

Max

EDIT: I should have been clearer: I am working in Desktop Real-Time, but in external mode.

Answer 1

[Jan Houska]

Hi Max,

you can use blocks from the DSP System Toolbox library. There is a Buffer block that packs single data samples to buffers usable for computing the FFT, and then there is a FFT block that actually computes the FFT. Both these blocks are supported with Simulink Desktop Real-Time.

But if the number 4 is not a typo, this is way too short a buffer for FFT. You cannot get any meaningful frequency information from just 4 samples. The buffer must be significantly longer to get any useful results.

Good Luck, Jan

[Maximilian Becker]

Hello Jan,

sorry for commenting this late, but I did not have time to validate your answer sooner.

I should have been more precise regarding the environment I am using, because I am working in "external mode" where these two blocks are not available (due to code generation I guess).

And yes, the 4 should have been a 40, but I now think that this is also too few.

Do you have another idea that might work for me?

Best regards,

Max

[Jan Houska]

Hi Max,

I believe that both the Buffer and FFT blocks do support code generation. Is there any error you are getting when trying to use these blocks in external mode? Please beware that the FFT buffer length should be a power of two for best performance. Code generation may even require the buffer length to be a power of two.

Good Luck, Jan

[Maximilian Becker]

Let's say I did some thing very stupid and you were (of course) right. I tried to use the scope block for displaying it..

I now use this set up:

copied from here https://de.mathworks.com/help/dsp/ug/transform-time-domain-data-into-the-frequency-domain-using-the-fft-block.html .

But the scope stays empty and does not advance in time. Am I missing something?