TMS570LS1224: Problem using DSP library when concatenating small FFTs

Miguel Amezcua Peregrina

Part Number: TMS570LS1224

I have been trying to use the FFT library (RFFT) to obtain the frequency of a signal. The data is on a buffer with length of 5120.

I tried to use two buffers of 2048 and two of 512, but the result didn't match with the one obtained on Matlab, because of the internal calculations. Its different 10 FFTs calculated to 10 points and concatenate them to only one FFT at 100 points.

I looked in the web and found this:

As you see, a Ntot is added and Nffts.

I tried some modifications on the DSP code (on the CFFT code, since RFFT used the CFFT to calculate everything) but didn't work. When tried to printf everything, the MCU collapses.I decided to make 10 ffts in order to keep an integer number as a factor to multiply (5120/2048=2.5) so decided to run the 10 FFTs and I modified the n2=fftLen to n2=fftLen*10 but the problem persisted....

Don't know if someone has tried to concatenate small FFTs to get a bigger one...

I would really appreciate any help...

Thanks and best regards!!

over 8 years ago

0 Chuck Davenport over 8 years ago

TI__Guru 59540 points

Hello Miguel,

We are looking into your question and will get back with you when we have information to share regarding the trouble you are having.

0 Miguel Amezcua Peregrina over 8 years ago in reply to Chuck Davenport

Genius 4015 points

Hi Chuck

I have been playing for the weekend and did some modifications to the library. those modification were incrementing the size of the tables to adapt them to my needs. But for some reason, Im having troubles when running the app. And I don't know why is that... I don't know if I could send you my codes for you to see what am I missing... I will upload an image of where the codes stops to work

0 Miguel Amezcua Peregrina over 8 years ago in reply to Chuck Davenport

Genius 4015 points

okok

so... for some reason I don't understand, the program aborts in this part:

I put a breakpoint in the i0=i0+1u; line and ran it a couple of times and then i put the breakpoint on twidCoefModifier and thats when it took a time, I paused it and was on the dabort.asm file...

any idea???

best regards

0 Chuck Davenport over 8 years ago in reply to Miguel Amezcua Peregrina

TI__Guru 59540 points

Hi Miguel,

My apologies for the long delay in getting back with you. Most likely, your data abort is from a stack overrun. In regard to the DSP math, I have been trying to find someone that could help with this questions; but, truthfully, this level of DSP math is beyond the scope of this forum and we do not have anyone that can provide assistance. From my understanding, there are independent DSP forums out there that could be more beneficial for this sort of question and I would suggest you give those a try.

0 Miguel Amezcua Peregrina over 8 years ago in reply to Chuck Davenport

Genius 4015 points

Hi Chuck

Today I found a way. Its using a STFT. You window the data, using the window of your preference, apply the window to the data, process the data*window and its works. You just need to make sure the window is a supported length.

Thanks a lot!!

0 Chuck Davenport over 8 years ago in reply to Miguel Amezcua Peregrina

TI__Guru 59540 points

Great to know, Miguel!! Glad you were able to figure it out.

If your able, post the solution as I am sure others would benefit from it.

Arm-based microcontrollers

Arm-based microcontrollers forum

TMS570LS1224: Problem using DSP library when concatenating small FFTs