topic Re: FFT example on MCUXpresso please in i.MX RT Crossover MCUs

FFT example on MCUXpresso please

ping1 — Wed, 13 Oct 2021 09:36:28 GMT

Dear all

Read this post https://community.nxp.com/t5/i-MX-RT/FFT-Fast-Fourier-Transform-operations-on-i-MXRT-platforms/m-p/1348398 here and try to do an FFT with arm_rfft_fast_f32.c, I am using RT1024. Read the example for Keil project too for some SDK (not included for SDK RT1024) and noticed it doesn't call the init function, so I am a bit confused on the procedure to let the library work, please help .

1. Do I need to call the arm_rfft_fast_init_f32 function first? if so, when ?

2. What is the way to link the library, I put it by link the libarm_cortexM7lfsp_math.a file, I hope that is the correct file for RT1024, please confirm.

3. Any example for MCUXpresso IDE?

Thanks!

Ping

Re: FFT example on MCUXpresso please

Omar_Anguiano — Wed, 20 Oct 2021 19:54:56 GMT

Hello
Hope you are well. I will gladly answer your questions:

1. Yes, I suggest you use the init function before using the FFT operation.

2. You can add the library in the project wizard while you are creating your project:

If your project is already created, you can add it in the SDK management(right-click project -> SDK Management -> Manage SDK Components)

3. You can create a new project including the CMSIS DSP library and copy the source code from the examples listed in the documentation: https://www.keil.com/pack/doc/CMSIS/DSP/html/group__groupExamples.html

Let me know if this is helpful, if you have more questions do not hesitate to ask me.
Best regards,
Omar

Re: FFT example on MCUXpresso please

ping1 — Thu, 21 Oct 2021 08:40:55 GMT

Hi, Omar

Thanks for answering me!

I got the precompiled libarm_cortexM7lfdp.a file from SDK RT1024 shown as below and linked the library into project, I hope that is ok too?

I tried the complex example provided for Keil uVision in MCUXpresso, and it works there - same index is resulted using the same data set. however, when input my own data, the result is not in agree with what I input. I am not sure whether I should use complex FFT or real FFT, the input data set is real anyway in both cases. I am interested in getting power FFT only. So I tried to use fast real one, and it ends up hard fault and also modified some variables has nothing to do with the input ot output of FFT. I wonder whether the real fast FFT function there works. Here is some test sample for a short 128 data FFT, it is 400Hz signal, I get close FFT result from Matlab.

float FLCdata[128] = {

148,94,7,7,129,148,148,7,7,7,

148,148,107,7,7,116,148,147,7,7,

7,148,148,119,7,7,104,148,148,7,

7,7,148,148,132,7,7,91,148,148,

7,7,7,148,148,146,7,7,79,148,

148,7,7,7,148,148,148,7,7,66,

148,148,17,7,7,148,148,148,7,7,

54,148,148,29,7,7,148,148,148,7,

7,42,148,148,42,7,7,148,148,148,

7,7,29,148,148,54,7,7,148,148,

0,0,0,0,0,0,0,0,0,0,

0,0,0,0,0,0,0,0,

};

But I cannot get it from the library using complex function. Try to use the real fast f32 type, cannot get it initialised - failure. Advice please?

Regards!

Ping