Using FFT in DM6437

Have you seen this wiki note that comes with usage example:

http://processors.wiki.ti.com/index.php/FFT_Implementation_With_No_Data_Scaling

Have you seen the example provided with the DSPLIB (http://focus.ti.com/docs/toolsw/folders/print/sprc265.html)

([INSTALL_DIR]\c64plus\dsplib_v210)

Which FFT kernel are you using?

The common challenge that I have seen is that the input data is not correctly scaled when using the FFT kernel. Note, the o/p of FFT can grow therefore it is important that the expected o/p can be represented by the fixed point data format selected

Regards,

Gagan

Hi Gagan when im trying to build fft example you said in wiki. i have this errors;

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 43:   error: 

               syntax error

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 43:   error: 

               section GROUP_1 not built

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 43:   error: 

               section GROUP_1 not built

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 43:   error: 

               section GROUP_1 not built

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 44:   error: 

               section GROUP_1 not built

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 48: warning: 

               multiple definitions of SECTION named '.bss'

      }>

>> C:\\CCStudio_v3.3\\c64plus\\dsplib_v210\\example\\fft_example\\lnk.cmd, line 48:   error: 

               syntax error

im trying to add dsp lib to linker but i'm taking errors again. what is my wrong? 

Ayhan, I just tried building the code and it builds fine. See below the log

--------------------------  fft_example.pjt - Debug  --------------------------
[gen_twiddle_fft16x16.c] "C:\Program Files\Texas Instruments\C6000 Code Generation Tools 7.0.4\bin\cl6x" -g -fr"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/Debug" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../.." -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x16" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x32" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft32x32" -d"_DEBUG" -mv6400+ -@"../../example/fft_example/Debug.lkf" "gen_twiddle_fft16x16.c"

[gen_twiddle_fft16x32.c] "C:\Program Files\Texas Instruments\C6000 Code Generation Tools 7.0.4\bin\cl6x" -g -fr"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/Debug" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../.." -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x16" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x32" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft32x32" -d"_DEBUG" -mv6400+ -@"../../example/fft_example/Debug.lkf" "gen_twiddle_fft16x32.c"

[gen_twiddle_fft32x32.c] "C:\Program Files\Texas Instruments\C6000 Code Generation Tools 7.0.4\bin\cl6x" -g -fr"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/Debug" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../.." -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x16" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x32" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft32x32" -d"_DEBUG" -mv6400+ -@"../../example/fft_example/Debug.lkf" "gen_twiddle_fft32x32.c"

[fft_example.c] "C:\Program Files\Texas Instruments\C6000 Code Generation Tools 7.0.4\bin\cl6x" -g -fr"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/Debug" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../.." -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x16" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft16x32" -i"C:/CCStudio_v3.3/c64plus/dsplib_v210/example/fft_example/../../src/DSP_fft32x32" -d"_DEBUG" -mv6400+ -@"Debug.lkf" "fft_example.c"

[Linking...] "C:\Program Files\Texas Instruments\C6000 Code Generation Tools 7.0.4\bin\cl6x" -@"Debug.lkf"
<Linking>

Build Complete,
  0 Errors, 0 Warnings, 0 Remarks.

Can you provide complete log for your build step?
What CCS version and Compiler version are you using?

Regards,

Gagan

thanks for your help gagan.

i'm using ccs3.3. i solved the problem chanching cmd file. now i can build too.

but my problem is still fft. i added fft functions to video_preview example. i'm taking all frame's fft like example. but when i compare fft image with matlab fft out image they aren^t same. what do you think my problem is? if you have example for this could you send me?

thanks 

Ayhan, great that you are making progress!
The easiest way for me to explain what is going on is if you can provide a simple example on what you do, what you get and what you expect. Can you do below:

* Provide a small buffer (say 16 samples)
* Suggest what DSP API you call and how
* Provide the result you get
* Provide the result you expect to get

I can then help you understand what is going on.

Regards,
Gagan

My project codes are in my office computer. tomorrow i will send you my source code. i think i'm doing this steps but i'm missing something. if you check my code tomorrow. it will be wonderful.

thanks.

Hi! i'm adding my source code's most important parts. could you check is there a problem?

And i wanna ask you something else why fft's out is short?

Matlab fft function's out is float type. You can say dm6437 works fixed point . so how it be real fft if values only change between 0- 255? am i thinking wrongly?

Thanks!

//main

//Frame sizes for capture and display

int height = 256;

int width = 256;

//FFT arrays

short FFTin[256 * 256 * 2];

short twiddle[2 * 256 *256]

gen_twiddle_fft16x16(twiddle,( 256 * 256));

  while ( status == 0 ) 

{

    FVID_exchange(hGioVpfeCcdc, &frameBuffPtr);

    phaseCorrelation( (void*)(frameBuffPtr->frame.frameBufferPtr), height, width);

    FVID_exchange(hGioVpbeVid0, &frameBuffPtr);

}

void phaseCorrelation(void* currentFrame,int yRows,int xPixels)

{

int i = 0, k = 0; 

FILE *out = NULL;

FILE *in = NULL;

FILE *fft = NULL;

    for( i = 1; i < (256 * 256)*2; i+=2 )

{

*( ( FFTin ) + k) = *( ( (unsigned char*)currentFrame ) + i ) ; 

k++;

    }

   DSP_fft16x16t(twiddle, (256* 256),(short *)FFTin,(short *) FFTin);

for( i = 1; i < (256 * 256)*2; i+=2 )

{

*( ( (unsigned char*)currentFrame ) + i ) = *( ( FFTin) + k); 

k++;

    }

printf("Frame Captured\n");

fflush(stdout);

}

Ayhan :

        Hi!

        How can you get the real fft result ? Were the results same with matlab 's results?

Regards

--Tiny

Hi Tiny.

That time i coulndt realize that fft input array must be complex form: real,imag,real,imag,real,imag....

after making this results will be same with the matlab results but just integer parts. reason of this is fixed poind cpu.

Ayhan:

    Hi!

    I have some other quesitions .After the 'fft32*32' function , the output  data is Q31 . How can I transform it ? And after the transformation ,the results are the same with the matlab' result. I read the sheet, I found the input and output are the Q31 formation. How can I use this function? Should I transform the input data to Q.31 and scale it ? In 'fft32*32',I found the input data need scale N. 

By the way , I use the davinci DM6446.    It's fixed point cpu.

 And are you chinese? I'm a chinese student.So my english is broken . Sorry!

    Regards,

    Tiny

Tiny im using DSP_fft16x32. you dont have scale fft result if you use this fft. if you really need to use 32x32 or 16x16 u have to scale results. 

no im from Turkey. 

Ayhan:

  Nice to meet you!

 I will test fft16*32 . Thanks for your help.

Do you have a mail address? And  I can exchange some problems with you expediently.My mail address  is wguoxd@gmail.com

Regards,

Tiny