EVM430-FR6047: FFT (via DSP Library API ) via LEA with 256 samples instead of 128 does not fit into existing memory

Hi Daniel,

Yes, the LEA RAM size is 4KB. You can see the definition in cmd file. Actually, the LEA need reserve some memory for stack. The size of LEA RAM can be used for data is 0xec8.

I think you can change the FFTsamples to 128 and compile the code to see the map file. In the map file, you can figure out what variables that occupy

the LEA RAM memory.

For your second questions, l'm not an expert of FFT algorithm, but I will consult our expert of FFT and com back to you later.

B.R.

Winter Yu

Hello,

still wondering about the memory issue. Any comment would be helpful.

Greetings

Daniel

Hi Daniel,

Could you provide your project? I would like to check your project on my side.

B.R

Winter

Hi Daniel,

When click reply, there will be "Insert Code, Attach Files and more..." link. Click it and enter a new reply page with "attach file" function. Please see below pictures.

B.R

Winter

Hey Winter,

thank you for the information. The function in this c file is simply to be called instead of USSLibGUIApp_Engine() in the USS_Water_Demo.

Best regards

Daniel

#include "hal.h"
#include "DSPLib.h"



#define FFTsamples     128
DSPLIB_DATA(input, MSP_ALIGN_FFT_Q15(FFTsamples)) _q15 input[FFTsamples];





void mainLoop(void) {

    WDTCTL = WDTPW + WDTHOLD;           //Disable watch dog timer

    hal_uart_Init();

    USS_configureUltrasonicMeasurement(&gUssSWConfig);
    USS_initAlgorithms(&gUssSWConfig);

    USS_Algorithms_Results_fixed_point algResFixed;
    USS_Algorithms_Results alg_results_float;



// Measure ----------------------------------------------------------------------------------------------

    while(1) {

        USS_startUltrasonicMeasurement(&gUssSWConfig, USS_capture_power_mode_low_power_mode_0);



// TI Algorithm -----------------------------------------------------------------------------------------

        USS_runAlgorithmsFixedPoint(&gUssSWConfig, &algResFixed);
        USS_getResultsInFloat(&algResFixed, &alg_results_float);



//         FFT Waveform ----------------------------------------------------------------------------------------- 20.06.18

        // Fill input array with ADC Waveform
        uint16_t* pUPSCap16 = (uint16_t*) (USS_getUPSPtr(&gUssSWConfig));
        uint16_t a = 0;
        while(a < FFTsamples) {
            input[a] = *pUPSCap16;
            pUPSCap16++;
            a++;
        }

        // FFT
        msp_fft_q15_params fftParams;
        fftParams.length = FFTsamples;
        fftParams.bitReverse = true;
        msp_fft_fixed_q15(&fftParams, input);



        __delay_cycles(1600000);                                  // 200 ms

    }

}

Hi Daniel,

I can't know the USSSWLib version you used. Actually, we already adopt Hilbert transformation to get the envelope in our algorithm in USSSWLib version after <USSSWLib  02_10_00_04>. And this algorithm also use LEA. In the newest <USSSWLib  02_10_00_07>, you can choose Lobe tracking or Hilbert Transformation by configuration. So you can compare the performance of two algorithm. Please see the release note for more new features information about USSSWLib.

B.R

Winter Yu

Hi Winter,

thank you for your reply. I am aware of those features since I am already using USSSWLib 02_10_00_07. But actually I want to run both your algorithm as well as mine and the problem is that yours already takes up almost all of the limited LEA-RAM so I cannot use the LEA for my own algorithm. This is about memory management which I clearly don't fully understand yet.

Sorry for the trouble.

Best regards

Daniel