• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Subscribers 30 subscribers
  • Views 637 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Original question:

AWR2944EVM: The Maxima Arrays from Advanced Statistics RAM in the sector Local Maxima Engine – Operating Mode Configurat...

AWR2944EVM: AWR2944 threshold calculation and interpolation

Henry Lin
Prodigy 210 points
Part Number: AWR2944EVM
Other Parts Discussed in Thread: AWR2944

Hi expert:

  My company are developing some applications using AWR2944，I have some questions when using SDK mmwave_mcuplus_sdk_04_02_00_03，

as below,

Q1: Why does the dB to linear formula is different between local maximum threshold and dopplerCFAR threshold?

local maximum threshold calculation in dopplerprochwaDDMA.c

/* 2D maximum value offset */ 
scratchVal = round(pow(10.0, (double)(obj->cfarAzimFFTCfg.localMaxCfg.azimThreshold)/20) * (1 << 10));
obj->dopplerAzimHwaCommonConfig.advStatConfig.max2DoffsetDim1 = -scratchVal;
scratchVal = round(pow(10.0, (double)(obj->cfarAzimFFTCfg.localMaxCfg.dopplerThreshold)/20) * (1 << 10));
obj->dopplerAzimHwaCommonConfig.advStatConfig.max2DoffsetDim2 = -scratchVal;

doppler CFAR  threshold calculation in mss_main.c

static uint16_t MmwDemo_convertCfarToLinear(uint16_t codedCfarVal) 
{
uint16_t linearVal;
float dbVal, linVal;

/* dbVal is a float value from 0-100dB. It needs to
be converted to linear scale..
First, recover float dbVal that was encoded in CLI. */
dbVal = (float)(codedCfarVal / MMWDEMO_CFAR_THRESHOLD_ENCODING_FACTOR);

/* Now convert it to linear value */
linVal = log2(pow(10, (dbVal / 20.0))) * (1 << 11) + 0.5;

linearVal = (uint16_t) linVal;
return (linearVal);
}

Q2:In Objectdetection.c  function DPC_ObjDet_estimateXYZ,  why is not necessary to change the value from log2 to linear when doing angle interpolation like other reference design (antilog2)

Q3:Is it possible to add interpolation mechanism in range and doppler calculation?

Thanks for reply.

Lin  Henry

  • 0
    TI__Guru*** 148304 points

    Hi,

    Due to local holidays there will be delays in responding to this question

    Thank you

    Cesar

  • 0
    Prodigy 210 points

    Hi, 

      Any update?

    Lin Henry

  • 0
    TI__Mastermind 20605 points

    Hi Lin,

    Thank you for your patience. Please find my response to your questions:

    Q1) We will get back to you regarding this question by the 17th of November.

    Q2) The performance is similar in both the cases when using log2 or linear.

    Q3) Can you give me more context here regarding what you are trying to achieve? What/How are you trying to interpolate in range or doppler?

    Let me know if you have any further questions.

    Regards,

    Kaushik

  • 0 in reply to Kaushik Gowda
    Prodigy 210 points

    Hi Kaushik,

      For question 3, we need more precise velocity and range results of detected objects,but the SDK for 2944  only output quantization results(index*resolution).

    Lin Henry 

  • 0 in reply to Henry Lin
    TI__Mastermind 20605 points

    Hi Lin,

    Thanks for the clarification. Interpolation should be possible after doppler FFT. During the range FFT, compression takes place, and it will not be possible to perform interpolation. 

    Regards,

    Kaushik

  • 0 in reply to Kaushik Gowda
    TI__Mastermind 20605 points

    Hi Lin,

    We have an update regarding your question 1. Please find the changed db to linear conversion formula :

    doppler CFAR  threshold calculation in mss_main.c

    #define CONST_LOG2_10  (3.3219)

    linVal = (uint32_t)(dbVal/20.0 * CONST_LOG2_10 * 2048.0);

    Regards,

    Kaushik

  • 0 in reply to Kaushik Gowda
    Prodigy 210 points

    Hi Kaushik:

      The changed of doppler CFAR formula is equivalent to above one only take the power tern out of (log2), I am asking the different between local maximum threshold calculation and Doppler CFAR threshold calculation . Can you help me clarify it?

    Lin Henry 

  • +1 in reply to Henry Lin
    TI__Mastermind 20605 points

    Hi Lin,

    Let me rephrase my answer to provide better clarity:

    Local maxima threshold:

    Before:

    scratchVal = round(powf(10.0f, (float)(obj->cfarAzimFFTCfg.localMaxCfg.azimThreshold)* (1.0f/20.0f)) * 1024.0f);

    obj->dopplerAzimHwaCommonConfig.advStatConfig.max2DoffsetDim1 = -scratchVal;

    scratchVal = round(powf(10.0f, (float)(obj->cfarAzimFFTCfg.localMaxCfg.dopplerThreshold)* (1.0f/20.0f)) * 1024.0f);

    obj->dopplerAzimHwaCommonConfig.advStatConfig.max2DoffsetDim2 = -scratchVal;

    Updated:

    #define CONST_LOG2_10 (3.3219f)

    scratchVal = round((float)(obj->cfarAzimFFTCfg.localMaxCfg.azimThreshold)* ((1.0f/20.0f) * CONST_LOG2_10 * 2048.0f));

    obj->dopplerAzimHwaCommonConfig.advStatConfig.max2DoffsetDim1 = -scratchVal;

    scratchVal = round((float)(obj->cfarAzimFFTCfg.localMaxCfg.dopplerThreshold)* ((1.0f/20.0f) * CONST_LOG2_10 * 2048.0f));

    obj->dopplerAzimHwaCommonConfig.advStatConfig.max2DoffsetDim2 = -scratchVal;

    Doppler CFAR threshold:

    Before:

    dbVal = (float)(codedCfarVal / MMWDEMO_CFAR_THRESHOLD_ENCODING_FACTOR);

    /* Now convert it to linear value */
    linVal = (uint32_t)(log2f(pow(10, (float)dbVal/20.0)) * additionalScaling * defaultScaling + 0.5);

    Updated:

    #define CONST_LOG2_10 (3.3219f)

    dbVal = (float)(codedCfarVal / MMWDEMO_CFAR_THRESHOLD_ENCODING_FACTOR);

    /* Now convert it to linear value */
    linVal = (uint32_t)(dbVal/20.0 * CONST_LOG2_10 * 2048.0);

    The dB to linear conversion in both the cases are reduced to the same form in the updated version.

    Thanks & regards,

    Kaushik