TI中文支持网Part Number:AWR1843
代码中的res->angleFftIn[txAntIdx * DPParams->numRxAntennas + rxAntIdx]是否就是存储用于测角的各天线的原始相位呢？代码来源是industrialtoolbox4_11_0中traffic monitoring的aoaprocdsp.c
static inline uint32_t AoAProcDSP_processMultipleVirtualAntennas
(AOADspObj*aoaDspObj,uint32_tnumObjsIn,uint32_t*numObjsOut,volatile uint32_t*waitTime
)
{uint16_t idx, rxAntIdx, rangeIdx, txAntIdx, dopplerIdx;uint16_t nextTransferRxIdx, nextTransferRangeIdx, nextTransferTxIdx;DPU_AoAProcDSP_HW_Resources *res;DPU_AoAProcDSP_StaticConfig *DPParams;cmplx16ImRe_t *inpDoppFftBuf;cmplx32ReIm_t elevationSymbols[SYS_COMMON_NUM_RX_CHANNEL];uint32_t numObjsOutLocal = 0;volatile uint32_t startTimeWait;int32_t retVal = 0;cmplx16ImRe_t *radarCubeBase;uint32_t pingPongIdx, nextTransferIdx;uint8_t channel;cmplx32ReIm_t *windowingBuf;cmplx32ReIm_t *azimuthFftOut;cmplx32ReIm_t *twoDfftOut;/* for extMaxVelocity feature: */cmplx32ReIm_t *hypothesesSymbols;cmplx32ReIm_t *azimuthFftOutTemp;int32_tdopplerSignIdx;int32_thypIdx, maxHypIdx;float32_t maxPow, maxPowTemp;int32_twrapStartInd;cmplx32ReIm_t *angleFftIn_copy;res = &aoaDspObj->res;DPParams = &aoaDspObj->params;radarCubeBase = (cmplx16ImRe_t *)res->radarCube.data;/* Setup overlayed buffers */windowingBuf = (cmplx32ReIm_t *)res->scratch1Buff;azimuthFftOut = (cmplx32ReIm_t *)res->scratch1Buff;twoDfftOut= (cmplx32ReIm_t *)res->scratch2Buff;/* Setup overlayed buffers for extMaxVelocity feature */hypothesesSymbols = (cmplx32ReIm_t *)res->scratch2Buff;azimuthFftOutTemp = res->elevationFftOut;/* Angle estimation for each object */for (idx = 0; idx < numObjsIn; idx++){rangeIdx= res->cfarRngDopSnrList[idx].rangeIdx;dopplerIdx = res->cfarRngDopSnrList[idx].dopplerIdx;/* Reset ping/pong index */pingPongIdx = DPU_AOAPROCDSP_PING_IDX;/* Trigger first DMA. First transfer is for [txAntIdx=0, rxAntIdx=0, rangeIdx=(Obj range from CFAR list)].Note: EDMA ping/pong scheme must support #TX antennas = 1,2,3 #RX antennas = 2,4 */retVal = EDMA_setSourceAddress(res->edmaHandle,res->edmaPing.channel,(uint32_t) &radarCubeBase[rangeIdx]);if (retVal != 0){goto exit;}EDMA_startDmaTransfer(res->edmaHandle, res->edmaPing.channel);for (rxAntIdx = 0; rxAntIdx < DPParams->numRxAntennas; rxAntIdx++){for (txAntIdx = 0; txAntIdx < DPParams->numTxAntennas; txAntIdx++){/* verify that previous DMA has completed */startTimeWait = Cycleprofiler_getTimeStamp();retVal = AoAProcDSP_waitInData (res, pingPongIdx);if(retVal != 0){goto exit;}*waitTime += Cycleprofiler_getTimeStamp() - startTimeWait;/*Find index in radar cube for next EDMA.*/nextTransferTxIdx= txAntIdx + 1;nextTransferRxIdx= rxAntIdx;nextTransferRangeIdx = rangeIdx;if(nextTransferTxIdx == DPParams->numTxAntennas){nextTransferTxIdx = 0;nextTransferRxIdx++;if(nextTransferRxIdx == DPParams->numRxAntennas){nextTransferRxIdx = 0;nextTransferRangeIdx++;}}nextTransferIdx = (nextTransferTxIdx * DPParams->numRxAntennas * DPParams->numDopplerChirps +nextTransferRxIdx) * DPParams->numRangeBins + nextTransferRangeIdx;/*Last computation happens when nextTransferRangeIdx reaches (rangeIdx + 1) was we are processingonly one range index. Therefore, do not trigger next EDMA.*/if(nextTransferRangeIdx == rangeIdx){/* kick off next DMA */if (pingPongIdx == DPU_AOAPROCDSP_PONG_IDX){channel = res->edmaPing.channel;}else{channel = res->edmaPong.channel;}retVal = EDMA_setSourceAddress(res->edmaHandle, channel,(uint32_t) &radarCubeBase[nextTransferIdx]);if (retVal != 0){goto exit;}EDMA_startDmaTransfer(res->edmaHandle, channel);}inpDoppFftBuf = (cmplx16ImRe_t *) &res->pingPongBuf[pingPongIdx * DPParams->numDopplerChirps];/* Remove static clutter? */if (aoaDspObj->dynLocalCfg.staticClutterCfg.isEnabled){AoAProcDSP_clutterRemoval(DPParams, inpDoppFftBuf);}/* The windowing output buffer is the input buffer for the Doppler FFT.It needs to be padded with zeroes if number of Doppler chirps is less thanthe number of Doppler bins.*/if(DPParams->numDopplerBins > DPParams->numDopplerChirps){memset((void*)(&windowingBuf[DPParams->numDopplerChirps]), 0,(DPParams->numDopplerBins - DPParams->numDopplerChirps) * sizeof(cmplx32ReIm_t));}/*Up to this point samples are in ImRe format, which is the radar cube format.The windowing function will do IQ swap and the Doppler FFT is computed inReIm format.*/mmwavelib_windowing16x32_IQswap((int16_t *) inpDoppFftBuf,res->windowCoeff,(int32_t *) windowingBuf,DPParams->numDopplerChirps);/*From this point forward samples are in ReIm format. In particular, FFT, twiddle are all in ReIm format.*//*DSPLIB notes for DSP_fft32x32():Minimum FFT size is 16. Size must be power of 2.No scaling done in FFT computation.*/DSP_fft32x32((int32_t *)res->twiddle32x32,DPParams->numDopplerBins,(int32_t *)windowingBuf,(int32_t *)twoDfftOut);/* Save the 2D doppler FFT output for the required doppler bin */res->angleFftIn[txAntIdx * DPParams->numRxAntennas + rxAntIdx] = twoDfftOut[dopplerIdx];angleFftIn_copy[txAntIdx * DPParams->numRxAntennas + rxAntIdx] = twoDfftOut[dopplerIdx];pingPongIdx ^= 1;}/*txAntIdx*/}
Shine：
请参考下面的帖子。https://www.ti2k.com/wp-content/uploads/ti2k/DeyiSupport_传感器_3391796
