AWRL6432: TrueGroundSpeed Offline Processing

Hello Santosh,

Thank you for your help. I have tried to program it, similar to how it is described here: AWR1843BOOST: Received signal separation on BPM MIMO

Can you tell me if it is correct? Unfortunately, I have not found the relevant code section in CSS. Is BPM something that is done internally in the hardware accelerator or frontend M3 controller? If there is a MATLAB example on how you do the BPM, that would also be very helpful.

fft_2d = np.fft.fft(rfft_1d, rfft_1d.shape[1], axis=1)
fft_2d_norm = fft_2d / fft_2d.shape[1]
radar_data_2dFFTMag = np.abs(fft_2d_norm) ** 2

# MIMO BPM on 2D FFT data: fft_2d_norm[3,32,129,10,2], 3Rx, 32 Doppler Bins, 129 samples (rfft), 10frames, 2TX

n_rx_chan = radar_data_2dFFTMag.shape[0]  #3RX

# Init BPM Arrays
bpm_even = np.zeros((n_rx_chan, fft_2d_norm.shape[1] // 2, fft_2d_norm.shape[2]), dtype=np.complex64)
bpm_odd = np.zeros((n_rx_chan, fft_2d_norm.shape[1] // 2, fft_2d_norm.shape[2]), dtype=np.complex64)

# Get the BPM MIMO data
for i in range(n_rx_chan):
    bpm_even[i] = fft_2d_norm[i, 0::2, :, Frame_num, 0]  # Even Chirps (Sa = S1 + S2)
    bpm_odd[i] = fft_2d_norm[i, 1::2, :, Frame_num, 0]   # Odd Chirps (Sb = S1 - S2)

# Reconstruct Signal
S1 = (bpm_even + bpm_odd) / 2
S2 = (bpm_even - bpm_odd) / 2
print('shape S1:', S1.shape) #shape S1: (3, 16, 129)
print('shape S2:', S2.shape) #shape S2: (3, 16, 129)

#? How to multiply Values on Virtual Array?
#compRangeBiasAndRxChanPhase 0.0 1.00000 0.00000 -1.00000 0.00000 1.00000 0.00000 -1.00000 0.00000 1.00000 0.00000 -1.00000 0.00000
#S1_corrected = S1 * np.exp(-1j * phase_correction)
#S2_corrected = S2 * np.exp(-1j * phase_correction)

Another question: When I do the calibration with the Industrial Visualizer to get the compRangeBiasAndRxChanPhase values, do I need to be already in Bpm mode? Or do I have to select TDM specifically for the calibration first?Thanks again for the quick help! The answers are always very helpful for me.

best regards Tobias

Hi, there:

For BPM demodulation, there are two methods implemented in SDK L demos. 

1) in the motion and presence detection demo, the BPM demodulation is done at range FFT as shown in the Doxygen documentation (file:///C:/ti/MMWAVE_L_SDK_05_04_00_01/docs/api_guide_xwrL64xx/MOTION_AND_PRESENCE_DETECTION_DEMO.html). 

2) in the mmwave_demo, the BPM demodulation is done at AoA2D DPU.  As explains in the SDK L Doxygen documentation at:  file:///C:/ti/MMWAVE_L_SDK_05_04_00_01/docs/api_guide_xwrL64xx/MMWAVE_DEMO.html.  Because Doppler compensation is needed before BPM demodulation.

Therefore, if you are working on an application that there are significant speed, then you should consider method (2).  Otherwise, you can apply BPM demod at the very beginning of the process.

For your second questions, for phase/gain calibration measurement, if you are running motion and presence detection demo, you can program it as TDM or BPM.    But if you are running mmwave_demo, then you can only programmed as BPM mode.   Again you can search inside doxygen documentation to find these information.

Best,

Zigang