• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 31 subscribers
  • Views 2671 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.

IWR6843: Understanding Doppler Estimation using Capon beamforming weights

Attiya Mahmood
Intellectual 575 points
Part Number: IWR6843

Hi,

I am replicating the DSP chain of people counting demo using the raw ADC data. I have understood and implemented the range FFT, Capon beamforming and CFAR-CASO in both range and angle domain.

I am trying to better understand on how beamforming weights are being applied before doppler is computed. From the C code it looks like to me that beamforming weights are w = A* inv(Rn), where A is the steering vector (1x8) corresponding to the angular estimate of target and Rn is the 8x8 covariance matrix. Next, the beamforming weights 'w' are multiplied point by point with the range FFT data corresponding to different antennas. In other words beamforming weights are allowing to combine data from 8 receive antennas to make one data point before the doppler FFT is applied across chirps. Is my understanding correct? Also can you point me to any documentation (apart from the code) which explains this better?

Also, I did not see peak grouping implementation in the demo after CFAR-CASO?

Attiya

  • 0
    TI__Mastermind 28055 points

    Hi Attiya,

    First, to answer the easy questions:

    1. Signal chain order of People Counting is Range -> Static Clutter Removal -> Angle of Arrival (Capon) -> CFAR -> Doppler Estimation.  So you are correct that the 8 Rx antennas data is combined before doppler calculation.
    2. We do not use any peak grouping in this demo. After the signal chain, the point cloud is passed to the tracker, which tries to find clusters in the point cloud, so more points is better.

    Regarding the Capon calculation, I have attached a 3 slide power point that gives the rough idea of what the signal chain is doing.  You can also read the TI design

    Regards,

    Justin

    CaponSlides.pptx.

  • 0 in reply to Justin Curewitz
    Intellectual 575 points

    Hi Justin,

    Thank you for the prompt response.

    I want to understand conceptually why you are multiplying beamforming weights with the range FFT data to get the doppler estimates. Why is it needed?

    Regards,

    Attiya 

  • 0 in reply to Attiya Mahmood
    TI__Mastermind 28055 points

    Hi Attiya,

    I am looping in an expert to provide an answer.

    regards,

    Justin

  • 0 in reply to Attiya Mahmood
    TI__Mastermind 40335 points

    Hi, Attiya:

    It is possible that for the same range bin, we have several target, and each target may have different angle and Doppler speed.  

    Traditionally, we apply Doppler FFT first, and for each detected (range, Doppler) peak, we apply angle of arrival (DOA) estimation.    

    Here, we calculate range-angle spectrum first, then for each detected (range, angle) peak, we apply Doppler FFT to estimate the speed.  Apply the beamform weight of the peak angle is the way to get to (range, peakAngle) information.    

    Let me know whether it makes sense to you.

    Zigang

  • 0 in reply to zigang Yang
    Intellectual 575 points

    Hi Zigang,

    Thank you for the reply. I think I have understood it and it makes sense to me.

    Regards,

    Attiya