• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 32 subscribers
  • Views 2464 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:

MMWAVE-SDK: In AoA DPU, is the "2D FFT" the same thing as in Doppler DPU?

MMWAVE-SDK: DopplerDPUHWA and DopplerDPUDSP

yifan zhu
Intellectual 255 points
Part Number: MMWAVE-SDK
Other Parts Discussed in Thread: AWR1843BOOST

Tool/software:

Hi:

I have some issue with understanding HWA processing for doppler frequency. It is clear that when we received a set of FMCW wave, we just perform a 2-D DFT so we get the delay-doppler profile. In TI's document, the first process is to compute RangeFFT, which makes sense that just perform 1D FFT for each virtual antenna and each chirp. 

After the RangeFFT is done, we go to DopplerDPU. In DSP version, it is quite clear that we perform a DFT operation across chirps and then sum the absolute value of the DFT results over all virtual antennas. This process is clear. However, for the HWA version, everything becomes confusing. I don't understand why there is a 2D FFT block and a 16-point FFT blocks in the HWA implementation. And what are the physical meaning of the output from two FFT blocks? I can tell that the first '2D FFT' performs DFT across chirps for each virtual antennas. and to me that should be the end of the this DPU right? So why should we do another DFT process here? If I understand it correctly, the '16-point FFT' is performed over all virtual antennas, which I guess is used for AOA estimation right? So what signal is stored in the Detection Matrix at the end of this DPU? 

Best 

Kevin

  • 0
    TI__Mastermind 22895 points

    Hi,

    Thanks for posting. Could you let us know what device/part you are using, what market and end application you are targeting?

    Thanks again,

    Ajay

  • 0 in reply to Ajay Narayan
    Intellectual 255 points

    Hi:

    I'm using AWR1843Boost. Actually I'm not targeting any specific application now cause I'm still walking through the out-of-box demo and I would like to implement my own physical layer denoising algorithm for some research purposes. 

    Best

    Kevin

  • 0 in reply to yifan zhu
    TI__Mastermind 22630 points

    Hi Kevin,

    You can check C:/ti/mmwave_sdk_03_06_02_00-LTS/docs/mmwave_sdk_module_documentation.html to find documentation of each DPU.

    Regards,

    Samhitha

  • 0 in reply to Samhitha Bobba
    Intellectual 255 points

    Hi Samhitha:

    Yes I did check these documents and they are very helpful, but what I'm looking for is a complete mathematical explanation that matches the description in the document for the DPUs. For example, in the document, it shows what blocks are used in Doppler DPU but there is no information about why the Doppler DPU is designed like that. As I mentioned, I cannot understand why the Doppler DPU uses the '2D-FFT' + '16-point FFT' combination (to obtain the doppler frequency, only a FFT cross chirps should be enough after range process). 

    So if there are some more detailed design document for the DPUs, especially from the bottom, that would be very useful.

    Best

    Kevin

  • +1 in reply to yifan zhu
    TI__Mastermind 22630 points

    Hi Kevin,

    Summation of all virtual antennas is computed for each Doppler bin using FFT. Check the below screenshot from the same document shared earlier.

    Regards,

    Samhitha