• State TI Thinks Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 5 answers
  • Subscribers 30 subscribers
  • Views 616 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.

IWR6843AOPEVM: Multiple points on a line, same azimuth/elevation

Morten Nissov
Intellectual 380 points
Part Number: IWR6843AOPEVM
Other Parts Discussed in Thread: IWR6843

I think I have an understanding problem.

For the IWR6834AOPEVM, if all antennas are used then the azimuth/elevation resolution should be 30 deg, given this sensor has 120x120 FOV this means that the FOV is discretized in a 5x5 grid right?

I was curious if it is possible for multiple points to have the same value for azimuth/elevation but different values of range and doppler. If the previous statement is correct I feel this must be the case, because the point clouds certainly don't seem to max out at 16 points. But intuitively two reflection at the same azimuth/elevation should mean that one is shadowed by the other right?

Hopefully you can shed some light on this.

Just to be sure, angular resolution in this context does mean that a point's azimuth and elevation has to be one of the following values:

  • -60
  • -30
  • 0
  • 30
  • 60
  • 0
    TI__Mastermind 37805 points

    Hi Morten,

    We have more than 5 candidate angles for a target to be at. Using the difference in phase across the antennas, we can localize the angle of radar returns to a much finer degree. In practice, we compare the radar returns of the target with the expected radar returns at many different angles, and select the angle that maximizes this comparison.

    The resolution you're mentioning here refers to the ability of the radar to separate two targets in the same range bin. Since we have four antennas in the azimuth, then if there are two targets spaced less than 30 degrees apart in the same range bin, then we will not necessarily be able to separate them. However, in practice, targets occupy many range bins, which helps us distinguish between them.

    Best,

    Nate

  • 0 in reply to Nathan Block
    Intellectual 380 points

    But when it says "range" or "velocity' resolution this is actually the bin size right? It's also the smallest difference which is distinguishable, but my understanding for range and velocity was that this was also the quantization, if that makes sense. Is the radar cube processing described in more detail somehwere?

    Regarding azimuth and elevation, is there an expected accuracy of the measurement written anywhere?

    Back to the original question as well, I suppose this azimuth/elevation is calculated for each range-doppler slice of the radar cube right? Which means that each of these cube slices has the possibility for a peak in the same azimuth-elevation bin, assuming the object associated with the first reflection isn't blocking too much those behind.

  • 0 in reply to Morten Nissov
    TI__Mastermind 37805 points
    Morten Nissov said:
    But when it says "range" or "velocity' resolution this is actually the bin size right?

    Assuming that the number of bins is a multiple of two then yes. 

    Morten Nissov said:
    Is the radar cube processing described in more detail somehwere?

    Fundamentals of Radar - https://www.ti.com/lit/wp/spyy005a/spyy005a.pdf?ts=1678461466105

    IWR6843 Out-of-Box Demo Documentation : C:/ti/mmwave_sdk_03_06_00_00-LTS/packages/ti/demo/xwr68xx/mmw/docs/doxygen/html/index.html

    Processing Flow is found in the index.html files here : C:\ti\mmwave_sdk_03_06_00_00-LTS\packages\ti\datapath\dpc\dpu

    Morten Nissov said:

    Regarding azimuth and elevation, is there an expected accuracy of the measurement written anywhere?

    If you're asking how accurately we can measure the angle of an object with respect to the radar, you will find in the fundamentals document that it varies with the angle the object is at (better accuracy at lower angles).

    Morten Nissov said:
    Back to the original question as well, I suppose this azimuth/elevation is calculated for each range-doppler slice of the radar cube right? Which means that each of these cube slices has the possibility for a peak in the same azimuth-elevation bin, assuming the object associated with the first reflection isn't blocking too much those behind.

    I believe this is possible based off the document here

    C:/ti/mmwave_sdk_03_06_00_00-LTS/packages/ti/datapath/dpc/dpu/cfarcaproc/docs/doxygen/html/index.html

    Best,

    Nate 

  • 0 in reply to Nathan Block
    Intellectual 380 points

    Actually you're right, the range/doppler quantization is really:

    (max value - min value) / total number of bins

    right? I think this is how I had been understanding resolution, but if this is true I had been thinking resolution was also the quantization. Which I suppose isn't necessarily true.

    If you're asking how accurately we can measure the angle of an object with respect to the radar, you will find in the fundamentals document that it varies with the angle the object is at (better accuracy at lower angles).

    Is there a way to express this quantitatively for a given radar, like is ranges from x to y degrees uncertainty from boresight to the edge of the FOV? If nothing else I suppose we can comment on the quantization right? As this results from the FFT, as far as I could read. From what I understand angle estimation uses the same number of bins as doppler right?

    believe this is possible based off the document here

    C:/ti/mmwave_sdk_03_06_00_00-LTS/packages/ti/datapath/dpc/dpu/cfarcaproc/docs/doxygen/html/index.html

    This link didn't mention azimuth or elevation. I was thinking mostly since these are calculated in bins in an FFT (I think) then it should be possible to have two points at two distinct range bins with the same azimuth/elevation right?

  • 0 in reply to Morten Nissov
    TI__Mastermind 37805 points
    Morten Nissov said:

    Actually you're right, the range/doppler quantization is really:

    (max value - min value) / total number of bins

    right

    Yes but sometimes there is some interpolation on top of that to artificially increase the resolution.

    Morten Nissov said:
    Is there a way to express this quantitatively for a given radar, like is ranges from x to y degrees uncertainty from boresight to the edge of the FOV?

    See this formula from the fundamentals doc I linked above.

    Morten Nissov said:
    If nothing else I suppose we can comment on the quantization right?

    The quantization will depend on the number of bins in the angle FFT, which is configurable in the cfg file.

    Morten Nissov said:
    From what I understand angle estimation uses the same number of bins as doppler right?

    This is not accurate. The number of bins in the angle FFT is configurable in the cfg file.

    Morten Nissov said:
    This link didn't mention azimuth or elevation. I was thinking mostly since these are calculated in bins in an FFT (I think) then it should be possible to have two points at two distinct range bins with the same azimuth/elevation right?

    This links shows the input to the angle FFT, which is described in the index.html file under AoAProc. It should be possible to have two points at the same location with different velocities.

    C:/ti/mmwave_sdk_03_06_00_00-LTS/packages/ti/datapath/dpc/dpu/aoaproc/docs/doxygen/html/index.html

  • 0 in reply to Nathan Block
    Intellectual 380 points

    Yes but sometimes there is some interpolation on top of that to artificially increase the resolution.

    this is based on which binary I use right?

    This is not accurate. The number of bins in the angle FFT is configurable in the cfg file.

    which cfg line affects this? I'm looking at the SDK user guide and can't see what it should be. I know frameCfg controls range and dopple bins.

    This links shows the input to the angle FFT, which is described in the index.html file under AoAProc. It should be possible to have two points at the same location with different velocities.

    That's not really what I was asking. Can you have two points with the same azimuth and elevation but at different ranges? So essentially two points (or any number >1 really) lying on a line intersecting at the radar origin.

  • 0 in reply to Morten Nissov
    TI__Mastermind 37805 points
    Morten Nissov said:
    Can you have two points with the same azimuth and elevation but at different ranges? So essentially two points (or any number >1 really) lying on a line intersecting at the radar origin.

    Yes

    Morten Nissov said:

    which cfg line affects this? I'm looking at the SDK user guide and can't see what it should be. I know frameCfg controls range and dopple bins.

    This is configured in the aoa DPU in either aoaprochwa.h or aoaprocdsp.h. Its default value is 64.

  • 0 in reply to Nathan Block
    Intellectual 380 points

    This is configured in the aoa DPU in either aoaprochwa.h or aoaprocdsp.h. Its default value is 64.

    Meaning quantization is ~2deg without considering any inter-bin interpolation. So the configure this value I need to recompile and reflash the binary?

    As a followup to

    Can you have two points with the same azimuth and elevation but at different ranges? So essentially two points (or any number >1 really) lying on a line intersecting at the radar origin.

    The source for this happening would primarily be two return with minor differences in angle which get lumped in the same bin by FFT right? I suppose the first object only partially blocking the chirp such that it reflects off the second is fairly unlikely.

  • 0 in reply to Morten Nissov
    TI__Mastermind 37805 points
    Morten Nissov said:
    So the configure this value I need to recompile and reflash the binary?

    Yes.

    Morten Nissov said:
    The source for this happening would primarily be two return with minor differences in angle which get lumped in the same bin by FFT right

    Two returns with minor differences in angle may get lumped into the same angle FFT bins.

    Best,

    Nate