SPT DOA

[iceice]

HI,

S32R274 and S32R294 rsdk only have dbf algorithm for Angle estimation. Can other algorithms with higher accuracy be supported? Is there any relevant demo for reference? Thank you

[GaryRK]

You are welcome.

The beamNum is the number of steering vectors that are applied in the DBF operation, basically it means that the field of view is split into 64 beams and angle of arrival will estimate which beam the peak is in. E.g. if FoV = 120° then DoA resolution as determined by DBF is 120 / 64 = 1.875°. Please note that the antenna design physically determines best achievable DoA resolution and accuracy beyond the signal processing technique.

The FoV is measured with broadside at 0° so an FoV of e.g. 60° means DoA estimates in range from -30° to 30°. A target at +40° therefore lies at -30° + 40° = 10° when FoV is set to 60°. 

[iceice · Accepted solution]

Thank you. When FOV is set to ±30°, is there any way to eliminate the 40° target that would otherwise be checked as 10°, causing a measurement error

[GaryRK]

No problem. To mitigate aliasing like this you could use a more aggressive time domain window function to reduce reflection power in the sidelobes, or make a new antenna design that is suited to the narrow FoV (also to reduce power in sidelobes at the wide angles you are not interested in)

[GaryRK]

Hello iceice,

Yes other algorithms for DoA estimation can be supported on S32R274 and S32R294. If the algorithm requires operations not supported SPT then there is the option to use the e200z7 cores, for added performance here you can consider using the PowerArch SPE2.0 (Signal Processing Extension) instructions which are supported on e200z7 cores.

I'm afraid there is no reference code specifically ported to S32R cores outside of the RSDK examples. However there are many examples of algorithms available in public web with C code implementations. These can ported to e200z7 core and evaluated.

[iceice]

Thank you for your answer！

What does beamNum = 64 mean in dbf algorithm? How can I modify this value.

In rsdk demo, fov can be configured as 120°, 90° and 60°. If I set fov as 60°, a real target is at +40°, while the target output by dbf algorithm is at 10°, but fov of 120° can be correctly calculated at 40°. May I ask why?