AWR1843: Do not see the nearFieldCfg CLI command in AWR1843 SDK

Near field correction is not implemented in SDK 3.1. Whether you need to apply correction depends on whether the objects you are trying to detect are in the very close range. Most use cases don't need to worry about this. Yes, you can do it manually (although not sure exactly what you are trying to do because correction is in real-time) using the method described in SDK 2.1 but if you are activating both azimuth and elevation, both will need to be corrected, the combined geometry will need to be used which will be more complicated than just the azimuth direction. We have not done such analysis and we don't have any correction data for elevation only or azimuth+elevation (3d) cases.

Thanks for your quick reply.

SDK 2.1 module document has an computed phase error result at the boresight , 400 cm distance has about 2.1 degree phase error. When an radar is installed in the vehicle with fixture in mass production, because the fixture has angle tolerance maybe 2~3 degree tolerance, so we need to do the calibration for each installed radar. Normally we use a plat or corner to do the calibration but we can not put the plat or corner too far  ( about 3 meter ) because the space limitation. so if we do not do the near field correction, we may do the wrong calibration because 300 cm has about 3 degree phase error.  I have questions below:

(1) If the radar is designed by using AWR1642, does near filed correction can solve the above issue?

(2) If (1) can solve the issue, if the radar using AWR1843, how to solve the issue if there is no near field correction function?

1) If you are referring to oob demo's rx gain and phase calibration procedure, then this procedure assumes object is located in far field i.e does not take into account near field geometry.

The assumption in our calibration procedure is that there is sufficient freedom in the setup to place object in far field so we didn't create it for near field. The intention for near field implementation was for actual detections in real-time of objects that are close to the sensor and in practice it only works well over a small near range that is also not too close to the sensor as you may have read in the documentation. Your case seems to be that probably you don't care so much about the actual real-time detection of the near objects accurately compared to the calibration that is in error due to near field but can affect angle accuracy of objects in the scene in the far field. From this point of view, one thing to note is that the physical azimuth angle (theta) error is not the same as the phase error. The azimuth angle error is roughly (this is a rough guess, not measured) inverse sin (phase error / pi / numVirtualAntennas) so if phase error is 3 deg, then azimuth angle error is (in Matlab asind(3/180/8)) or 0.12 deg. If you care about this kind of error (at a max range of say 150 m, it will be equivalent cross-range error of approximately 150*2*sind(0.12/2) = 0.31 m or 31 cm), then you will need to study the current far field implementation and adapt it to the near field geometry.