A code mistake of Sionna RT

[HongdaXu]

The mistake occurs on the class "FieldCalculator" in, which makes NLOS paths get wrong arrival angles, causing wrong extractions of incident e-field values. The incident e-field values are named as "tgt_patterns" in class "FieldCalculator".
The accurate position is in the loop while dr.hint(active, mode=self.loop_mode, exclude=[wavelength]): in the method "_compute_cir": ki_world = dr.select(active, ko_world, ki_world) should be placed before active &= (depth <= max_depth) & ~next_is_none. The corrected code is shown as follows.

[HongdaXu]

tgt_patterns is directly affected by ki_world: # Receive antenna pattern tgt_patterns = [antenna_pattern_to_world_implicit(tgt_antenna_pattern, tgt_to_world, -ki_world, direction="in") for tgt_antenna_pattern in tgt_antenna_patterns], the original code would get the next active, causing a wrong operation ki_world = ki_world, the right operation should be ki_world = ko_world.

[faycalaa]

Hi @GH518dada ,

Thank you for reporting this issue! It is indeed a bug.
Would you like to open a pull request?

[HongdaXu]

Yes, I’d like to open a pull request. I'll create an issue and open a pull request in NVlabs / sionna-rt.