Fewer Paths Returned By PathSolver

[Mohkared]

Hi Sionna Team,

I really appreciate the work the Sionna team has done on the v1.1.0 release. I appreciate all the work you’ve put into it.

I'm observing a sharp decrease in the number of propagation paths returned by PathSolver in Sionna v1.1.0, compared to v0.19.1. The result is a much lower channel impulse response (CIR) energy, reflected in reduced a, tau, and delay spread values.

This is Sionna 0.19.1:

This is Sionna 1.1.0:

Also, I have noticed that the rx and tx in Sionna v1.1.0 has to be very close to get some extra paths:

[faycalaa]

Hi @Mohkared,

It is surprising that you get no paths with the second setup. Here are some sanity checks you could try:

• The object in your scene seems to have a fairly high amount of primitives, and diffracted paths could be calculated on the wedges connecting the primitives with Sionna 0.19. As Sionna 1 does not currently support diffraction, you could try to disable it with Sionna 0.19 for a fair comparison.
• Diffuse reflections are enabled for the setup with Sionna 0.19 (scattering=True) and the first setup with Sionna 1.1. It is surprising that no diffusely reflected paths are found for the latter. Did you ensure that the scattering coefficient of the material making the object is set to a value > 0? The third setup has diffuse reflection disabled.

[Mohkared]

Thank you @faycalaa for the reply.

For the first point, I have disabled it in Sionna 0.19 and the paths got a little bit fewer but still it is more than Sionna v1.1

For the second point, both scenes have scattering coefficient equal to zero. Also, I have changed the scattering coefficient for the scene in Sionna v1.1, but the result is still the same!

For the last thing mentioned in the second point. I have enabled the diffuse reflection but it still does not produce any paths.

[Mohkared]

I have noticed that when I change the scattering coefficient and enable the diffuse_reflection, I got a huge number of paths:

But still, why there is no direct specular reflection. Is there something I might be missing in the configuration that would prevent specular paths from being computed?

[faycalaa]

Sionna 0.19 generates diffusely reflected paths even when the scattering coefficient is set to 0. However, these paths should carry no energy (the path coefficients should equal 0) if the scattering coefficient of the reflector is set to 0. This is not the case with Sionna 1, which does not generate diffusely reflected paths when the scattering coefficient is set to 0, even if diffuse reflection is enabled. Could this be what you observe in the first figure, i.e., diffusely reflected paths carrying no energy? Assuming the scattering coefficient is set to 0, a quick check would be to look at the path coefficients you get with Sionna 0.19 and see if they equal 0, or too look at their types. Another option would be to disable diffuse reflection with both Sionna 0.19 and 1 and see if you observe a difference in the output.

[Mohkared]

Hi @faycalaa,

Thank you for the clear and insightful explanation. It really helped clarify what I’m observing.

[hafezmg48]

@faycalaa @Mohkared This does not yet explain why in the sionna v1.1 the channel path energy are much lower than the sionna v0.19. I am getting about 10 times lower energy. Thanks for the discussion.

[faycalaa]

Hi @hafezmg48,

Could you provide a minimal code snippet that illustrates the issue you are referring to?

[hafezmg48]

Dear @faycalaa
Sorry for late reply. I created a new issue here because this question was already answered.
#970