AODT for rural terrain coverage: does the EM solver handle multi-ridge diffraction, and how can a student without an NVIDIA GPU try it?

Accelerated Computing Aerial Forum
1

Hi all, I’m Sahi, a high-school student in the NVIDIA 6G Developer Program. I built an open tool that shows, for every public school in Arizona, whether a carrier’s FCC 4G claim there has ever been checked by a speed test. It also checks some claims with terrain physics.

The key finding: 16.1% of schools on tribal land have zero speed tests within 2 km, compared with 0.4% of other schools. So physics is the only way to check those claims.

I cross-checked with Sionna RT on 30 m terrain. It matches free space on line-of-sight links. For links behind ridges, though, it finds paths only at very high ray counts, and they come out a median 44 dB below NTIA Longley-Rice. That fits Sionna RT’s first-order-only diffraction. (Details are in a Sionna GitHub discussion: First-order diffraction on real terrain, measured: shadowed links found only at high ray counts, median 44 dB below ITM · NVlabs/sionna · Discussion #1194 · GitHub)

Two questions:

  1. Does the AODT 1.5 EM engine model higher-order or multi-edge diffraction over terrain, for example for the continental-US coverage runs mentioned in the Developer Day FAQ? Or is it best suited to urban scenes?
  2. I have an Apple-silicon Mac and no NVIDIA GPU. Is there a supported way for program members to run AODT or Sionna on GPU, such as LaunchPad labs, Brev credits or a student program? My budget is about $25 a month.

I’d also welcome any feedback on the tool from people working on rural or tribal coverage.