RCWA Backend Parity Report
Generated on 2026-06-11 from the real PixelStack -> RCWA adapter path.
This report runs a small 1.0 um BSI PixelStack through each RCWA backend adapter with the same TE source and wavelength list. It is a backend-health gate, not a converged optical benchmark: the Fourier orders are deliberately small so adapter failures and normalization mismatches are visible quickly.
Executive summary
- 2 of 4 RCWA adapters produced bounded, energy-consistent R/T/A values in this smoke test.
- 1 of 4 adapters met the parity target against
torcwa(max R/T/A delta <= 0.05). - Current result:
torcwais the usable reference path for patterned PixelStack runs; the other adapters need normalization/API work before they should be used for production pixel comparisons.
Low-order smoke test
This report does not claim converged QE. It checks whether adapters can run the same patterned stack and return plausible R/T/A numbers. Converged solver settings belong in the convergence report.
R/T/A spectra
Health and runtime
Backend table
| Backend | physical ok | parity ok | R@550 | T@550 | A@550 | mean QE@550 | energy residual | max delta vs torcwa | runtime s | first warning/error |
|---|---|---|---|---|---|---|---|---|---|---|
| torcwa | yes | yes | 0.0342 | 0.0007 | 0.9652 | 0.0057 | 0 | 0 | 0.9607 | - |
| grcwa | yes | no | 0 | 0 | 1 | 0.9101 | 0 | 0.1137 | 0.05 | - |
| meent/numpy | no | no | 180.8335 | 0.8235 | 0 | 0 | 332.3061 | n/a | 3.7173 | meent: R+T=16.5237 > 1 at λ=0.4500um (numerical instability for multi-layer 2... |
| fmmax | no | no | 0 | 0 | 0 | 0 | 1 | n/a | 0.4479 | fmmax failed at lambda=0.4500um, pol=TE: 'Expansion' object has no attribute ... |
Interpretation
physical okmeans finite, bounded R/T/A values and |R+T+A-1| <= 0.05.parity okmeans physical output and max R/T/A delta againsttorcwa<= 0.05.mean QEis the adapter's photodiode-allocation proxy averaged over the four Bayer pixels. It is useful for detecting broken absorption allocation, not for design decisions at this low order.- A backend can import successfully but still fail this report if the adapter uses stale upstream APIs or incompatible normalization.
Regeneration
powershell
uv run python scripts\generate_rcwa_backend_parity_report.pyGenerated metrics are stored at docs/public/reports/rcwa-backend-parity/rcwa_backend_parity_metrics.json.