Skip to content

Troubleshooting

Unexpectedly cropped flux

Increase pixels_per_subaperture, fft_oversampling, or the pyramid pupil separation/output margin. Cropping is reported in wfs_captured_photons_s and is never silently renormalized.

Shack–Hartmann flux exceeds the source rate or spots appear in the wrong place

Versions up to 1.1.0 aliased the lenslet field when the subaperture window in lambda/d (pixels_per_subaperture / spot_sampling_pixels_per_lambda_over_d, at the shortest wavelength) exceeded numerics.pupil_samples_per_lenslet. Current versions refine the lenslet field automatically; see Lenslet-field sampling.

Spots have the wrong scale

Check spot_sampling_pixels_per_lambda_over_d, sensor.wavelength_m, and the physical input.grid_extent_m. OPD is metres, not phase radians; phase inputs must declare their reference wavelength.

Noisy frames differ between runs

Pass an integer seed to expose. Unseeded calls intentionally advance the getframes camera RNG. Inspect frame.truth to separate optical photon-rate changes from detector noise.

LGS image is not elongated

A thin LGS is intentionally point-like. Add lgs_ranges_m and normalized lgs_range_weights to a Shack–Hartmann source. The current approximation uses one mean-range OPD plus geometric elongation; it does not synthesize range-resolved turbulence.