aocore

Shared conventions, conformance checks and optics primitives for the adaptive-optics simulation stack: aobasis, pyturb, getframes, makewfs, solvephase, shmpipeline-ao and pyRTC.

The packages of the stack hand each other OPD maps, images, wind vectors and slopes. aocore makes sure they mean the same thing by them:

• CONVENTIONS.md is the contract: array axes, pixel centres, units, propagation sign and normalization, RMS and Strehl definitions, Zernike orientation, wind and slope conventions, and who owns which primitive.
• aocore.conformance holds executable checks of the contract that every package runs against its own implementation in its test suite.
• The primitives themselves, implemented once:
  • a NumPy/CuPy array backend;
  • anti-aliased and segmented pupils with telescope presets, and sub-pixel translated copies of them (Pupil.shifted) for pupil-shift and misregistration studies;
  • FFT, matrix-Fourier, focal-plane and angular-spectrum propagators with exact adjoints;
  • wavefront metrics (rms, rms_unweighted, rms_tiptilt_removed) and the measured Strehl ratio (strehl_peak);
  • perfect-coronagraph images and azimuthal profiles for contrast curves (perfect_coronagraph_image, azimuthal_profile);
  • least-squares phase unwrapping, fast pixel binning (block_sum, block_mean), periodic sub-pixel shifts (fourier_shift), coordinate grids built on either device, and unit constants;
  • set_threads, one switch for the BLAS, OpenMP, FFT and Numba thread pools, for multiprocessing sweeps.

pip install aocore              # CPU
pip install "aocore[cuda12]"    # + CuPy for CUDA 12.x

import aocore as ac

pupil = ac.Pupil.vlt(128)
prop = ac.FocalPlanePropagator(pupil.shape, pupil.pitch, 1.6e-6, 1.6e-6 / 8 / 2, (64, 64))
print(ac.rms(opd, pupil))  # CONVENTIONS 4.1: weighted, piston removed
print(ac.rms_unweighted(opd, pupil.mask))  # plain quadratic mean, piston included
print(ac.rms_tiptilt_removed(opd, pupil))  # weighted, piston/tip/tilt removed
y, x = ac.coordinate_grid(pupil.shape, pupil.pitch, backend="auto", dtype="float32")

# Raw contrast behind a perfect coronagraph, and its azimuthal profile in lambda/D.
halo = ac.perfect_coronagraph_image(opd, pupil, prop)[0]
radii, contrast = ac.azimuthal_profile(halo, pixel_scale=1 / 2)  # 2 pixels per lambda/D

# In any package's tests:
from aocore import conformance

conformance.check_tilt_direction(
    my_image_from_opd, pupil_shape=(64, 64), pitch=..., pixel_scale=...
)

aocore depends only on NumPy, SciPy and threadpoolctl. CuPy is optional.

License

MIT. Every package in the stack except pyRTC (GPL-3.0-or-later) is MIT, and none of them may copy GPL code (see CONVENTIONS 8.4).