The measurement problem
Different optical aberrations can produce similar focal-plane images. A known change to the optical phase—called phase diversity—can help distinguish them. A fixed schedule chooses those changes in advance. An adaptive schedule uses the observations acquired so far to choose what to measure next.
The choice becomes harder when the instrument’s response differs from its calibration. A requested correction may produce a larger or smaller change than expected. We study the effect of that mismatch on measurement selection and the final residual wavefront error.
The experiment
The retained command-response study contains 24 independent base wavefronts, 288 simulated episodes, and 2,016 exposures. Every episode receives seven exposures and 14,000 expected photons. Fixed and adaptive schedules are evaluated with actual command response at 0.9, 1.0, and 1.1 times the nominal calibration.
The optical model uses six low-order aberration modes, monochromatic Fourier propagation at 550 nm, a 32-pixel circular pupil, and Poisson–Gaussian detector noise. The command-response mismatch is constant within each episode. The estimator receives nominal calibration and camera images; the evaluator retains the true response.
Both schedules use settings frozen from the earlier classical selection study. Their photon and exposure budgets match. Their computational budgets differ: the adaptive schedule refits after successive observations, while the fixed schedule fits once.
What we found
| Actual / nominal response | Fixed | Adaptive |
|---|---|---|
| 0.9 | 4.993 nm | 3.442 nm |
| 1.0 | 1.687 nm | 1.920 nm |
| 1.1 | 5.078 nm | 3.497 nm |
Adaptive probing reduced pooled error under both mismatched responses. Fixed diversity retained the pooled advantage under correct calibration. Fixed diversity also performed better in the 10–25 nm initial-aberration stratum even with calibration error. The pooled result therefore cannot justify using the adaptive schedule in every operating condition.
The full analysis reports each stratum and paired differences with descriptive bootstrap intervals. The 24 base wavefronts are the independent units; individual frames and conjugate wavefronts are not counted as independent samples.
What the result establishes
This is a descriptive development study in a static, low-order simulator. It shows that the preferred measurement schedule can change with command-response mismatch and initial aberration. It does not establish a working online selector, held-out controller superiority, or performance on a physical instrument.
The aberration strata use simulator truth, which an instrument would not have. Selecting a schedule in operation requires observable inputs, a new prospective protocol, separate selection and evaluation cohorts, and an explicit cost for additional measurements.
The next questions
Can the measurement history distinguish optical aberration from an incorrect command response? When does another exposure add enough information to justify its cost? When should a system retain a fixed schedule or withhold a correction?
We are working toward comparisons against stronger calibration-aware baselines, tests on previously unseen conditions, and experiments that connect the simulated commands to a measured optical response.