The Gold Standard of Wavefront Sensing

At a chosen time, a wavefront joins locations of equal optical phase. For an ideal plane wave it is a plane; a focusing beam has a different reference shape. An error must therefore be defined relative to the wavefront the system is meant to produce.

A bright beam can still have a distorted phase pattern. Measuring intensity in one plane does not generally supply the full optical field. Wavefront sensing provides additional constraints or uses a model to infer the missing phase information.

Begin with a drawing of the pupil, propagation direction and intended reference. Those choices determine what ‘flat’, ’tilted’ or ‘corrected’ means.

Contents · Next chapter

Let W represent optical path difference relative to a stated reference. Its units are length. At wavelength λ, the corresponding phase difference is φ = 2πW/λ in radians.

For a teaching value W = 100 nm and λ = 500 nm, W/λ = 0.20 waves and φ = 0.40π ≈ 1.257 radians. At λ = 1,000 nm, the same path difference is 0.10 waves, or about 0.628 radians.

Always report the wavelength when expressing error in waves. A number in nanometres cannot be compared with a number in waves until that conversion is defined. Also distinguish optical path error from a physical surface-height error; their relationship depends on the optical geometry.

Contents · Previous chapter · Next chapter

A Shack–Hartmann sensor uses an array of small lenses to form spots. Relative to a calibrated reference, the displacement of a spot indicates the average wavefront slope over the corresponding lenslet region, under the sensor model.

The recorded data are spot positions and intensities. A wavefront map is a further reconstruction. The distinction matters because the number and arrangement of samples, spot assignment and measurement uncertainty influence what can be inferred.

The research literature on registration and centroid errors shows why illumination and geometry belong in the measurement description. An apparently precise spot centre is not automatically an unbiased slope estimate.

Contents · Previous chapter · Next chapter

In a small-angle lenslet model, local slope α ≈ Δx/f. Let the reference-corrected displacement be 2 micrometres and lenslet focal length be 10 mm. Convert both to metres: α ≈ (2 × 10⁻⁶)/0.010 = 2 × 10⁻⁴ radians.

That is 0.20 milliradians. If this slope were constant over a 2 mm interval in a one-dimensional optical-path model, the path-difference change would be approximately αΔs = 4 × 10⁻⁷ m, or 0.40 micrometres.

The second step adds a constant-slope assumption. A lenslet provides an average over a finite region; it does not reveal every fine variation inside that region. Keep the measured displacement, estimated slope and assumed interpolation as separate steps.

Contents · Previous chapter · Next chapter

Slopes describe changes. Adding the same constant to W everywhere leaves its derivatives unchanged. Consequently, local slope data alone cannot recover that common constant, often called piston, without an additional reference or constraint.

In two dimensions, measured slopes also contain noise and may not integrate into one perfectly consistent surface. Reconstruction methods choose a wavefront using a model, a fitting rule and boundary or reference conditions.

A smooth-looking output can therefore reflect both evidence and imposed constraints. Ask which details the measurements resolve and which arise from the chosen representation. A reconstruction is most useful when those limits are visible.

Contents · Previous chapter · Next chapter

Suppose four equally weighted optical-path samples are 0.20, 0.20, 0 and 0 micrometres. Their mean is 0.10 micrometres. After subtracting that common offset, the residuals are 0.10, 0.10, −0.10 and −0.10 micrometres.

The RMS residual is √[(0.01 + 0.01 + 0.01 + 0.01)/4] = 0.10 micrometres. The RMS of the original values relative to zero is about 0.1414 micrometres. Both calculations are mathematically valid, but they describe different references.

Report whether piston, tilt or other fitted components were removed, and state the pupil and weighting. A lower RMS cannot be interpreted fairly when the definitions differ.

Contents · Previous chapter · Next chapter

A reference spot pattern, detector geometry and lenslet focal length connect image coordinates with slopes. Changes in alignment, illumination or detector response can alter that connection. Repeatability alone does not rule out a stable bias.

A useful check compares the sensor response with an appropriate known or independently characterised change. Also inspect whether saturated, faint or overlapping spots have entered the reconstruction as though they were ordinary samples.

The goal is to locate the first unreliable step. If the spot assignment is wrong, refining a later surface fit will not repair the original measurement.

Contents · Previous chapter · Next chapter

Not every system uses a lenslet array. JWST documentation describes wavefront information retrieved from images with known focus diversity, followed by controlled mirror adjustments. Such phase-retrieval methods constrain an optical model using deliberately informative observations.

This is a different data path from local spot displacements, although both aim to estimate optical error. Compare methods through their measurements, assumptions, sensitivity and ambiguities rather than expecting them to output identical raw data.

Sensing and correction are also distinct. A reliable estimate can guide a correction, but the correction system has its own range, calibration and response.

Contents · Previous chapter · Next chapter

A small numerical RMS does not describe every aspect of image quality. Spatial pattern, pupil coverage and wavelength matter. A sensor can miss structure finer than its effective sampling, and a single summary statistic can hide local features.

A flat reconstructed map does not establish zero error in every unobserved mode. Likewise, a measured tilt does not identify its physical cause uniquely. Several optical or alignment changes can produce similar data.

Write a bounded conclusion: over this pupil, at this wavelength, using this reference and reconstruction, the measured data support this pattern. That sentence is more useful than an unexplained claim of perfect optics.

Contents · Previous chapter · Next chapter

With the same 10 mm focal length, a 1 micrometre displacement gives a slope of 1 × 10⁻⁴ radians. Under the same constant-slope approximation across 2 mm, the path change is 0.20 micrometres. Explain why halving displacement halves this modelled result.

Now add 0.50 micrometres to all four RMS-example samples. The residual RMS after subtracting their mean remains 0.10 micrometres. Show why a common offset disappears from that particular statistic.

For transfer, compare a lenslet sensor with a focus-diverse telescope method. Identify the recorded observations, reference information, fitted model and a plausible ambiguity in each. Keep the evidence path visible even when the instrument changes.

Contents · Previous chapter · Next chapter

Optical foundations

• Optics — connect rays, fields and image formation
• Wave Optics — develop phase and interference reasoning
• Photonics — connect optical fields with detectors and signals
• Measurement — inspect calibration and uncertainty
• Astronomy — place telescope measurements in context

Testing and applications

• Optomechanical Engineering — connect alignment with physical stability
• Optical Manufacturing — understand verification of optical performance
• Optical Engineering — connect specifications with the full system
• Telescope correction mathematics — extend the atmosphere and image-quality question
• Teaching Astronomy — connect scale, light and honest inference

Contents · Previous chapter · Next chapter

Frequently asked questions

Does a slope sensor measure absolute piston? Its slope data alone do not supply a common path offset. Is wavefront sensing always a Shack–Hartmann measurement? No; image-based methods provide other routes. Is a correction command proof of corrected performance? The result needs a suitable verification measurement.

Choose a documented optical test and trace its data path in four sentences: recorded observation, calibration, reconstruction and supported conclusion. Use supplied images or numbers for practice.

Primary research and observatory sources

• Shack–Hartmann registration research — spot displacement and local-slope measurement
• Centroid error from non-uniform illumination — a source of measurement bias
• JWST wavefront sensing and control — image-based inference and mirror alignment

Contents · Previous chapter · Education root