10 November 2026

Best-Fit Optical Axis: Choosing the Right Weighting Strategy

Understand ordinary, weighted, and orthogonal least-squares methods and learn how weighting choices affect best-fit optical axis analysis for lens assemblies.

In a multi-lens optical system, individual surfaces may have different amounts of decentration and tilt. Centration measurement provides the decentration data for each surface. The next problem is how to combine those data points into one reference axis that represents the complete optical system.

This is the role of best-fit optical-axis calculation.

Centration measurement software with surface data

Finding the optical system’s principal axis

In a multi-lens system, each optical surface may be decentered or tilted to a different degree. Centration measurement supplies surface-level decentration data, but combining those measurements into a single system reference requires best-fit-axis analysis.

What methods can be used to fit the optical axis?

1. Ordinary least squares (OLS)

OLS determines the fitted line by minimizing the sum of squared residuals between the measured points and the line. All data points receive the same weight, making this method suitable for reducing the squared residual sum under uniform assumptions.

2. Weighted least squares (WLS)

WLS assigns weights according to the characteristics of different measurements, changing the influence of each data point on the fitted result.

3. Orthogonal least squares / total least squares (TLS)

This method considers the perpendicular distance from each measured point to the fitted line. It is suitable when errors in multiple coordinate directions must be considered together.

The selected method depends on the measurement-error model and the fitting objective.

Weighting is the counterweight on the balance

Consider a system with multiple optical surfaces, each requiring decentration measurement. Measurement conditions and measurement accuracy are not necessarily equal across surfaces. With equal weighting, every surface has the same influence on the fitted axis. Once differentiated weights are applied, the calculated axis changes accordingly.

Weight allocation is therefore how the engineer controls the contribution of different measurement points to the final optical axis.

Multiple weighting choices, multiple engineering positions

Weighting type Basic principle Typical application
Equal weighting All measurement points receive the same weight Measurement accuracy is similar and sampling is uniform
Inverse-variance weighting Smaller measurement variance produces greater weight A defined statistical error model is available
Inverse-radius weighting Smaller radius of curvature produces greater weight Surfaces with substantially different radii participate in the fit
Manual weighting The engineer assigns weights according to the measurement objective Specific surfaces or positions have special significance
Robust weighting Reduces the influence of abnormal measurements Outliers or anomalous measurement data may be present

Why is there no completely unified weighting standard?

Weighting options in best-fit optical axis software

The measured objects differ: single lenses, lens assemblies, infrared optics, and large-aperture optical systems operate under different conditions.

Measurement objectives also differ: some users focus on the overall system axis, some prioritize a specific surface, and others need an assembly-alignment reference.

Measurement principle, accuracy, sampling, and data-processing methods also produce different error characteristics. Weight selection should therefore match the measured object, measurement uncertainty, and engineering requirement.

The best-fit axis is not a property of the data alone; it is a property of the data plus the question being asked.

Best-fit-axis analysis is only one evaluation metric for lens assembly. Its exact influence on final imaging performance is difficult to quantify. This article therefore presents common weighting strategies only; analysis of complex lens groups containing planar surfaces is outside its scope.

Probes and light beams: two different ways to measure

1. Contact measurement reference

Lens assembly measured on a precision rotary stage

Definition: Contact measurement collects coordinate or dimensional data through direct contact between a probe and the workpiece surface, then uses those measurements to establish a reference datum.

Common references: the outer diameter and mechanical locating faces of an optical element, the lens barrel or lens mount, and other locating datum surfaces.

Advantages: direct acquisition of mechanical geometry and position data; suitable for parts with clear mechanical datums; convenient for analyzing assembly and positioning relationships.

Limitations: probe force may damage thin-walled, easily deformed, or surface-sensitive optical elements; complex curved or special optical surfaces are difficult to measure; results can be affected by probe shape, contact point, and measuring force.

2. Non-contact measurement reference

Precision stage and sensor setup for lens alignment

Definition: Non-contact measurement uses optical sensing, reflection, or transmission to acquire measurement data without touching the optical surface.

Common optical-centration methods: reflected measurement of a spherical surface’s center of curvature; transmitted measurement of beam or imaging characteristics; optical inspection of the axis features of aspheric and other rotationally symmetric surfaces.

Advantages: no contact damage; direct collection of optical-feature information; suitable for surfaces that cannot tolerate contact; useful for analyzing the optical-center relationships of single lenses, cemented lenses, and multi-element assemblies.

Limitations: requirements on the optical properties and clear aperture of the measured surface; results can be affected by surface quality, optical parameters, and environmental factors; different configurations vary in material compatibility, surface suitability, and measurement range.

In optical autocollimation, for example, the reflected image from a spherical surface can determine the position of its center of curvature relative to the instrument’s rotation axis, allowing the centration error of the optical surface to be analyzed.

Contact vs. non-contact reference

Comparison Contact Non-contact
Data source Mechanical surface geometry Optical or surface-geometry features
Measurement method Direct probe contact Optical sensing and related methods
Primary reference Outer diameter, end face, locating surface, etc. Center of curvature, optical axis, surface axis, etc.
Main purpose Mechanical positioning and geometric analysis Optical centration and optical-axis analysis
Selection basis Mechanical structure and assembly requirements Optical properties and measurement requirements

Conclusion

The best-fit axis is generated by fitting multiple sets of optical-surface data. The fitting method and weight allocation both change this invisible axis. For different optical structures, workpieces, and assembly objectives, the weighting strategy must also change.

The method should be selected from the actual requirement, with suitable weights and a matching measurement reference.

Want to build a repeatable multi-element alignment workflow? Read our guide to aligning a lens assembly to the best-fit optical axis, or compare Clotho centration systems.

Frequently asked questions

Does weighting always improve best-fit-axis accuracy?

No. Weighting improves the result only when the weights reflect a valid error model or a clearly defined assembly objective. Arbitrary weights can make the axis less representative.

Why use inverse-radius weighting?

Surfaces with smaller radii can have different measurement sensitivity and positional influence in an assembly. Inverse-radius weighting can increase their contribution when that better reflects the engineering objective, but it should not be applied automatically.

Should the best-fit axis replace surface-level tolerances?

No. The best-fit axis is an aggregate alignment metric. Individual surface tolerances, mechanical datums, and final image-performance checks remain necessary.