Abstract
Monolithic multi-freeform optical structures are revolutionizing high-end optical systems by integrating multiple complex surfaces into single components. However, their manufacturing and measurement are challenging, requiring simultaneous verification of surface form and relative positioning. This paper proposes a deterministic form-position deflectometric measurement method based on Bayesian multisensor fusion. By integrating calibration priors and a fully probabilistic framework, the method overcomes the inherent absolute positioning limitations of deflectometry. Experimental results demonstrate measurement accuracies of approximately 100 nm for surface forms and a few microns for surface positions.
Introduction
Modern optical systems are increasingly moving toward "monolithic multi-freeform" structures—single pieces of material featuring multiple complex, non-spherical surfaces. These designs are compact and high-performing, making them ideal for advanced aerospace and astronomical instruments. However, measuring these components is notoriously difficult because it requires high-precision data for both the shape (form) of each surface and its absolute position relative to others on the same part.
The Challenge: Slope-Height Ambiguity
A primary tool for measuring such surfaces is Phase Measuring Deflectometry (PMD), valued for its high resolution and low cost. However, PMD suffers from "slope-height ambiguity," where it cannot easily distinguish between a change in surface height and a change in tilt. While excellent for measuring surface form, PMD struggles to determine the absolute position of surfaces relative to a global datum.
Methodology: Bayesian Multisensor Fusion
To address this, researchers developed a multisensor fusion system that integrates three key components:
- Phase Measuring Deflectometry (PMD): To capture high-resolution surface gradients.
- Chromatic Confocal Sensor (CCS): A distance sensor that provides a "rigid constraint" for absolute distance.
- Precision Rotary Table: To facilitate measuring different surfaces from multiple angles.
The core of this approach is a Bayesian multisensor fusion framework. Unlike traditional sequential optimization, which can be prone to numerical bias or local optima, this framework treats calibration and measurement as a state estimation problem. By modeling physical uncertainties and marginalizing calibration priors into the measurement model, the system accounts for its own inherent uncertainties, leading to higher fidelity and preventing overfitting to measurement noise.
Key Results and Insights
The study evaluated the system using a custom-machined monolithic structure with four functional surfaces.
- Surface Form Accuracy: Achieved approximately 100 nm RMS.
- Surface Position Precision: Achieved 3.9 to 9.3 μm Peak-to-Valley.
The success of this method lies in sensor complementarity: PMD is highly sensitive to angles but weak on distance, while the CCS is the exact opposite. Fusing them creates a system with "absolute" vision. Furthermore, using the nominal design as a geometric constraint helps shorten the error propagation chain, ensuring multiple surfaces are positioned correctly relative to a consistent datum.
Conclusion and Future Directions
This Bayesian fusion method provides a robust solution for the deterministic measurement of complex monolithic optics. The most promising direction for future work is in-situ measurement. Integrating this system directly onto single-point diamond-turning machines would allow for real-time measurement and corrective machining without removing the part, significantly increasing manufacturing efficiency for high-precision optics.
Summary updated