At SPIE Photonics West 2026 in San Francisco, Phaseform presented a machine-learning approach to controlling the Deformable Phase Plate (DPP), the optofluidic wavefront modulator at the core of Delta 7. The talk was given on 20 January 2026 in the MOEMS and Miniaturized Systems XXV conference; the proceedings paper appeared on 4 March 2026.
The problem
A DPP is an electrostatically actuated optofluidic wavefront modulator: electrodes deform a fluid-filled membrane to shape the wavefront that passes through. Conventional open-loop control assumes that the device responds linearly to its control signals. That assumption breaks down at high strokes, where electrode gap variations and fluidic coupling between actuators introduce nonlinear effects. The result is a gap between the wavefront that was requested and the wavefront the device actually produces.
The approach
The team trained a gradient boosting regression model on a dataset of interferometrically measured Zernike-mode wavefronts. The model learns the inverse mapping from a desired wavefront shape to the control signals that produce it, nonlinearities included. With this model in the loop, the DPP reaches a residual wavefront error below 20 nm RMS in open-loop operation.
For users of the Delta 7, the practical consequence is better control fidelity and robustness without a wavefront sensor: the shape you ask for is the shape you get, also at large amplitudes.
Reference: S. Klykov, P. Rajaeipour, L. Fixl, Ç. Ataman, S. M. Weber, "Machine-learning-based open-loop control of an optofluidic deformable phase plate," Proc. SPIE 13907, MOEMS and Miniaturized Systems XXV, 13907-14 (2026). doi.org/10.1117/12.3080509
See the Delta 7 product page for the wavefront modulators this control method runs on.
