UC Berkeley

The dynamic patterning of 3D optical point clouds has emerged as a key enabling technology in volumetric processing across several applications, including biological microscopy, material processing, and near-eye displays. Yet patterning throughput is doubly constrained by the computation times required for computer-generated holography (CGH) and the switching times of spatial light modulator (SLM) arrays. This work seeks to alleviate the electronic drive and computation burdens associated with high-speed, real-time, point cloud CGH operation through judicious, application-driven allocations of the space-bandwidth product and precision available for modulation.I first demonstrate the utility of fast-settling piston-motion micromirrors as SLM pixels by successfully designing and operating a circular micromirror array intended for dwell-capable, random-access focus tuning under CMOS-compatible drive. The array was annularly partitioned into distinctly addressable rings to achieve 2π phase shifting at optical wavelengths <1200 nm and refresh rates >15 kHz under 32-channel, <30 V drive. I subsequently present a scalable SLM architecture that improves micromirror planarity, stability, fill factor, and uniformity through expansions to the fabrication process. The updated architecture was designed to accommodate compact chip-scale driver integration for both co-wired SLM arrays intended for primitive optical modulation and full-fledged SLMs capable of single-pixel addressing. Finally, I present a fast, non-iterative, 3D point cloud CGH algorithm that optimally allocates available degrees of freedom across target points for maximal efficiency and employs lightweight deterministic phase calculations under a patchwork approach to hologram construction. The algorithm’s speed advantage relative to existing iterative algorithms was evaluated in simulation and experiments and was found to increase with SLM pixel format, reaching >100,000x at formats as low as 512x512. The algorithm was also found to make optimal use of time-averaged operation, lending itself to ongoing improvements to spatiotemporal SLM bandwidth.