UC Berkeley

In this work, optical, computational, and electronic considerations for high-speed MEMSmicromirror-based spatial light modulators (SLMs) are presented with a focus on holographic systems that utilize these elements for point cloud generation. High-speed 3-D holographic patterning of light into point clouds is a prominent technique in a variety of applications ranging from AR/VR displays and holographic projectors to 3-D printing and biology. All-optical neural interfaces utilizing this technique stand to provide a minimally-invasive pathway to circumvent the fundamental limitations of electrical inter- faces. However, the refresh rate and temporal capabilities of such holographic systems are heavily bottlenecked by both the computationally intensive computer-generated holography (CGH) algorithms used to compute the hologram and the slow-settling phase-modulating elements in the SLMs used to project the hologram. I first present a computationally light CGH algorithm for point cloud patterning that closely matches the performance of state-of-the-art algorithms at 2-6 orders of magnitude faster computation times. Its non-iterative and memory-light architecture allows for CPU- based computation in ms timescales. Fast computation easily lends itself to time-multiplexing- based approaches for target throughput increase and speckle reduction. Experimental verifi- cation confirms the simulation results across SLM formats, target counts, and refresh rates. I then present the analysis, design, and verification of two generations of MEMS mirror- based SLMs. Firstly, a reduced-degree-of-freedom SLM built from high-speed piston-motion micromirrors is discussed. This device consists of an annular array comprising >23000 mi- cromirrors arranged into 32 concentric rings, and a custom-designed driver ASIC capable of correcting for the global process variations of the MEMS fabrication. The array was used in random-access varifocal operation, demonstrating optical functionality. A second- generation family of piston-motion micromirror-based SLMs is presented next, with a path- way to achieving high-degree-of-freedom SLM operation with array sizes of up to 64x64 individually-addressable mirrors. The integration scheme and actuation voltage require- ments for these devices necessitated the design of a second-generation ASIC, which is also presented.