High real-time surface shape control based on FPGA hardware acceleration

In high-precision optical systems, the surface shape accuracy of components significantly influences system performance. Major engineering projects require nanoscale precision on optical surfaces, approaching the limits of existing deterministic machining technologies. While deformable mirror (DM) technology achieves nanoscale surface control, conventional Central Processing Unit (CPU)-based control algorithms are constrained by serial instruction mechanism and jitter, failing to meet high real-time requirements. To address this, we propose a real-time surface shape control method based on Field Programmable Gate Array (FPGA). The FPGA control system employs a subregion row-level ping-pong buffer architecture, enabling time overlap between data transmission and computation through alternating read-write operations between dual-buffer. Subregions are partitioned based on actuator array distribution, with non-sequential data access via address jumps to compute averages. A reference surface is fitted, and the tilt component is eliminated through differential operation. The global Peak-to-Valley (PV) value is calculated to evaluate compliance with target specifications. For error correction, proportional-integral (PI) control algorithm is used to compute control voltages. Experimental validation was conducted on a self-developed high-precision surface shape adjusting device equipped with a piezoelectric ceramic actuator array. Results demonstrate that for surface shape data with 600×600 resolution, a single computation cycle requires only 390.1μs, achieving a ~200× improvement in real-time performance compared to software-based solution. This method leverages FPGA’s parallel computing advantages, offering a high-speed, high-precision surface shape control solution for applications such as planar optical elements, demonstrating significant engineering value.