A 1600-Channel Retinal Prosthesis SoC with On-Chip Edge Detection and Ambient Light Adaptation

Abstract

This work presents a 1600-channel subretinal prosthesis system-on-a-chip (SoC) that employs edge-based stimulation with adaptive ambient light compensation to improve spatial resolution and ensure stable operation under varying illumination conditions. High-density retinal prostheses often suffer from current dispersion and image saturation caused by simultaneous activation of neighboring pixels. To address these issues, the proposed SoC employs a Retinal Prosthesis Edge Detection (RPED) algorithm that selectively stimulates edge-related pixels, thereby suppressing current spreading and reducing excessive power consumption. In addition, an adaptive ambient light sensing scheme dynamically adjusts the stimulation reference voltage to maintain consistent system performance across a wide range of lighting environments. The system adopts a pulse-frequency modulation (PFM) architecture to directly digitize photodiode currents using charge-stacking pulses, eliminating the need for per-pixel analog-to-digital converters and enabling low-latency, low-power operation in a large-scale pixel array. The digitized signals are processed by an on-chip digital controller that generates charge-balanced biphasic stimulation pulses through integrated current drivers, ensuring safe and reliable stimulation. The proposed 1600-channel SoC is fabricated in a 180 nm CMOS process. Mixed-signal simulations and experimental measurements demonstrate effective edge extraction, accurate charge-balanced pulse generation, and improved image clarity compared to conventional full-pixel activation methods.