Biomimetic Retinal Sensor Architecture and CUDA Digital Twin for Objective Evaluation of Display Temporal Dithering (FRC) and Cortical Mach Band Load

Conventional display metrology relies on stationary CMOS frame cameras or single-point photodiodes to evaluate display flicker and spatial image quality. While suitable for characterizing deep Pulse-Width Modulation (PWM) luminance dips, stationary camera sensors fundamentally fail to capture the neuro-ophthalmic load induced by sub-LSB Temporal Frame Rate Control (FRC) dithering and bit-depth quantization banding. A conventional camera sensor integrates photons linearly over a fixed exposure window on a static pixel grid. Conversely, the human visual system processes light through dual parallel retinal pathways (Parvocellular sustained integration and Magnocellular transient differentiation) coupled with horizontal-cell lateral inhibition during continuous oculomotor motion [1, 2, 3]. This paper presents two contributions to display metrology and visual ergonomics: A hardware specification for a Biomimetic Retinal Sensor (Mixed-Signal CMOS-MEMS ASIC) incorporating parallel Parvocellular (P) and Magnocellular (M) analog pixel branches, an on-chip Center-Surround lateral inhibition grid, and an active MEMS oculomotor smooth-pursuit optical stage [8, 9]. A Software Digital Twin implemented in CUDA C++ that emulates this retinal architecture on high-speed video captures or raw display composition framebuffers, computing three objective biophysical metrics: the Magnocellular Flicker Fatigue Index (MFFI), Cortical Mach Band Load (CMBL), and Retinal Dynamic Pursuit Instability (RDPI) [10].