Characterization and avoidance of in-field defects in solid-state image sensors

As solid-state image sensors become ubiquitous in sensing, control and photography products, their long-term reliability becomes paramount. This thesis experimentally examines the nature of in-field faults and demonstrates two combined hardware-software approaches for detecting and mitigating them. Characterization experiments found that most tested commercial cameras developed hot pixels that c reate image bright spots and degrade dynamic range. Faults appear spatially point-like and uniformly distributed, and they develop continually over time. Silicon displacement damage, induced by terrestrial cosmic rays, is the likely cause. A fault tolerant active pixel sensor is developed to isolate hot defects to a portion of the pixel, enabling software algorithms to correct the faults without sacrificing dynamic range. Experimentally-emulated hot pixels can be corrected within ±5% error. A new statistical software approach is developed to identify and calibrate stuck and abnormal-sensitivity faults from only regular photographs. Monte Carlo simulations verify the detection accuracy in complex environments.