Abstract
Low-power image sensors are highly demanded for Internet of Things (IoT), wireless sensor network (WSN), and machine vision of AI. In these applications, multiple imaging modes provide a good flexibility for different environment. Also, pulse-width modulation (PWM) imagers, as compared with traditional image sensor, have the advantages of low-voltage operation and potentially lower power. However, the image quality of PWM imagers are easily degraded by process variation and circuit noise that causes either spatial or temporal noise issues. These reasons motivate the study of PWM pixel for multiple imaging modes in standard 0.18μm CMOS technology. This thesis introduces the motivation at first, and then provides traditional voltage- and time-domain image sensors in second and third part, respectively. The fourth and fifth parts are the main content that demonstrates two major works: (a) a high-dynamic-range (HDR) PWM imager and (b) a linear-response (LR) PWM imager, both include energy harvesting operations. The conclusion and future work will be presented at the end. The first work presents a self-powered HDR CMOS imager with dual-mode operation: HDR imaging mode and energy harvesting (EH) mode. In HDR mode, a dual-exposure extended-counting (DEEC) scheme is proposed with a programmable current-controlled threshold (PCCT) generator, which achieves a high DR of 137dB and a low iFoM of 8.1fJ/pixel·code. In energy harvesting mode, the sensing pixel turns out to be an energy harvesting pixels with an additional global micro solar cell, which generates 455mV and 14μW at 60klux (sunny day) and supports a self-powered imaging operation at 4.1fps. The second work presents a PWM imager with three operation modes which includes an additional high quality LR mode. In LR mode, an adaptive-multiple-sampling (AMS) scheme is developed with dual-slope ramping (DSR), pixel-wise signal level judgment, and n-time multiple sampling to improve the signal-to-noise ratio (SNR) for low illuminance pixels. The reset noise, photon shot noise, and thermal noise of the PWM pixel are considered and analyzed to provide good image quality. The 0.4V linear-response PWM pixel achieves a LR DR of 60.1dB, a non-linearity of +0.36/-0.29%, and a FPN of 0.159%. HDR and EH mode are also preserved to provide monitoring and self-powered capability.