| As the continuous trend of scaling in standard CMOS process, CMOS image sensor is more and more easily affected by the noise and the crosstalk exist in the ciruit. While digital image sensor is a kind of sensor that achieve analog to digital converting within the pixel, which can largely alleviate the drawbacks of the scaling process. There are three main ways to build a digital image sensor, which are pulse width modulation,pulse freqrency modulation and address event representation. Pulse width modulation can transfrom information from voltage-domain to time-domain, at the same time maintain proper power consumption and circuit area. However, limitation of dynamic range exists in conventional PWM readout sensors constrained by in-pixel memory. The detection of lower incident light is significantly affected if the integration time is limited.This paper presents a modified pulse width modulation(PWM) readout method for image sensors with block-based self-adjusted reference voltage to extend its dynamic range. In this structure, pixel array is divided into blocks with the same size, and each block contains a reference voltage generator to obtain the reference voltage based on the incident light within the block. Dynamic range, photo response, distortion, and noise are carefully analyzed in theory. Pixel array of 64×64 pixels are designed under 65 nm CMOS process, including in-pixel comparator, in-pixel memory, reference voltage generator, 8-bit single slope ADC and etc. It can double the dynamic range from 48 dB of typical PWM digital image sensors to 96 dB in theory, and the simulation shows it can achieve dynamic range of 88.16 d B. Peak signal noise ratio(PSNR) indicating the distortion of the scene is improved to 28.9d B PSNR on average, better than 20.11 dB with fixed reference voltage method and 17.25 dB with ramp reference voltage method. Comparisons between outputs and theory values are made under high and low luminance, and the errors between them are analyzed. |
PDF Full Text Request