A 640×512 CMOS image sensor with ultrawide dynamic rangefloating-point pixel-level ADC

摘要

Analysis results demonstrate that multiple sampling can achievenconsistently higher signal-to-noise ratio at equal or higher dynamicnrange than using other image sensor dynamic range enhancement schemesnsuch as well capacity adjusting. Implementing multiple sampling,nhowever, requires much higher readout speeds than can be achieved usingntypical CMOS active pixel sensor (APS). This paper demonstrates, using an640×512 CMOS image sensor with 8-b bit-serial Nyquist ratenanalog-to-digital converter (ADC) per 4 pixels, that pixel-level ADCnenables a highly flexible and efficient implementation of multiplensampling to enhance dynamic range. Since pixel values are available tonthe ADC's at all times, the number and timing of the samples as well asnthe number of bits obtained from each sample can be freely selected andnread out at fast SRAM speeds. By sampling at exponentially increasingnexposure times, pixel values with binary floating-point resolution cannbe obtained. The 640×512 sensor is implemented in 0.35-Μm CMOSntechnology and achieves 10.5×10.5 Μm pixel size at 29% fillnfactor. Characterization techniques and measured quantum efficiency,nsensitivity, ADC transfer curve, and fixed pattern noise are presented.nA scene with measured dynamic range exceeding 10000:1 is sampled ninentimes to obtain an image with dynamic range of 65536:1. Limits onnachievable dynamic range using multiple sampling are presented