Study on consistency driving method of stitching pixel array based on self-adaptive correction technique

With the application of stitching technology in CMOS image sensors of large arrays, the traditional clock tree synchronization design method is not suitable for the two-side drive circuit of pixel arrays, resulting in the serious problems of DC penetration and bad row of the two-side drive of pixel arrays in the same row. In this paper, in terms of the idea of self-adaptive correction on chip, a consistency driving method which can be applied to stitching pixel arrays is proposed. The method can be adapted to the change of chip working environment, and has the characteristics of simple structure and high reliability. Based on 55 nm technology, the present method is applied and verified in a large array CMOS image sensor with 150M pixel size and 77 mm×84 mm chip area of 12 288×12 288. The experimental results show that, under a master clock of 500 MHz and a row clock of 125 kHz, the inconsistency of bilateral row drive is reduced from 17.5 ns to less than 2 ns (one clock period), and the consistency is improved by above than 9 times, ensuring that the frame frequency of the super-array image sensor with a scale of 100 million pixels can reach more than 10 frames. 随着拼接工艺在超大阵列CMOS图像传感器中的应用, 传统的时钟树同步设计方法已不再适用于像素阵列的双侧驱动电路中, 同一行像素阵列的双侧驱动面临直流穿通和坏行的严重问题。基于片上自适应校正思想提出了一种可应用于拼接型像素阵列的一致性驱动方法。该方法根据芯片工作环境的变化进行自适应实时校正, 具有结构简单和可靠性高等特点。基于55 nm工艺, 采用所提方法在一款12 288×12 288的150M像素规模、芯片面积高达77 mm×84 mm的超大阵列CMOS图像传感器中进行了实际应用和全面验证。实验结果表明：在500 MHz的主时钟、125 kHz的行时钟下, 双侧行驱动的非一致性由17.5 ns减小至2 ns(一个时钟周期)以内, 一致性提高了9倍以上, 确保了亿级像素规模的超大面阵图像传感器的帧频达到10帧以上。