A Study on High-Sensitivity Image Sensors Using Signal Processing Techniques for Noise Reduction

在这项研究中，新的降噪技术的基础上过采样信号处理的高灵敏度图像传感器进行了研究。所提出的技术不需要任何光电荷倍增，这是高灵敏度成像的主要技术。其中一种方法是在非破坏性读出模式下对图像阵列输出进行多次采样。该技术基于像素输出的非破坏性模式中的中间信号电平中的相关性来预测信号分量，并且极大地降低了输出随机噪声。另一种方法使用集成在图像传感器的列处的增益自适应放大器阵列。基于0.25 μ m工艺的实验图像传感器芯片已经开发出来，并在CMOS图像传感器中实现了260 uV的最低噪声水平。为了研究低噪声传感器接口，建立CMOS图像传感器接口电路的噪声分析方法是必不可少的。我们已经建立了一个高增益开关电容放大器的噪声分析方法，适合于集成在列的图像传感器。噪声分析模型包括热噪声和1/f噪声。分析表明，如果在放大器内部节点处采样的噪声分量被抵消，则双级噪声抵消放大器具有极低的噪声电平。

In this, study, new noise reduction techniques based on oversampled signal processing for high-sensitivity image sensors have been investigated. The proposed technique does not require any photo charge multiplication, which is a major technique for highly sensitive imaging. One of the methods uses a multiple sampling of the image array output in non-destructive readout modes. This technique predicts the signal component based on the correlation in the intermediate signal levels in the non-destructive mode of pixel outputs and greatly reduces the output random noise. Another method uses a gain-adaptive amplifier array integrated at the column of the image sensors. An experimental image sensor chip based on 0.25um technology has been developed and the lowest noise level of 260uV in CMOS image sensors is achieved. In order to investigate the low-noise sensor interface, establishment of a noise analysis method for CMOS image sensor interface circuits is indispensable. We have established a noise analysis method for high-gain switched capacitor amplifiers suitable for integrating them at the column of the image sensors. The noise analysis model includes both the thermal and 1/f noises. The analysis shows that a double stage noise-canceling amplifier has extremely low noise level if a noise component sampled at the inside node of the amplifier is cancelled.