Abstract
The low-light performance of image sensors can be enhanced by designing high-effective conversion gain signal paths from photon conversion sites to chip outputs. This can be achieved by using high-gain in-pixel amplifiers, high-gain column amplifiers, and by reducing the conversion capacitance of photon sensing nodes. However, each of these approaches presents unique challenges and limitations that have restricted their widespread adoption in low-light applications. This paper reviews the use of in-pixel gain amplifiers and their signal chain electronics in high-conversion-gain CMOS image sensors over the past two decades. In-pixel gain amplifiers are classified into different categories according to the type of amplification technique used. Analyses of the column-referred conversion gain and the noise of each topology are presented alongside the different metrics used to characterize CMOS image sensor pixels for low-light imaging applications. The performance metrics of various in-pixel gain amplifiers are compared, providing a framework that highlights the best achieved input-referred noise in CMOS image sensors over the past fifteen years. Furthermore, different trade-offs are examined between optimizing conversion gain, pixel full-well capacity, and input read noise in both voltage and charge domains.