Analysis and Simulation of Low-Light INL in CMOS Image Sensors
This paper describes some sources of low-light nonlinearity in CMOS image sensors. With the number of ADCs increasing to read larger pixel arrays at the higher frame rates, disturbances on shared nets can cause nonlinearity. Traditional sources of nonlinearity are well-known with corresponding techniques to correct them during post-processing, though linearity degradation on the low end of the signal range can be more difficult to correct. Furthermore, scene-dependent low-light nonlinearity is even more of a challenge, as the required corrections now need to be modulated based on what may be happening elsewhere in the image. With smaller routing pitch and a larger number of ADCs, array effects are an increasingly important source of nonlinearity. These array effects typically have a scene-dependent component. When single-slope ADC architecture is used, ADCs convert at the same time range for the same signal level and can potentially cause crosstalk as an adverse array artifact. This crosstalk is more severe in low-light conditions as the shot noise is insignificant. Hence, array effects cannot be masked. In this paper, we explain some mechanisms that can contribute to low-light INL and offer plausible solutions. A method to simulate low-light nonlinearity is also presented, and a comparison is made to show improvements.