Characterizing spatial crosstalk effects in small pixel image sensors

Abstract

The popularity of miniaturized CMOS image sensors in embedded platforms, such as mobile telephones, is driving the move to increasingly small pixel pitches. The resulting pixels suffer from increased sensitivity to microlens misalignment and degradation in crosstalk performance, as a direct result of their reduced pixel size. This paper presents a novel application of pixel scan techniques to characterize microlens misalignment, the effect of microlens misalignment on crosstalk, and crosstalk performance in general. Pixel scans are performed on 2.2μm pitch sensors, under monochromatic light. A series of scans are taken for each device under test, sweeping the incident light across and beyond the visible spectrum. The captured data is remapped from the image space into a pixel space. Analysis of how the scans develop over the course of the spectral sweep provides insight into the primary directional sources of crosstalk. Further processing derives approximations of pixel spectral responses at various microlens misalignments. It is likely that the device under test has its microlens layer misaligned by an unknown amount, which must be corrected for. This misalignment is characterized by identifying common positional offsets between the peaks of in-band channels in the recorded scans. The spectral responses can be then used to estimate the effects of microlens misalignment on colour and crosstalk performance across the imaging array. The techniques detailed in the paper are designed to be run on unmodified product dice and do not require expensive test devices.