One of the primary limitations to image quality in CMOS arrays is fixed pattern noise (FPN), caused by mismatch in device parameters. The primary means of reducing FPN is correlated double sampling (CDS). The output is sampled before and after placing the photogenerated charge on the readout capacitance, and the difference is ideally proportional to signal charge. FPN is evaluated as a function of irradiance for three CMOS pixels: a charge readout pixel, a voltage readout pixel, and a brightness adaptive pixel. FPN for the charge readout pixel is greatly improved by CDS, but is still limited at low irradiance by non-removable offset mismatches such as pixel and column dark current mismatch. In the voltage readout pixel, CDS converts most offset mismatch sources into "gain-like" mismatch sources due to the nonlinear photodiode capacitance and source follower transfer characteristic. Gain mismatch sources are unaffected by CDS. The brightness adaptive pixel is very similar to the voltage readout pixel, but has superior FPN performance at high irradiance due to the effective reduction of gain error.
This file (317 kb) is the set of foils presented at the Solid State Circuits Technology Workshop on CMOS Imaging Technology.
(last updated: February 16, 1996)
Mailing address
Steven Decker
MIT Room 39-611
Cambridge, MA 02139
Phone: (617) 253-0710
FAX: (617) 253-8806