As opposed to the random sources of noise which are temporal, Fixed Pattern Noise (FPN) is a spatial noise and refers to device mismatches in pixels and colour filters, variations in column amplifiers and processing elements. FPN is present in both CMOS and CCD cameras. However CCD image sensors only suffer from pixel FPN due to spatial variation in photo-detector geometry and dark current, neither the CCDs nor the output amplifier (which is shared by all pixels) cause FPN. Thus, the FPN associated with CCD cameras is much lower than that of CMOS cameras.
As previously mentioned, CMOS sensors come in two ‘flavours’ Active Pixel Sensors (APS) and Passive Pixel Sensors (PPS), APS suffers from higher pixel FPN than PPS but PPS generally suffers from higher column FPN as shown in the figure below.
FPN is typically reported as the standard deviation of the spatial variation in pixel outputs under uniform illumination (not including temporal noise) as a % of voltage swing (or well capacity) - values of < .1% to > 4% have been reported.
A technique commonly used to reduce FPN is Correlated Double Sampling (CDS) and is used in many analogue circuits to reduce offsets and reset noise. The output is sampled twice, once immediately after reset and the second time with the signal present. Taking the difference eliminates offset and reset noise. CDS only reduces offset FPN (gain FPN cannot be reduced using CDS). The figure below shows the improvement in FPNdue to CDS.