Linear pixels like CCD and CMOS A PS, have a limited dynamic range of 2-3 decades, while wide dynamic range logarithmic pixels have reduced sensitivity at low light. Pixels, which integrate the photo-generated charge for low intensity and use a logarithmic load for high intensity, combine the better features of the two types of pixel. However, device mismatch between pixels and the linear-logarithmic response combine to create several sources of fixed pattern noise. In this paper a device-physics based model of the pixel response is used to characterize these sources of fixed pattern noise and study their impact on the output image produced from pixels manufactured in a typical CMOS process.