How Linear are Typical CCDs ?

Abstract

The typical linearity achievable with CCDs was questioned at the previous ESO Optical Detector Workshop in 1993. This paper describes the efforts at CTIO since then to develop a technique for measuring linearity accurately, and rapidly enough to make the exploration of CCD operating parameters practical. A simple circuit has been adopted which stabilizes the intensity of a pulsed LED. It is compact enough to be installed behind the shutter so that linearity can be monitored at the telescope while the dome lights are on. Methods are described for verifying the linearity and stability of the measurement system itself. These requirements are very stringent, since CCD gain variations have typically been measured to be 0.2% or less over most of the operating range. Methods are proposed for reducing non-linearity further, and for generating each new linearity curve from a single exposure, after having taken several calibration exposures. Several discrepancies have been seen between linearity of point sources and linearity measured with flat illumination: the amplitudes were similar but the gain variations had opposite slopes over the central part of the curve, and a more pronounced signal deficit has been noted at low light levels for point sources, which is equivalent to the subtraction of few electrons from every pixel.