Dark signal correction for a lukecold frame transfer CCD. Application to the SODISM solar telescope on board the PICARD space mission
When Charge Coupled Devices are used for scientific observations, their dark signal is a hindrance. In their pristine state, most CCD pixels are `cool'; they exhibit low, quasi uniform dark current, which can be estimated and corrected for. In space, after having been hit by an energetic particle, pixels can turn `hot'. They start delivering excessive, less predictable, dark current. The hot pixels need therefore to be flagged so that subsequent analysis may ignore them. The image data of the PICARD SODISM solar telescope (Meftah et al. 2013) require dark signal correction and hot pixel identification. Its frame transfer E2V 42-80 CCD operates at -7°C. Both image and memory zones thus accumulate dark current during, respectively, integration and readout time. These two components must be separated to estimate the dark signal for any observation. This is the purpose of the Dark Signal Model presented in this paper. The dark signal time series of every pixel is processed by the Unbalanced Haar Technique (Fryzlewicz 2007) in order to timestamp when its dark signal is expected to change. In-between those instants, both components are assumed constant and a robust linear regression vs. integration time provides first estimates and a quality coefficient. The latter serves to assign definitive estimates. Our model is part of the SODISM Level 1 data production scheme. To check its reliability, we verify on dark frames that it leaves a negligible residual bias (5 e-), and generates a small RMS error (25 e- rms). The cool pixel level is found to be 4 e-/pxl/s, in agreement with the predicted value. The emergence rate of hot pixels is investigated too. It legitimates a threshold criterion at 50 e-/pxl/s. The growth rate is found to be 4% of the frame area per year. Aspects of the method (adaptation of the Unbalanced Haar Technique, dedicated robust linear regression) have a generic interest.