A Digital CCD Noise Reduction Technique Experimentally Tested on a Large Batch of Scientific Sensors

Abstract

In this article, a digital optimal-filtering method is proposed for reducing readout noise in charge-coupled devices (CCDs). The technique includes a number of noise sources that other methods omit but may restrict the sensor's effectiveness. A procedure is described to obtain the noise directly from the video signal under a standard readout sequence, instead of using an indirect noise measurements approach that does not capture all the interference sources. The proposed research method underwent extensive experimental testing on a large number of scientific sensors, reducing noise in comparison to standard techniques. A significant improvement in the performance comes from using a nonstationary noise model for the video signal and the calibration of the dynamics of the charge transfer signal for each sensor. The average reduction in the noise standard deviation of all tested devices was approximately 40%. Moreover, the spread of the standard deviation values was reduced compared to the standard technique. These two aspects indicate that this technique may increase the selection yield for the construction of multiple-sensor instruments.