Simple and Efficient Method to Eliminate Spike Noise from Spectra Recorded on Charge-Coupled Device Detectors

Abstract

Charge-coupled device (CCD) detectors have several characteristics which are suitable for spectroscopic measurements at very low light levels. The quantum efficiency of a CCD is high over a wide wavelength range, and the dark count rate is greatly reduced by cooling the detector. Further reduction of the dark count can be made by setting the pixel binning frame to the detector area that is actually illuminated. If the detector is sufficiently cooled and the signal readout is performed after appropriate binning, the only significant noise source, aside from inevitable photon shot noise, is the noise produced in the signal readout process. Since the readout noise is independent of the duration of exposure, longer exposure provides a better signal-to-noise (S/N) ratio, in principle. However, the practical limit to exposure time is set by spike noise, which is generated by cosmic rays as well as possible γ- and α-rays emitted from materials around the detector chip. Each quantum of such high-energy rays striking the CCD chip usually generates thousands of electrons in a pixel, while a single photon absorbed by the detector produces less than one electron on the average. Thus the spike noise sometimes obscures the very low light level spectrum. Distinction of spike noise from the true photon signals and reliable elimination of the spikes are required in order to further elongate the exposure time and to achieve a better S/N ratio. Occasionally, data processing with median filters is employed to remove the spikes from spectra recorded on CCD detectors. However, this filtering method is a kind of smoothing and has a possibility of distorting the spectrum, particularly when spectral linewidths are comparable to or less than those of spikes. We have developed a simple and efficient method to eliminate the spike noise without losing the original spectral information. Here, we describe the method and demonstrate its effectiveness in Raman and atomic emission spectroscopy.