CCD Reductions

The bias level was subtracted from each frame, using the average of several CCD bias frames. Cosmic rays were denoted by hand and removed by nearest-neighbor interpolation. Bad columns were also denoted by hand and then removed by sinc interpolation across each row of the bad column. The sinc function was selected for use in the interpolation because of its favorable Fourier properties: its Fourier transform is the top-hat function, implying even sampling of the image noise structure over a finite range of frequency.

A flatfield image was constructed for our data by summing a series of frames of a white source, scanning the etalon through a free spectral range. Any first-order slope in the illumination pattern was removed with a two-dimensional linear fit. (Sky flats were not available.) The flatfield was then divided by a smoothed ($\sigma \sim 3$ pixels) version of itself, normalized, and divided into each data and calibration frame. Due to problems with on-chip binning, the movement of transient defects between the images of each cube, and temperature fluctuations of the CCD, the flatfielding process did not substantially alter the noise content of most frames. Fortunately, the pixel-to-pixel sensitivity variations comprise a noise source at a level of only $\sim$1.6%, a small value compared to other Fabry-Perot photometry uncertainties.