The CCDs used at Keck can be read in one-amp or two-amp mode. Often observers may choose to observe in one-amp mode simply because it is simpler in concept, and despite the higher overheads from longer readout times.

However, two-amp readout is actually relatively simple to understand and reduce. The lower readout times can be particularly advantageous during calibrations and for science programs that involve significant numbers of short-exposure images.

First, realize that the two amps will have slightly different bias levels and gains. In general these are fairly well-matched, but the eye will still see a discontinuity between the two halves of the image in either bias images or images with significant background flux (flats or sky images). These two effects come out readily in data reduction. The difference in DC offset or bias level is removed by subtracting either bias frames or the overscan region. (More on the latter in a bit!) This leaves a difference in gain, manifest by a different number of electronis per DN. This is readily removed when the image is divided by a flat field, which is also subject to the same gain difference. An image which has been bias-subtracted and flat-fielded properly will show no discontinuity across the boundary between different halves of the CCD.

The keywords written into the FITS header provide the information for defining the prescan, postscan, and data regions in the image:

PREPIX	# of prescan columns per amplifier
NAXIS1	total number of columns
NAXIS2	total number of rows
POSTPIX	# of postscan columns per amplifier
NUMAMPS	# of amps in use

Missing from this list is the designation of the starting row and column. This can be dependent on the data reduction package being used, usually being 0 or 1. (E.g. IDL is zero-based.) Let's use SC and SR for these values.

The two data regions are given by:

start col 1 = SC + NUMAMPS*PREPIX
start col 2 = start col 1 + NAXIS1/NUMAMPS

# cols = NAXIS1/NUMAMPS

The bias regions are given by:

start col 1 = SC + NAXIS1 + NUMAMPS*PREPIX
start col 2 = start col 1 + POSTPIX

# cols = POSTPIX

These can be readily coded into any data reduction program.