At the beginning of each MULTIACCUM observation, when the detector is reset, a
net DC bias with a value of order 11,000 DN is introduced. For details, see Section 7.7.2 of the Instrument Handbook.
The overall bias level is somewhat different within each readout quadrant. On top of this signal is a fixed pattern representing the inherent pixel-to-pixel reset levels. In standard MULTIACCUM processing, a two-step process is used to remove these signals. First, reference pixels are used to remove any changes in the overall bias level that may occur during the exposure. The 5 rows and columns of reference pixels surrounding the IR detector are insensitive to illumination, and are used to measure the bias level of the detector at the time of each readout. During calwf3
processing, the mean measured signal of the reference pixels in each read is calculated and subtracted from all of the science pixels on the detector for that read. This method has the advantage of removing any bias drifts with time. WFC3 ISR 2002-06
details the method of reference pixel bias subtraction. This processing is performed in the BLEVCORR step of calwf3
After this, pixel-to-pixel and quadrant-to-quadrant differences in reset levels are
removed by subtracting the zeroth readout of the exposure from all subsequent readouts. This is performed in the ZOFFCORR step of calwf3
. Bias signal is therefore not a component of any of the calibration reference files, because it is removed using data contained in the science exposure itself.