There are two calibration matrices included with UVIS PDS data.  The first is a 
two-dimensional matrix that is of the same spectral and spatial size as the 
data.  In most situations, a single detector readout can be calibrated by 
multiplying by this matrix.  The second calibration matrix included represents 
that which would be required for an equivalent observation with no binning or 
windowing (full detector readout), and therefore always has a dimension of 
1024x64.  The following text describes the situation in which this second 
calibration matrix may be required.

The UVIS FUV detector has pixels whose performance is not well understood.  
These anomalous pixels are distributed across the entire detector.  The 
calibration approach adopted by most of the UVIS team is to eliminate these 
detector elements from analysis by setting calibration values to not-a-number 
(NaN).  Applying the calibration matrix to the data will have the effect of 
forcing these anomalous pixels to also have a value of NaN.  One method of 
dealing with such an irregular array is to interpolate across these spurious 
pixels.

When binning is used, any binned pixel in the calibration matrix will assume 
a value of NaN if any of its constituent pixels were of that value.  In certain 
circumstances of high binning, most (if not all) of the resulting calibration 
matrix will have a value of NaN.  In such situations, a modified approach to 
that outlined above is required to retrieve a meaningful measurement.  One 
method is to interpolate across NaN values in the supplied unbinned 
calibration matrix, bin and window the matrix to match the data, and then 
apply the matrix to arrive at a calibrated data product.  A multiplicative 
adjustment of 1.1 is then required to account for the effect of including the 
anomalous pixels in the data.