Computing the time information on a row-by-row basis proved to be very inefficient, as the OAL functions were optimised for vector inputs and constant CCD number. The rows were therefore arranged into into blocks of constant CCD number (in the case of PN, this was derived from the CCDID and QUADRANT columns), and the OAL functions were applied to each block. The computed time values were then mapped onto the correct rows of the modified table.
This resulted in an 100-fold (approx) improvement in efficiency.