 |
 |
On Mon, 31 May 2010, Edward Chapin wrote: > > So, this led me to a bug in my code... the file that was loading fast was > constructed incorrectly :) (i.e. it should always have been painfully slow to > load in data in the past). The problem is that SCUBA-2 data are acquired with > a system with a very jittery clock... so we need a time lookup table for each > plane (a linear approximation won't do it unfortunately). I think we could get away with doing a linear fit to the time series and creating a WinMap. For the model data the error should be fairly small and we don't use the LutMap when regridding the data. When we overlay time series from different sources if the error is less than 5ms we won't see that in any of the plots even if we zoom in quite a lot. I probably could have created a WinMap in the data acquisition library but it was less work (for me and the CPU) to do the accurate thing. In fact we could even do a linear fit when using sc2concat. -- Tim Jenness JAC software http://www.jach.hawaii.edu/~timj