There is a lot that can happen within a few days: after rolling out the first version of our quicklooks for the inspection of the data from the pre-environmental tests before the upcoming calibration campaign, being confronted with real-world data spawned a flurry of changes to the code. First of all there was performance: extracting the track of the laser diode shining onto the detector pixels turned out to be prohibitive for larger datasets. It took a while going through the underlying routine in order a) not only to identify where exactly processing time was burned up but b) come up with an improvement. That the latter was indeed the case became evident from a number of timing runs: with a factor 4-5 speed-up visual inspection of the data from the laser diode scans became feasible. However, as seen from one of the first real-world results are number of additional modifications where in order…

Scanning data quicklook

If the motion of the scanning laser across the plane of the detector no longer is a single smooth motion, but rather a number of subsequent scans (e.g. along one of the detector pixel coordinate axes), the resulting plots contain clusters of data points for which the time-sequence information – at least not without some additional visual encoding – becomes lost. As a first measure to gain back some insight, the data points are now drawn using a color scale following the chronological order. Though already I have seen cases where this still might not be enough to disentangle multiple sub-sets of laser measurements, the colouring greatly improves readability of the plots and understanding the nature of the measurements.