At Henriette’s very successful Master’s thesis defense yesterday a few good questions were asked by the opponents and the audience. Some of them we were prepared for, but a few were novel:
1) A question about the quality of FFs was a little mixed up, but the essence was ‘are they good because they are almost overexposed’? Now that we know about the CCD nonlinearity which starts even at 15000 counts and is not obviously in the ‘blooming phase’ until you pass 55000 we might consider the impact of non-linearity on various quality estimates again. However, the high quality of the hohlraum FFs is due to the hohlraum being a good uniform source.
2) If we had a way to rotate the CCD we could indeed investigate field gradients. This need only be done once. As far as I know the CCD stops its rotation because it is now held by alignment of two steel pins. Let us ask Ben to comment on whether a 180 degree turn of the camera is possible in the present setup.
3) After which interval does lunar libration and sunrise on the Moon cause so many changes in the illumination of the lunar surface that stacking images is pointless? This can be investigated with our forward model system. A long time ago I tested that and found that half an hour was the limit, but this should be redone and focused on areas of the Moon – there are obvious changes near the terminator after juts a few minutes but what is the effect near the disc edge, and so on. It takes some 3-4 minutes to get two bias frames and a stack of 100 images.
Added later: A quick investigation using synthetic images of the quarter Moon suggest that the maximum change in a single pixels’ intensity occurs on the terminator and is roughly 1% per minute. Away from the terminator – in typical DS and BS pixel positions – the change in DS/BS ratio is like 0.03% per minute. This is thus an issue mainly for BBSO, who do not expose DS and BS simultaneously.
4) What would be the error due to alignment problems alone? We can investigate that with the forward model.
5) There is no WCS in the file headers. Could alignment be aided with such a system? WCS is the World Coordinate System – this is what we can find by using astrometry.net – the pointing of the telescope is found with sub arcsecond precision and written into the FITS header. We may be able to do that if there are stars in the frame – since we are using coAdd mode at the moment it seems impossible to get stars in the frame except occasionally (e.g. tau Tauri).
6) In the precision error budget independence is assumed – how much de-pendence is OK for the budget still to be valid? This is a good question about statistics fundamentals! We sought to separate the dependent and independent errors in two budgets, but in general the question can be handled with formal methods or simulations where synthetic examples are investigated.
7) Scale the bias frame better by following the cooler periodicity. This was countered by a warning: Beware that the periodicity seen may depend on camera load.
The counter was good because the nice periodicity is obtained when taking long sequences of bias-only images, right? Under observing conditions the heat load on the camera due to taking science frames may be different – and may vary depending on exposure time and number of frames taken. We can investigate this by studying bias frames from regular sequences of image-taking: are there signs of a periodic bias level in those images? We know the period and we know the time of exposure so it should be possible to perform a signal analysis. A worst-case, ‘upper limit’ analysis could be performed by simply studying how much the mean value of scaled superbiases varies from regular observing sequences: If the variability of the mean level is more than that seen in the bias-only sequences then we know that we have additional heat load on the cooler during science frame taking.