This is the first evening of real autumn! Today was partly cloudy and fairly warm, maybe 75F – but tonight the temperatures are rapidly dropping and it is quite windy. The low tonight should be in the mid-upper 40s. At the beginning of the evening the seeing was just slightly better than two arcseconds, but now (9:55pm) it is worse than 3 arcseconds. Very clear but very unsteady air. UPDATE – around 10:00pm the seeing started to improve markedly – to around 2 arcseconds. UPDATE – by 10:30pm the seeing had again gone to around 3 arcseconds!
Rig #1 is running another sequence of spectra for the U Montreal campaign – Deneb and PCyg, both at H-alpha, then Deneb at the Si lines. Rig #2 is again shooting a sequence of images of GD552. This may be my last night on this object as Joe Patterson is requesting data from several other targets.
There have been several nights of good observing for which I’ve not kelp a log. I’ve also been working on squashing bugs in the PPX program (previously PhotProc-X). I found a bug in the PPX code that caused signal-to-noise to be very much under estimated. It had to do with the way the weighting mask was being built and resulted, in fact, in the mask becoming almost a straight aperture photometry mask with softer edges. After fixing the bug photometry of the typically fainter program objects improved markedly while the standard deviation of comparison star light curves increased somewhat. Other software such as Maxim and Mira both show lower standard deviations in comparison-minus-check light curves, but PPX’s curves for fainter program objects is clearly “tighter”. That is to be expected as the weighting mask is optimized for the program object and not the (in this case much brighter) comparison stars.
I also had some problems with the star matching program MATCH_STARS. Once I had fixed the PPX bug I went back and re-reduced all of the data I had for GD552 over the past four years. The data I’d acquired in 2011 had been taken using the ST9 camera and on a couple of the nights following meridian flip there was not much overlap between the template and the flipped images. I re-wrote the main loop to allow the program, if it could not get a good set of matched stars, to go back and add another star and re-run the analysis. The user can choose how many iterations are run before it gives up and returns zeros for every one of the fit coefficients. There was also a small bug that caused the program not to properly order the triangle vertices with their corresponding opposite sides. It wasn’t a huge problem but the program now is much more likely to find matched triangles when there are few overlapping stars. Finally, once a good fit is found, all of the stars sent to MATCH_STARS are then fit which greatly increases the number of stars used to create the final transformation coefficients, the hope being that the transforms thus derived come from a larger overlap area than they might given just a few bright stars. Changes in the resulting photometry, thus far, have been small to none.
I’m beginning to have some suspicions regarding how stable the grating micrometer is. For the U Montreal campaign I have to change the grating angle twice during the night. What I think I’ve noticed is that following a change in grating angle it might take a few minutes for things to “settle” – I think I’ve seen the lines move very slightly between two comparison spectra taken only a minute apart. This is something I can easily check – maybe tonight after the final program spectrum.
Out at 1:45 am – nice night – the first night that it’s felt like autumn!