Being finished with the first step of data reduction, I'm now going back through each image and looking to see how good of a job the routines did of subtracting all the stars. When I first started playing with these programs, it was catching about 96% of the light from the stars, which means that there would be a 4% systematic error in the results. Not good. Most good astronomy is done at or below the 1% level, which is why I spent so much time tinkering with the program at the outset.
But it seems now that things worked pretty well. In all of the frames, the bottom 1/3 of the image is not well focused, which means that the fitting routine doesn't do as well. For most images, we still have that <1% residual, but a few are larger. In particular, the images taken in the B filter are the worst. Unfortunately, it doesn't look like much can be done about this.
To do this error checking, I have to take the image, and look at the intensity of the star before and after the subtraction. Ideally it should go to zero. Sometimes it doesn't quite get everything, and other times, it overestimates how bright the star was and subtracts too much. To get a real good feeling, I divide the images into 9 pieces mentally (a 3x3 grid) and pick the brightest star from each one to check.
At some point, I'll try to go through the concept behind the process of what I've been doing thus far. Conceptually it's not too hard, but it would be best explained with some pretty images.
I only ended up making it through about half of the 21 images to check. Then we headed to the beach again. Waves were much better, but I couldn't borrow a boogie board and ended up body surfing (which is fun nonetheless).