At the telescope we take flat-field exposures to try to estimaterelative variations in the sensitivity of pixels across the detector. Some pixels might be less sensitive or less strongly illuminated than others, and we want to be able to capture those differences
Going forward in this step, we might:
use multiple quartz-lamp flat-field images to make a median, unnormalized, flat-field image
identify a mask of which pixels are actually illuminated by the dispersed spectrograph slit (possibly unnecessary)
estimate pixel-to-pixel variations by filtering out large-scale features in the median flat due to the quartz lamp Planck spectrum and the spectrograph sensitivity; this sensitivity map represents the variations a star's spectrum will see if it wobbles slightly in the slit
Going backward in this step, we might:
start from an image of photons hitting the pixels and multiply by the pixel-to-pixel sensitivity to introduce defects in the image
To test/explain this step, we might:
explain in friendly words what's happening
visualize the individual flat frames as an animation, along with the median
visualize the unnormalized flat-field frame, the smoothed large-scale features being removed, and the pixel-to-pixel sensitivity
test that we recover the right answer from a simulated image
At the telescope we take flat-field exposures to try to estimaterelative variations in the sensitivity of pixels across the detector. Some pixels might be less sensitive or less strongly illuminated than others, and we want to be able to capture those differences
Going forward in this step, we might:
Going backward in this step, we might:
To test/explain this step, we might: