flat¶

flat builds a master flat, normalised to a mean of 1, from twilight, dome or night-sky flats of one filter.

The flat maps the response of the whole light path: pixel sensitivity, vignetting, dust and filter non-uniformity. Each input is corrected for bias and dark, then the frames are combined using their photon statistics, which suppresses cosmic rays and other defects.

For a good flat, expose to about half of the full well, take more than 13 frames, and make one flat per filter with the FILTER keyword set, so that phcorr can check that flat and frame match.

Usage¶

munipack flat [options] file(s)

Options¶

Option	Default	Description
`-gain`	header `GAIN`, else 1	Gain [e⁻/ADU]; overrides the header.
`-bias`	—	Master bias subtracted from every flat.
`-dark`	—	Master dark subtracted from every flat.
`-xdark`	exposure-time ratio	Dark scaling factor t.
`-bitmask`	—	Mask frame (0 = skip pixel; set to 1 in the master).
`-st`, `--saturate`	header `SATURATE`	Upper limit of raw values used [ADU].
`-th`, `--threshold`	tiny positive value	Lower limit of raw values used [ADU].
`--approximation`	`standard`	`standard` (full fit) or `basic` (quick robust mean).
`-o`, `--output`	`flat.fits`	Output file.
`-B`, `--bitpix`	−32	Output type; keep −32, integer types lose precision.
`--verbose`	off	Print per-frame statistics and a gain estimate.
`-h`, `--help`	—	Show the help.

Output¶

The master flat (mean 1), with a STDERR extension holding the error of each pixel. Give it to phcorr with -flat.

Examples¶

# V flat from bias- and dark-corrected twilight flats
munipack flat -bias bias.fits -dark dark30.fits -o flat_V.fits flat_V_*.fits

# estimate the gain of the camera
munipack flat --verbose -gain 1 -bias bias.fits -o flat_test.fits flat_V_*.fits

Estimating the gain¶

With --verbose, flat reports the gain relative to the one it applied. Start with -gain 1, multiply by the reported ratio and repeat until the ratio is close to 1. Read noise and non-linearity bias this estimate, so compare it with the manufacturer’s value.

Notes¶

flat keeps all input frames in memory (about 8 bytes per pixel per frame), so a large set of big frames needs plenty of RAM.
Pixel rejection, masks, header checks and input lists work as for all masters: see Common to bias, dark and flat.