Photometricity versus Calibration¶
Note: see SV Calibration Plan page for further iteration on Jim's plan here.
My thinking on this is that there are two parts to photometric calibration,
"photometricity" and "calibration".
The limiting factor to photometricity in wide field cameras is scattered light
reaching the focal plane CCDs. Thus we should plan to measure, limit, and correct
for the scattered light. Recall that any scattered light in the flat field image
translates to incorrect qe corrections and thus directly into photometric errors.
1) light scattered from the flat field lamp and screen (and deCal system)
- affects calibration directly
- BRO model suggests light path is primarily from primary cell around primary
shining directly into the camera and from light scattered by the secondary vanes
onto the primary mirror and thence into the camera
Pinhole camera in filter slide to image what the camera sees.
Eliminate what light from outside primary can be.
2) light reflected off of the focal plane and down to filters or optics, and reflected
back onto ccds. AKA: pupil ghost. Amplitude from ray tracing calculations ~3%,
but shape and amplitude varies with filter height.
- must be subtracted from flat fields. Since varies over a ccd must
also be subtracted from data, alas.
- calculable, but not with precision necessary
- a "standard candle" for stray light measurements
- define to be azimuthally symmetric
- probably can be measured with a flat field
- probably a correction image constructed from ray tracing radial profile sufficent
- but an opportunity: if left uncorrected in the flats, both star flat
and ubercal technologies should be able to recover, easily.
Flat field azimuthal average and fit to radial profile
Star flat/ubercal try to reproduce
3) Remaining stray light.
a) if it is in the data, and is smooth on a scale of a ccd, ignore: it sky subtracts out
b) if it is in the flat field, it must be measured via star flat or ubercal
In the DESDM terminology this is an "illumination correction". It should not be
built from sky flats because a) the sky is not flat, b) star flat/ubercal makes
better estimates, and c) there are no subtle vignetting effects in this wide field camera.
- Arguably this correction is made at the image processing level ("illumination correction")
rather than at the catalog level.
10 x 10 or perhaps 62 observations of a Stripe 82 field with the central target
moved from ccd to ccd (62), or moved from grid point to grid point (10x10).
(DLT: Since DESDM is solving for photometric zeropoints on a CCD-by-CCD
basis, maybe the appropriate star flat measurement is on the scale of
a CCD rather than the scale of the focal plane -- i.e., perhaps we should
consider a grid pattern of, say, 5x10 small dithers with a grid cell size
of about 100 arcsec x 100 arcsec. This would result in an individual star
flat for each of the 62 science CCDs -- 62 star flats in total -- which
then could be converted into FITS images for use as illumination correction
This path leaves us with a clean focal plane: the flat field is what would be measured
from a perfect camera.
It is worth noting that several of the CCDs on the focal plane have been absolutely
calibrated, so we know the number of photons/s/cm^2 reaching them. Juan Estrada has
4) Stray/Scattered light from the Moon and very bright stars
- Are there light paths onto the focal plane without going through filter?
- To what extent does the moon leave light structured on the scale of a CCD
on the focal plane, as a function of moon to boresight distance?
- We'll be working in moony conditions; the moon is so bright that unexpected things occur.
- Alpha Phoenix is ~0 mag star ~30' from edge of SN field E1
Data from near the moon; Stripe 82 is probably perfect because the moon
is, near full and during Sept-Dec, very near Stripe 82 and the SDSS data
(taken under moonless conditions) is quite good.
Data on SN field next to alpha phoenix; do they need to move the field?
What fraction of ccds have their sky brightness raised by 1 mag due to
a mag=3 star on boresight? At 1.2 degrees of boresight?
The focal plane may be perfect and well baffled, but the tranmission varies as a function
on focal plane position.
5) Bandpass variations as a function of focal plane position
- there are, due primarily from filters, but also CCDs and optics
- instrument response measurements have to be made
- in best of all possible worlds, atmospheric transmission would also be measured
- variations ("color terms") can be corrected for at catalog level
Stripe 82 or SN fields + system response engine
From calibrated spectra or stellar population models, recover observed colors as
a function of focal plane position
Even if we reach a perfect instrument we have to ensure it remains so during the survey.
6) Instrumental stability
- flat field lamp stability
- deCal stability, versus time and temperature
- deCam y,z bandpass stability as function of focal plane temperature
- star flat/ubercal results from late commissioning and from mid-late SV
Measurements: straight forward to describe
The grid of data take for the star flat is perfect for ramping up the
understanding of the ability of the survey to do relative calibration, taking
all observations onto a single well defined instrumental magnitude.
7) Instrument Relative Photometry
- relative calibration involving only zeropoints ("global calibration")
- relative calibration involving solving for instrumental and atmospheric effects ("ubercal")
- incorporating focal plane position dependent bandpass ("color terms"/"george II")
Measurements: the 10x10 or 62 grid of observations
Now that the instrument is behaving and is understood enough to find the same
flux from the same star observed on different parts of the focal plane, it
is time to turn that flux into photons/sec/cm^2/nm
1) Demonstrate the abilty to do "all-sky photometry"
- take observations of ~ 100 standard fields
- calculate zp, extinction, perhaps the color terms for a ccd or two
- take observations of white dwarfs and special stars
Now we turn to the ~100 sq-degrees of area in SV taken with up to 10 tilings.
This is the ideal data set to test global relative calibration/ubercal.
2) Demonstrate calibration via massively overlapping images
- rederive pupil ghost using sf/uber technology, just to show it can be done
and that it isn't different from star grid data derived one. Compare
precisions of various technologies
- remove pupil ghost and rederive scattered light corrections.
compare with star grid derived one, look for temporal changes, compare
precisions of various technologies
- apply scattered light correction and derive zeropoints using grc/uber tech.
learn just what model for the focal-plane+atmosphere is needed for ubercal
As many tilings of the ~100 sq-degree field as possible. Only single image
processing needed, as this is all work done with star catalogs from individial images.
Pan-STARRS/Chris Stubbs seems to be limited by aerosols, though I would have
thought that component of atmospheric transmission is exactly what ubercal tech would have been
good at. Resarch needed.
3) Learn how we are to deal with the atmosphere
- just incorporate into ubercal model?
- use GPS water vapor measure to calibrate a MODTRANS transmission spectra
for use in ubercal and/or george II?
- go all out for atm, and use its output !MODTRANS transmission spectra
in ubercal and/or george II?
same as above, but incorporating the GPS stream and whatever comes out of the
atm testing campaign this fall