Post on 28-Dec-2015
transcript
Background on WFC distortion
• General difficulty calibrating HST– Need high-density field, accurate positions– No satisfactory fields exist– Need self-calibration
• ISR on HRC distortion released a year ago– WFC more complicated:
• Largest HST field
• PSF spatially variable
Overview of this talk
1) PSF issues– Spatial variation
– Time variation
– Fitting stars
– A useful program
2) Distortion solution– Difficulty of calibration
– Form of solution
– Time variability
3) How-to Astrometry with the WFC
PSF Issues (1)
• Need a PSF to measure stars to solve for distortion– Several routines are coming out– My routine: img2xym_WFC.09x10.F
• Similar to the my HRC routine
• Operates on _flt images
• Uses an array of PSFs to deal with spatially dependent charge diffusion
– Between 17% and 24% of a star’s light in central pixel
– Affects photometry at the +/- 4% level
– Affects astrometry at the 0.01 pixel level
PSF Issues (2): treating the PSF
• The base PSF model• 9x10 array of PSFs• 101x101 pixels• 4x super-sampled• Use bi-cubic interpolation• Covers out to r = 12.5 pix• “Effective” PSF
• Time variability• Typically 5% in the core• Treat as perturbation:
PSF(dx,dy;x,y,NIM) = PSF(dx,dy;x,y) + PSF(dx,dy;NIM)
PSF Issues (3): the program
• Operation of program:– Take _flt image– Simple finding criteria– Return (x,y,m) for sources– User collates with other observations
• Measurement quality (internal precision)– Photometry: 0.005 magnitude– Astrometry: 0.01 pix
Distortion Solution (1): Why?
• Need for distortion solution– Image rectification
– Stacking to go deep
– Source identification
– Spectra slit/fiber placement
– Lensing analysis
– Astrometry
• Different applications require different accuracies
Distortion Solution (2): Solving for
• Ways to solve for– Best way: calibrated reference frame
• None exists with density/precision useful for HST
– Alternate way: self-calibration• Compare two WFC images of a good-density field
• Hard to know where the distortion error is
• Hard to visualize distortion– 2-d function over a 2-d surface
• Hard to measure distortion outright– But easier to test for errors
Distortion Solution (3): History• Solution history
• Meurer GO-9028• F475W of 47Tuc
• 4th-order polynomial
• Linear-term degeneracy
• Anderson GO-9443• Took orthogonal observation
• Used several filters
• Filter-dependent residuals
• Slightly different quadratic terms
• 68.2666-column pattern, amplitude 0.01 pixel
Distortion Solution (4): Form
• Final form of solution1) Column correction: amplitude 0.01 pixel
2) Polynomial: amplitude 40 pixel
3) Filter-based look-up table: 0.05 pixel
• Software now available for 12 filters• Better for some filters than others
• Used in the drizzle pipeline
• Supplementary program to improve solution for F606W and F814W: GO-10252• Use inner field in Omega Cen: 88,000 stars, even density
• Tables to be improved, PSFs obtained
• Problem: out of focus, just provides a check
• Other checks on solution
Distortion solution (4): Check #1
• Checking the distortion solution– Easier to check than to solve for
– Three tests: short-term, long-term, out-of-focus
• Short-term time variations– GO-10424 (PI Richer)
– 126 orbits taken over 4 weeks
– Each orbit: F814W, F606W, F814W
– Compare each to the average
– Hard to separate distortion variation from PSF variation
– Typical variation is much less than 0.02 pixel
Distortion Solution (5): Check #2
• Long-term variation– Outer field in 47 Tuc– Observed over 300 times by WFC– Inter-compare exposures, allowing for linear
transformation• Examine astrometric and photometric residuals• Linear variation of linear skew term: 0.1 pixel over
three years• Typical systematic residuals are 0.02 pixel
Distortion Solution (6): Check #3
• Calibration supplement program GO-10252– 1 orbit for each of F606W, F814W– Aim to improve the fine-scale solution
and provide good empirical PSFs– PSF very much out of focus
• 10% low in central pixel
• Use as comparison test
Distortion Solution (7): Summary
• Short term (weeks)– Linear and quadratic good to 0.02 pix
• Long term (years)– Linear has systematic trends– Quadratic stable to 0.02 pixel
• Out of focus – Errors up to 0.03 pixel at edges
Prescriptions for astrometry (1)
• Accessing the solution– ISR coming very soon, with FORTRAN
programs for finding/measuring/correcting– Included in the drizzle pipeline
• Planning observations– Accuracy: 0.01 pixel per exposure, but…– Beware small systematic errors of ~0.02 pixel
• Planning can minimize/identify these• Ideal dithering depends on goal of project
– Dense field: may be able to solve for PSF– Sparse field: need large dithers to average out spatially
dependent errors
Prescriptions for astrometry (2)
• Reductions– Measure _flt images only (x,y,m)
– Correct for distortion
– Cross-ID stars in different images
– Carefully perform transformations• 6-parameter linear
• Go local if necessary
– Combine similar things first• Identify systematic errors
• Get a handle on random errors
• Lots of Astrometry left to do!