CTE in the Dark An Empirical Pixel-Based Correction for CTE TIPS/JIM

advertisement
CTE in the Dark
An Empirical Pixel-Based
Correction for CTE
TIPS/JIM
January 21,2009
Jay Anderson
Luigi Bedin
Shuffle
30s, 47 Tuc Outer field
CTE/CTI
Steadily increasing problem for:
– STIS, ACS’s WFC, … WFC3?
– Was also bad for WFPC2, HRC
Symptoms:
– Charge trails
– Loss of flux
Cause:
readout
– Traps within pixels that delay readout
– Trap density increases linearly over time
Traditional mediation:
– Photometric correction ; astrometric?
– Experimental pixel-based corrections
• STIS: Bristow 2002+
• ACS: Massey et al 2010
– Theoretical plus empirical
• New approach here: purely empirical
observed
One Raw Dark, post SM4
Stack of 168 Post-SM4 Darks
“Peak” Map
A Purely Empirical Plan
PASP Paper in preparation with L. “R”. Bedin
Inspired by
– HVS project (PI-Oleg Gnedin)
– Massey et al. (2010): WPs in COSMOS science data
Plan:
• Examine WPs in darks
• Study two dimensions:
– Profile scale: dependence on WP intensity
– Profile drop-off: dependence on n
• Focus: “Just numbers”
– Lots of trails
– First step: mechanics of measuring the trails
CR Tail Measurement
Empirical
Trails
Faint
No “notch”
channel
apparent!
Bright
TOTAL IN TAIL
Total
Power in
Tails
WP INTENSITY
A Simple, Empirical Model
• Different traps affect different electrons
– More traps affect lower-hanging electrons
– (q): traps per pixel at each chg level
• Different traps may have different release times
– Follow release out to 100 pixels
– Model: keep track of each trap’s state
– All the charge?
• What about shadowing?
– Well known background effect
– Leading pixels?
– How to resolve?
What about
Shadowing?
WP

?
CR
X
CR
?
WP~5000
Yes! Shadowing is
essentially “perfect” !
WP~2500
A Simple, Empirical Model
[1] Different traps affect different electrons:
(q) = trap density (total number)
[2] Different traps may have different release times
(n;q) = release profile
[3] Perfect shadowing!
Instantaneous filling of traps
Readout Model’s Four Stages:
(1) Shuffle out current cloud
(2) Release trapped charge
(3) Shuffle in new cloud
(4) Trap new electrons
Correction Scheme
Start with a readout model
–
Two parameters:
1) Trap density:
(q)
2) Release profile: (n;q)
–
Input PIX(j)
Iterate
–

output PIX(j)
Find source function PORIG(j) that produces POBS(j)
Optimize model:
–
Minimize trails in darks by varying (q) and (n;q)
Independent tests:
(1) Trails
(3) Astrometry
(2) Photometry
(4) Shape
Corrected
WP Trail
Residuals
Faint
Bright
Adjust by hand
the model
parameters
1) density: (q)
2) profile: (n;q)
Corrected WP Deep
(q)
Trap Density - vs - q
Trap Profile
(n;q)
Two Components of the Model
Detailed Model Example
The tests…
1)
2)
3)
4)
Aesthetic test:
Photometry:
Astrometry:
Shape:
trails gone?
flux back?
flux in right place?
flux really in the right
place?
339s, 47 Tuc Outer field
339s, 47 Tuc Outer field
30s, 47 Tuc Outer field
30s, 47 Tuc Outer field
30s, 47 Tuc Outer field
30s, 47 Tuc Outer field
47 Tuc Calibration Catalog
53,000 stars
x,y,m
BRIGHT
FAINT
Photometric Residuals in
Deep 339s 47 Tuc Images
BRIGHT: Near Saturation
FAINT: about 50 e- max
Photometric Residuals in
Short 30s 47 Tuc Images
BRIGHT: Near Saturation
FAINT: about 50 e- max
Astrometric Residuals in
Short 30s 47 Tuc Images
Bright
Faint
Corrected
What
about
shape?
Summary
2-component model (q) and (n;q)
– Pameters based solely on WPs in darks
– Readout model, invert to get original pixels
Tested against stars:
–
–
–
–
Images with backgrounds of 1.5 DN2 and 15 DN2
Trails removed
Photometry/astrometry generally restored
Shape surprisingly good
Remaining issues?
Remaining Issues
• Reminder: just a proof of concept
• When best to do? _flt or _raw?
– Either is ok
• Pipeline modifications?
– Use of darks, biases
• Improvements:
– Speed: 5 iterations = 10 minutes/exposure
– Faint and bright extremes poorly constrained
• Verify linear time behavior (pre SM4…)
• Read-noise amplification
– Should apply algorithm only to real structure
• X-CTE
– Yes, but…
Ask me!
THE END
Backup Slides
READOUT SCHEMATIC
30s, 47 Tuc Outer field
30s, 47 Tuc Outer field
Original
“Smoothed”
RN Component
Decomposition
Original
Repaired Original Repaired Modified
Actual Change
Change for Original
Change for RN-Smoothed
Just the change
Serial CTE
Serial CTE linear trends
Serial CTE
Parameters
Download