STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

1

Removing SAA-Persistent Cosmic Ray Fluxfrom NICMOS

Anton Koekemoer (INS)with Elizabeth Barker, Vicki Laidler, Eddie Bergeron

Overview of SAA persistence

Algorithm for removing SAA persistent flux

Implementation & testing of “saaclean” task

Future plans

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

2

SAA Persistence Overview of Persistence:

NICMOS HgCdTe/CdTe bulk material contains small flaws that trap electrons produced by detected photons

The trapped electrons are not read out together with all the other electrons detected in a given pixel

Thermal excitation causestraps to release the electronson longer timescales with alogarithmic decay

Resulting flux appears insubsequent exposures, witha distribution that dependson the original intensity andthe population of traps

(Bergeron & Dickinson, NICMOS-ISR-2003-10)

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

3

South Atlantic Anomaly (SAA): Region in Earth’s magnetosphere where Van Allen belts dip closest to the

surface, resulting from misalignment in Earth’s magnetic axis HST passes through it ~50% of the time, 8-9 orbits/day Although NICMOS is powered off, cosmic ray electrons are still trapped Cosmic ray rate is so high that persistence can increase total noise by 4-5x

in subsequent exposures obtained long after the SAA passage(Bergeron & Dickinson, NICMOS-ISR-2003-10)(O. Lupie, WFC3-ISR-2002-01)

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

4

Removing SAA Persistent Flux Post-SAA darks:

After each SAA passage when NICMOS transitions from SAAOPER to OPERATE, a pair of dark exposures is obtained:

– ACCUM mode, NREAD=25, EXPTIME=256s– Darks start 174s and 444s after exiting the SAA

Calibrate the darks to remove other instrumental signatures (pedestal etc) Create average post-SAA dark image:

– scale the 2nd dark to the level of the 1st dark, using the decay time constant– average the 2 darks to remove CRs accumulated during the dark exposures

(Bergeron & Dickinson, NICMOS-ISR-2003-10)

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

5

Subtracting the SAA model: The average of the 2 post-SAA darks is the SAA model, representing all

the CRs accumulated during SAA passage Scale and subtract from the calibrated (pedestal-corrected) science image Scale factor is determined iteratively:

– multiply SAA model image by small scale factors– subtract from science image– measure the resulting noise (FWHM of Gaussian fit to pixel histogram)– Plot noise as a function of scale factor, fit parabola to determine optimal scale

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

6

Fitting for High and Low pixels: Some CRs during SAA have higher energy and/or occur later than others Need to fit different scale factors, depending on the level of the residual signal In practice, divide the pixel histogram into 2 regimes: low and high Determine separate scale factors, and only apply a correction if the reduction in

noise is > 1%

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

7

Example Results

(Barker & Koekemoer, 2005 HST Cal Workshop

Before SAAclean After SAAclean

SAA CR Persistence modelPost-SAA Dark

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

8

Implementation of SAAclean task Original script:

IDL, provided by E.Bergeron New version:

Python code, written by Vicki Laidler, following original IDL prototype Installed as Pyraf/STSDAS hst_calib.nicmos.saaclean Some slight differences in algorithms (mostly related to IDL fitting

functions that are unavailable or different in Pyraf) Inputs:

– calibrated exposure (optionally corrected for pedestal effect)– Post-SAA darks (read from the image header)

Output:– calibrated exposure with the SAA model subtracted– SAA model (in off-line version)– Lots of output text containing information about the fitting process– Fitting parameters also stored in header keywords

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

9

Testing Procedures Compare IDL and Pyraf versions:

run on the full variety of datasets, ranging from extreme SAA events to datasets that are not impacted

resulting model files and output images compared, examining:– noise in the images– individual pixels– dependence on quality of bad pixel files

Results:– some differences between IDL and Pyraf; generally only in treatment of bad

pixels, while SAA-impacted pixels showed the same behaviour

Full test of the Pyraf version: Ran on all SAA-impacted NIC1,2,3 exposures ever obtained (>6,000) Examined images, as well as the output values of:

– chi-sq– nhigh/nlow pixels and the threshold value– scale correction factor– noise reduction values

STScI TIPS19 January 2006

Removing SAA-Persistent Cosmic Ray Flux from NICMOSAnton Koekemoer (INS)

10

Future Plans Off-line version:

Prototype version was released publicly in STSDAS 3.4 (Nov 1, 2005) Available for use within STScI and in the outside community; feedback

solicited Finalize addressing minor issues turned up in very extreme datasets (very

low or very high chi-sq, or low/high pixel regimes not well modelled):– appears that all of these can be fixed by using the correct bad pixel files

Add some additional parameters to improve flexibility of iterations Next release likely in Feb/Mar 2006

Pipeline version: Aim to incorporate all improvements in off-line version by Feb/Mar 2006 Likely release to OPUS in Apr/May 2006 Initially only run on data that is within 1 orbit of SAA passage Correction only applied if noise reduction is >1% Continue to solicit feedback from community