Change page style: 

Public Data Readme

GPI Public Data Early Release Readme


Contributors:

GPI first light observing team: Bruce Macintosh, James Graham, Stephen Goodsell, Les Saddlemyer, Dave Palmer, Jeff Chilcote, Andrew Cardwell, Jennifer Dunn, Ramon Galvez, Gaston Gausachs, Markus Hartung, Pascale Hibon, Patrick Ingraham, Marshall Perrin, Lisa Poyneer, Carlos Quiroz, Fredrik Rantakyro, Naru Sadakuni, Dmitry Savransky, Andrew Serio, Sandrine Thomas, Kent Wallace, Schuyler Wolff

GPI data analysis team: Marshall Perrin, Jerome Maire, Patrick Ingraham, Dmitry Savransky, Christian Marois, Rob De Rosa, Jeff Chilcote, Rene Doyon, Zack Draper, Quinn Konopacky, Franck Marchis, Max Millar-Blanchaer, Jennifer Patience, Laurent Pueyo, Abhi Rajan, Jean-Baptiste Ruffio, Sandrine Thomas, Jason Wang, Kim Ward-Duong, Schuyler Wolff


Description of Targets


The targets for this data release fall into five categories:

  • Photometric/Spectroscopic standards. Most of these were observed open loop (i.e. AO control off) and unblocked by the coronagraph to measure system throughput while avoiding saturation. Some targets were observed with and without the apodizers and Lyot stops to measure the relative throughput of the coronagraph masks. The targets are from the IRTF Spex spectral library (HD 51956, HIP 30979, HD 20619), plus the white dwarf companion to HD 8049 from Zurlo et al. 2013.  

  • Astrometric Standards. Trapezium stars Theta1 Orionis A and B (compare with Close et al. 2013), and HD 1160 (compare with Nielsen et al. 2012).

  • Bright overhead test targets (but otherwise undistinguished stars).  During the first run we tested the AO system and coronagraph on a series of targets chosen primarily for brightness and position in the sky. We include these as examples of GPI coronagraphy in a range of atmospheric conditions and observing bands, but caution that (a) these were early nights and instrument performance was very much in flux as operating parameters were adjusted, and (b) the seeing was relatively poor.

  • Stars with Known Planets and/or Disks.  HD 95086 (H and K1 bands), Beta Pic (J band only), HD 10647 (J), HD 61005 (J).

  • Resolved Solar System Targets. Neptune (H band)


A subsequent data release(s) will add targets in polarimetry mode and in GPI’s non-redundant aperture masking (NRM) mode. The GPI team is still working to improve calibration of these modes and handling of some systematics before releasing these data.  


Finally, other targets have been observed but remain proprietary to the GPI team for the time being, including Beta Pic (H, K1, K2 bands) and HR 8799 (K1 and K2 bands).  



Important Cautionary Notes


The released FITS files represent the GPI instrument team’s best effort at calibrating and reducing these observations, but instrument calibration is necessarily a work in progress and incomplete. GPI is a highly complex instrument that we are still getting to know. Use your own experience and scientific judgement when analyzing and publishing data from GPI.


In particular, the stability over time of GPI’s absolute orientation with respect to the sky has not yet been characterized. Position angles should be considered uncertain to ~ 1 degree.


Every commissioning comes with surprises. For GPI these included:

  • More vibration than desired, due to the cryocoolers on the IFS. This manifests as tip-tilt and focus jitter in these data, and reduces contrast by ~ 0.3 mag for many targets. Instrument modifications in January and February 2014 are expected to mitigate this. Note also that contrast degradation is a function of angular separation.

  • Somewhat lower throughput was achieved than expected, with particularly sharp falloff beyond 2.2 microns in the K2 band. This is still being investigated. GPI’s design and optical coatings were optimized for performance at shorter wavelengths such as H band.

  • An artifact identified before commissioning—an optical ghost called the ‘seagull’ that can be seen in flat field or arc lamp images—is due to a caustic reflection inside the IFS of light from outside the field of view. A new baffle was installed in January 2014 to mitigate this.

  • The brightness ratios of the satellite spots relative to one another are variable at a level of up to a few tens of percent. This is believed to be due to an interference phenomenon between speckles and the diffracted spots. Quantitative models and calibrations are under development. Users should be cautious in assessing uncertainties when conducting photometry relative to satellite spots, and should not rely solely on any one single spot.

  • The lenslet (spaxel) pixel scale appears to be exactly as expected, 14.3±0.1 mas/lenslet. Distortion has not yet been characterized on sky; laboratory tests measured median distortion of 0.26 spaxels over the full field of view, and a correction based on these lab measurements is included with the data pipeline, which should reduce residual distortion to <0.04 spaxels.

  • Due to IFS internal flexure, the microspectra shift locations with respect to the detector, generally less than one pixel but with occasional larger shifts. Calibrations for this are still being improved. The wavelength calibration of the reduced data files should be considered uncertain by ~0.2% (~ 3 nm).  Shifts will vary from image to image depending on changes in elevation.



Data Reduction Notes


We summarize here the steps by which the raw data were reduced for this data release. Readers should also consult the GPI data pipeline documentation for additional descriptions of the individual steps in processing. As with all outputs of the GPI data pipeline, detailed reduction history and all reduction parameters are saved in FITS headers, so consult those for full details.


The data were reduced with v1.0rc1 (release candidate 1) of the GPI data reduction pipeline using the following primitives:

 

  • Flag Quicklook

  • Load wavelength calibration

  • Subtract dark background

  • Destripe science image (a)

  • Interpolate bad pixels in 2D frame

  • Update spot shifts for flexure

  • Assemble spectral datacube

  • Divide by lenslet flat field

  • Interpolate wavelength axis

  • Correct distortion

  • Measure satellite spot locations (b)

  • Measure satellite spot peak fluxes (b)


(a) - Destriping only performed on J-band closed-loop data.

(b) - Satellite spots only measured in a subset of the closed-loop data.


Sky background observations were obtained for a subset of targets,  offset by ~ 30 arcsec from the science objects. For this subset of observations, a sky cube was constructed through a median-combination of each reduced sky cube, which was then subtracted from each reduced science cube. For observations that lack contemporaneous sky data, no sky subtraction was performed in these reduced files.


Calibration files were selected either automatically or manually for the following primitives:

Load wavelength calibration -- automatic

Subtract dark background -- automatic

Interpolated bad pixels in 2D frame -- automatic

Divide by lenslet flat field -- manually (S20131212S0004_lensletflat_spec.fits)

Correct distortion -- manually (S20121210S0066-distor.fits)


The default recipe parameters were only changed in two instances; (1) when an offset was needed to align the wavecal solution with the spectra in the unprocessed 2D images, and (2) when the automatic satellite spot location algorithm failed due to low SNR on the satellite spots:


1 - Wavecal offset


For a subset of the reductions, the offset obtained from the lookup table in the ‘Update spot shifts for flexure’ primitive were not sufficient to fully correct for the flexure of the instrument. This problem manifests itself as a strong Moire pattern within each slice of the reduced data cube, which is discussed under ‘Reducing your science data’ here:


http://docs.planetimager.org/pipeline/usage/quickstart.html


For observations where the offsets obtained from the lookup table led to a poor reduction, GPItv was used to determine the correct offset to apply by visually aligning the spectra within the 2D image with the overplotted wavecal solution. This offset was then entered into the ‘manual_dx’ and ‘manual_dy’ parameters of the primitive, with the ‘method’ parameter changed to ‘manual’, and the data were re-reduced. As the expected change in the offset caused by instrument flexure as a function of target elevation was small for a given set of observations on a target (typically <0.2px), a single offset value pair was used for each sequence of a given target. This procedure was necessary for the following observations:

All Y-band

All J-band

H-band observations of Neptune

All K2-band


Improved automation for flexure compensation is under development for future versions of the data pipeline.

2 - Satellite spot finding


For a subset of closed-loop observations taken under poor seeing conditions, or for which the satellite spots are only marginally detected, the automatic spot-finding algorithm (‘Measure satellite spot locations’) fails during the reduction process. In this case, the data are initially reduced without the final two satellite spot primitives. The pixel position of the spots in the 15th frame were measured manually, and the data were re-reduced with the following modifications to the parameters in the ‘Measure satellite spot locations’ primitive:


reference_index = 15

loc_input = 1

x*, y* = [x, y pixel position of the corresponding satellite spot]


The next release of the data pipeline will include refinements to the automated spot-finding algorithm which will hopefully increase its success rate.  However, images with very low SNR in the spot locations will almost always need to be reprocessed manually to locate the spots.


Using the Satellite Spots:

The measured satellite spot locations and fluxes are written to the extension header in each FITS file, along with the derived mean star locations (keywords SATSi_j for sat spot locations, SATFi_i for sat spot fluxes, where i gives the wavelength index and j in {0-3} indexes the four spots.) Keywords PSFCENTX and PSFCENTY give the star locations;  Note that these are the mean locations across the spectral axis, but due to atmospheric differential refraction the star’s apparent position will generally vary with wavelength. (These data were all obtained prior to GPI’s atmospheric dispersion corrector being commissioned for use, so ADR is uncompensated).  The wavelength dependent position may be computed from the SATSi_j keywords and used to align the wavelength channels.


The satellite spots can also be used for photometry relative to the central star. The satellite spot intrinsic brightness ratios have been measured in the lab for each apodizer (one per filter). See the file config/apodizer_spec.txt included with the data pipeline.  Datacubes may be calibrated into physical flux units using the “Calibrate Photometric Flux” primitive in the data pipeline.