The initial step in the Kepler science data pipeline is performed
by the software module termed CAL . CAL converts raw data numbers into
calibrated pixels, which are then passed to the
Photometric Analysis module for compilation into a light curve. The Science Operations Center
at Ames receives pixel data (Haas etal 2010) from the Data Management Center at STScI. Data from
each Kepler CCD is formatted as
FITS files, including
collateral pixel data, collected for calibration. CAL operates on both the 30 min (long
cadence) and 1 min (short cadence) observations, as well as on the full frame
images. Users of
CCD data will be familiar with most functions of the calibration module, however we note a
few aspects peculiar to the operational modes of Kepler. In general, most users of Kepler
photometry will not work directly with raw counts, so the information provided here is to
inform users on the the process by which calibrated pixels are generated.
The purpose of this webpage is to provide a brief guide to
CAL functions. GOs and archival users are urged to examine the primary documentation, and
look on this page for discussion of specific issues. Primary documentation includes:
- Kepler Instrument
Handbook, Version 1, 15-July-2009.
- Release Notes,
which provide detailed information for each data release, either initial or reprocessed
data per quarter.
- Jenkins etal,
2010, "Overview of the Kepler Science Processing Pipeline".
- Caldwell etal,
2010, "Instrument Performance in Kepler's First Months".
- "Pixel-Level Calibration in the Kepler Science Operations Center Pipeline", by
E. Quintana etal, SPIE Conference on Astronomical Instrumentation, June 2010
(paper available July 2010).
CAL performs a number of functions familiar to CCD
photometrists to transform raw data
counts into calibrated pixels, operating on a single CCD channel at a time. Detailed models of
each CCD were developed during pre-flight testing, and are combined with full-frame images obtained
during the commissioning perior both before and after ejection of the aperture cover.
Each CCD channel consists of 1070 rows by 1132 columns, of which only a subset (1024 x 1100)
is used for photometry.
Block diagram of a Kepler CCD, showing the science and
collateral pixel locations.
The focal plane CCD models are applied within CAL to execute
the following tasks:
The CCD bias level is determined using collateral pixels obained with each CCD read.
Users should note that the term "black level" is used in much of the Kepler documentation,
as a synonym for the more commonly used "bias level". The bias level contains both a 2D
map, and an additional 1D "dynamic" bias correction. Once the 2D black level is removed,
a fit to the residual bias is used to estimate a 1D black correction. Note the difference
between Kepler's bias subtraction method, and the approach commonly used for shuttered,
operation, in which the user takes separate bias frames, before, during and after data
Dark current is estimated from the masked and virual smear pixels. Since the focal plane is
maintained at −85 C, the effective dark current is essentially zero. (Caldwell etal
Kepler observes continuously, with no shutter; therefore stars illuminate the CCDs
during readout. This "staring" mode produces trails along columns that contain
stars, as charge is smeared out during the read. Each pixel in a given column of the image
receives the same smear signal. These values are typically small, since each pixel only
“sees” a star for the readout time, 520 milli-seconds, divided by the number of rows 1070.
The smear level corrction in each image is determined using the masked and virtual smear
pixels set aside for this purpose, as seen in the figure above.
The gain function associates observed photoelectrons (e-) to the analog/digital units
generated by the A/D converter. Gain is the average slope of the transfer function, and has
a median value of 112 e-/ADU across the focal plane (Caldwell 2010).
A measure of the deviation from the linear transfer function is estimated at each ADU signal
level; this nonlinearity model provides appropriate corrections.
An issue noted during pre-flight testing can be described as a large, signal-dependent trailing
undershoot in the image, traced to an amplifier in the local detector electronics. An undershoot
model is applied to correct affected pixels.
A flat field correction is the last step in CAL, applied to photometric pixels to correct
for spatial variations in pixel sensitivity to a uniform light source. CAL uses a local flat
field, also termed the pixel response non-uniformity (PRNU) map. The PRNU image maps each
pixel’s relative brightness variation from the local mean, expressed in percent (Van Cleve &
Caldwell 2009). The median standard deviation of the pixel values in the PRNU image across the
focal plane is 0.96 %. The flat field was developed during ground testing prior to launch; there
are no separate flatfield exposures obtained on orbit.
CAL produces the basic photometric product of Kepler: a series of
calibrated target pixel images within a pre-set aperture. The photometric time series is
constructed from these images, in the next step, Photometric Analysis, or