THE INFRARED SPECTROGRAPH
1
(IRS) ON THE SPITZER SPACE TELESCOPE
J. R. Houck,
2
T. L. Roellig,
3
J. van Cleve,
4
W. J. Forrest,
5
T. Herter,
2
C. R. Lawrence,
6
K. Matthews,
7
H. J. Reitsema,
4
B. T. Soifer,
8
D. M. Watson,
5
D. Weedman,
2
M. Huisjen,
4
J. Troeltzsch,
4
D. J. Barry,
2
J. Bernard-Salas,
2
C. E. Blacken,
2
B. R. Brandl,
9
V. Charmandaris,
2,10
D. Devost,
2
G. E. Gull,
2
P. Hall,
2
C. P. Henderson,
2
S. J. U. Higdon,
2
B. E. Pirger,
2
J. Schoenwald,
2
G. C. Sloan,
2
K. I. Uchida,
2
P. N. Appleton,
8
L. Armus,
8
M. J. Burgdorf,
8
S. B. Fajardo-Acosta,
8
C. J. Grillmair,
8
J. G. Ingalls,
8
P. W. Morris,
8,11
and H. I. Teplitz
8
Receivved 2004 March 30; accepted 2004 June 3
ABSTRACT
The Infrared Spectrograph (IRS) is on e of thre e science instruments o n the Spitzer Space Telescope.TheIRS
comprises four separate spectrograp h modules c overing the wa velength ra nge from 5.3 t o 38 m with spectral
resolutions, R ¼ k=k 90 and 600, and it was optimized to take full advantage of the ve ry low background in
the space environment. The IRS is pe rforming at or better than the prelaunch predictions. A n autonomous target
acquisition capability enables the IRS to locate the mid-infrared centroid of a source, providing the information
so that the spacecraft can accurately offset that centroid to a selected slit. This feature is particularly useful when
taking spectra of sources with poorly known coordinates. An automated data-reduction pipeline has been
developed at the Spitzer Science Center.
Subject headinggs: infrared: general — instru mentation: spe ctrographs — sp ace veh icles: instrumen ts
1. INTRODUCTION
The design of the IRS was driven by the objective of
maximizing sensitivity given the 85 cm aperture of the Spitzer
Space Telescope (Werner et al. 2004) and the then -availa ble
detectors. Dividing the optical trains of the IRS into four
separate spectrographs substantially reduced the complexity
and ove rall cost of the system. The result i s four separate
modules, known by their wavelength coverage and resolution
as Short-Low (SL), Short-High (SH), Long-Low (LL), a nd
Long-High (LH). The slit widths are set to k
max
/85 cm, where
k
max
is the longest wavelength for the module. Fur ther in the
geometric limit the monochromatic slit image covers two
pixels. Two Si : As d etectors, 128 ; 1 28 pixels in size, collect
the light in the SL and SH modules, while two Si : Sb detectors
of the same number of pixels are used in the LL and LH
modules. In addition to its spectrographs, the IRS contains two
peak-up imaging fields, which are built into the SL module
and have bandpasses centered at 16 m (‘‘blue’’) and 22 m
(‘‘red’’). A picture of IRS is presented in Figure 1, and Table 1
gives the parameters of each module.
The two long-slit low-resolution modules were de signed for
optimum sensitivity to dust features in the local and distant
universe and are effectively limited in sensitivity by the
zodiacal and/ or Galactic backgrounds. The two cross-dispersed
high-resolution echelle modules were designed to achieve the
highest possible resolution for the given array dimensions, and
they were optimized for sensitivity to emission lines.
In the following sections we describe the instrument, its
operation and calibration, and the reduction of its data using
the pipeline developed at the Spitzer Science Center (SSC).
The Spitzer Observer’s Manual (SOM) provides a more
thorough discussion, and it is up dated freq uently.
12
2. DESIGN, MANUFACTURE, AND GROUND TESTING
The IRS co ntains no moving parts. The diamond-machined
optics are bolted directly to the precision-machined module
housings, and all are constructed of a luminum so that the as-
sembled modules retain their focus and alignment from room
temperature to their operating temperature near 1 .8 K without
any further adjustment. The focal plane assemblies are equipped
with custom plates to interface with the module housing and
account for the individual dimensions and locations of the ar-
rays. Each module contains two flood-illuminating stimulators
to monitor the performance of the arrays .
The low-resolution modules each contain two subslits, one
for the first-order spectrum and one for the second-or der
spectrum. Wh en a source is in the subslit f or the second order,
a short piece of the fi rst order a ppears on the array; this
‘‘bonus’’ order improves the overlap between the first and
second ord ers. The high-resolutio n modules are cross dis-
persed so that 10 orders (11–20) fall o n the array.
The m odules were extensively tested at op erating temper-
ature at Ball Aerospace (Houck et al. 2000) . Light from ex-
ternal sources passed through the Dewar window and a s eries
1
The IRS was a collaborative venture between Cornell University and Ball
Aerospace Co rporation funded by NASA through the Jet Pro pulsion Labo-
ratory and the Ames Research Center.
2
Astronomy De partment, Cornell University, Ithaca, NY 14853-6801;
jrh13@cornell.edu.
3
NASA Ames Research Center, MS 245-6, Moffett Field, CA 94035-1000.
4
Ball Aerospace and Technologies Corporation, 1600 Commerce Street,
Boulder, CO 80301.
5
Department of Physics and Astronomy, University of Rochester,
Rochester, NY 14627.
6
Jet Prop ulsion Laboratory, Califor nia Institute of Technology, MC
169-327, Pas adena, CA 91125.
7
Palom ar Observatory, California Institute of Technology, Pasadena, CA
91125.
8
Spitzer Science Center, California Institute of Technology, MC 220-6,
Pasadena, CA 91125.
9
Leiden University, 2300 RA Leiden, Netherlands.
10
Chercheur Associe
´
, Observatoire de Paris, F -75014 Paris, France.
11
NASA Herschel Science Center, IPAC/Caltech, MC 100-22, Pasadena,
CA 91125.
12
See http://ssc.spitzer.caltech.edu /documents/SOM/ for the most current
version of the SOM.
18
The Astrophysical Journal Supplement Series, 154:18–24, 2004 September
# 2004. The American Astronomical Society. All rights reserved. Printed in U.S.A.
of ne utral density filters at liquid N
2
and He temperatures. An
external optical system projected both p oint and extended
images from either a blackb ody or a monochromator onto
each slit or subslit. These tests verified the resolution, internal
focus, and alignment, as well as the correct centering of the
spectrum on the arrays. Fro m the beg inning of testing through
integration and lau nch, the focus and alignment of the mod-
ules has remained constant on a scale of one-ten th of a pixel.
The monochromator tests a llowed us to map the wavelength
positions roughly on the arrays in prep aration fo r flight.
Uncertainties in the transmission of the neutral-density filter
stacks pr even ted a useful d eterminat ion of the ov erall sensi-
tivities prior to launch. Instead, t he sensitivities of the modules
were predicted by an analytic model of the system. Extensive
observations using a pro totype o f the SH module at the Hale
5 m telescope verifie d the procedures used in the sensitivity
predictions to the 20% level (van Cleve et al. 1998; Smith &
Houck 2001).
When the IRS is operating, all four of its dete ctor arrays are
clocked simultaneously, bu t it is only possible to capture da ta
from one array at a time. Two techniques are used in da ta
collection, the double-correlated sampling (DCS mode) and
raw data collection (Raw mode, or ‘‘sample up the ramp’’).
Science data are collected in Raw mode, while peak-up
employs DCS. In DCS mode following an initial series of bias
boost and re set frames, each pixel is sa mpled, and then after a
number of nondestructive spins through the array each pixel
is sampled again via a destructive read and the difference
between the two samples is stored as an 128 ; 128 pixel im-
age. In the Raw mode after the same initial b ias boost, reset
frames, and first sampling of pixels t here are a number of spin
frames followed by a nondestructive read. This pattern is re-
peated n times. When all pixels are sampled a gain the result is
a 128 ; 128 ; n cube, w here n is the number o f nondestructive
reads, and the values are n ¼ 4, 8, or 16. A fina l 128 ; 128
image is created by calculating the signal slope of eac h pixel
of the Raw cube.
3. IN-FLIGHT OPERATION AND CALIBRATION
Mapping the relative positions of each field of view and the
spacecraft’s Pointing Calibration Reference Sensor (PCRS)
was the most critical step of the in-orbit checkout (IOC)
phase. The positions must be measured to an accuracy of
better than 0B14 radial (0B28 for the lon g-wavelength m odules)
to meet the 5% radiometric requirements. The positions were
measured iteratively, starting with ground-based estimates and
proceeding suc cessive ly through ultra coarse, coarse, and fine
focal plane surveys. Combined with the determin ation of the
focus between the telescope and the slits, this process took
place over nearly six of t he e ight weeks of the IOC period.
The estima ted uncertainties in the final measured pos itions of
the slits and the peak-up a rrays are better than the require-
ments in all cases, ranging from 0B09 to 0B12 .
Our knowledge of the internal focus and optical calibration
of the spe ctral orders were updated in-flight using a combi-
nation of photometric standard sta rs and emission line objects.
The widths of the orders were derived from the zodiacal
light at maximum intensity, while order curvatures, tilts, and
wavelength solutions were rederived from spectral maps of
emission line sta rs such as P Cygni and plan etary n ebulae
such as NGC 6543, NGC 7027, and SM P 083. Figure 2
presents the hyd rogen recombination spectrum of the Be star
Cas, which provides a goo d check of the wa velength cali-
bration for SH.
Flat-fielding and spectrophotometric calibration are deter-
mined from mappi ng and staring observations of well-known
standards, includ ing Dra (HR 6688, K2 III), HR 7310
(G9 III), and HR 6606 (G9 III). Calibration of the low-
resolution modules also used fainter standa rds (e. g., HD 42525 ,
A0 V). Morris et al. (2003) provide more details abo ut the
calibration scheme . Decin et al. (2004) discuss the stars and
synthetic spectra used for in-flig ht spectrophotometric cali-
bration. We are also verifying the c alibration using spectral
templates generated as described by Cohen et al. (2003) and
observations of additional s tanda rd stars.
Fig. 1.—Infrared Spectrograph on Spitzer. The four IRS modules, SH, SL
(which includes the peak-up cameras), LH, and LL are marked. A schematic
of the location of the spectrograph slits on the Spitzer focal plane is presented
in Fig. 2 of Werner et al. (200 4).
TABLE 1
Properties of the IRS
Module Array
Pixel Scale
(arcsec) Order
Slit Size
(arcsec)
k
(m) k/k
Short-Low .......... Si : As 1.8 SL2 3.6 ; 57 5.2–7.7
a
80–128
SL1 3.7 ; 57 7.4–14.5 64–128
‘‘Blue’’ peak-up 56 ; 80 13.3–18.7
b
3
‘‘Red’’ peak-up 54 ; 82 18.5–26.0
b
3
Long-Low........... Si : Sb 5.1 LL2 10.5 ; 168 14.0–21.3
a
80–128
LL1 10.7 ; 168 19.5–38.0 64–128
Short-High.......... Si : As 2.3 11 –20 4.7 ; 11.3 9.9–19.6 600
Long-High.......... Si : Sb 4.5 11–20 11.1 ; 22.3 18.7–37.2 600
a
The bonus orders cover 7.3–8.7 m (SL) and 19.4–21.7 m(LL).
b
This is the full width at half-maximum of the filter.
INFRARED SPECTROGRAPH ON SPITZER 19
The overlaps i n w avelength coverage between the various
modules a nd in ord ers within each module aid the internal
cross-calibration of the IRS. These overlaps were also used to
search for leak s in the order-sorting filters by o bserving
‘‘cold’’ and ‘‘hot’’ sources in two overlapping orders. In the
case of the first-order filter for Long-Low (LL1), which had
partially delaminated prior to launch, we used the combination
of Uranus and Neptune as the cold source s and spe ctropho-
tometric standard stars as the hot sources. At this time the
analysis limits any possible filter leak in LL1 to a maximum of
5% for a Rayleigh-Jeans spectrum. Further on-going efforts
will refine this limit.
The third largest solar proton flare in the past 25 yr
began on 2003 October 28, resulting in an integrated proton
flux through the arrays of 1:9 ; 10
9
protons cm
2
, equivalent
to the 50 percentile dose expected during the first 2.6 yr of the
mission. As a result, the sensitivity of 4% of the pixels in
LH and 1% of the pixels in the other modules have been
degraded. The degree of damage is consistent with the pre-
launch damage experienced by nonflight arra ys that were
exposed to a beam of 40 MeV protons at the Harva rd Cy-
clotron Facility. T he IRS is more sensitive t o d amage than the
other science instrume nts on board Spitzer because it is op-
erating at lower background conditions where even a small
increase in dark current has a sign ificant effect.
The measured responsitivity of the system in flight was on
average 2 times better than the prel aunch model p redic tion.
The detector noise measured in the unilluminated parts of
the arrays are the same as measured prelaunch. Therefore,
potentially in the limit of small signals the sensitivity is on
Fig. 2.—SH spectrum of the Be star Cas. The lower panel shows a continuum-corrected spect rum with the identifications of the hydrogen recombination lines
used to check the wave length calibration of this module. The crosses mark the location of unusable wavelength elements.
HOUCK ET AL.20 Vol. 154
average 2 times better than the mode l predictions. Most of this
increase is due to d esign margin, an assumed factor to accou nt
of errors and nonmod eled effects. The assumed margins were
a factor of
ffiffiffi
2
p
for the short-wavelength modules and a fac tor
of 2 for the long-wa velength mo dules.
Updated estimates of sensitivity are available in the current
version of the SOM. The SSC provides a sensitivity calculator
for the IRS known as SPEC-PET that includes detector noise,
source shot no ise, and background shot noise.
13
At the current
time the sensitivity is limited by systematic effects, such as
noise from the flat-field, fringing, and /or extraction edge
effects, which are not included in the ca lculations of SPEC-
PET. The observed sensitivities are approximately 3 times
worse than would be achieved by the above responsivities and
detector noise. It is anticipated that the realized sensitivities
will improve as the system calibration improves.
4. USING THE IRS
4.1. Spectroscopy
The IRS currently has two modes of operation, the spectral
‘‘staring’’ and ‘‘mapping’’ modes. In both cases the target
is acquired and then observed in a fixed sequence starting
with SL2 (Short-Low second-or der), SL1, SH, LL2, LL1, LH,
skipping any unrequeste d slits .
The target acquisition method is selected by the observer.
One may simply rely on the blind telescope pointing to point
the slits to the specified position on the sky. A second option is
to use the PCRS ( Werner et al. 2004) to c alculate the centroid
of a re ference star, use the offset between the star and the
science target, and move the science target into the slit. A third
and more accurate option is to use the IRS peak-up c ameras to
centroid on a 16 or 2 2 m image of the science target or a
reference star before commencin g the requested spectroscopic
observations (see x 4.2). Using the PC RS or IRS pe ak-up on a
reference star requires accurate coordinates for both the off set
star and the science target.
TheIRSstaringisthemorebasicmodeofoperation,andit
results in observing the target at the 1/3 a nd 2/3 positions
along the slit (its two nod positions), each with the integration
time specified for that slit. The m apping mode steps the slit
parallel and /or perpendicular to the slit according to the
number and the size of the steps specified by the observer for
each slit. The mapping mode does not perform the 1 /3 and 2/3
nod observations that are automatically done in the staring
mode.
Both the staring and mapping modes accept multiple target/
position inputs, as long as all of the positions are within a 2
radius in the sky. Multiple targets are specified in a ‘‘cluster’’
list by either their (1) absolute positions, (2) right ascen-
sion and declination offsets, or (3) parallel /perpendicular slit
offsets with respect to an absolute fiducial position. If using
IRS in spectral mapping mode with cluste r inputs, parallel/
perpendicular slit offsets cannot be used. If the cluster speci-
fication is selected, all sources/offsets in the cluster are
observed with the same slit before proceeding to the next slit.
Use of the cluster specification, as opposed to re peating th e
same observation one source at a time, reduces the observatory
overhead and can lead to substantial savings in total ‘‘wall
clock’’ time.
The SOM describes the IRS observing modes and the
various methods of target acquisition in more detail. In the
sections below, we concentrate on using the IRS p eak-u p
camerastoplaceatargetinthedesiredslitanddescribehow
an observer can use the peak-up fields for imaging a t 16 and
22 m.
4.2. IRS Peak-up
The IRS SL module conta ins two peak-up imaging fields.
Theirfieldofviewis55
00
; 80
00
,withascaleof1B8per
pixel, and their bandpasses are centered at 16 and 22 mfor
the ‘‘b lue’’ and ‘‘red’’ camera, respectively (see Table 1). The
IRS peak-up mod e en ables the placement of a source on a
spectrographic slit or series of slits more accurately than just
using blind pointing of the spacecraft alone. Th e telescope
blind pointing has a positional accura cy of 1
00
(1 rms
radial). An on-board algorithm determin es the centroid of the
brightest source in the specified peak-up field and co mmuni-
cates the offsets required to accurately position the target in
the requested slit to the spacecraft. As long as the coordinates
of a target are accurate enough to place it on the peak-up
imaging field and it is the brightest object in the field, the IRS
will accurately offset to the selected slits. The peak-up images
are supplied to t he observers and can also be used scientifi-
cally if desired.
The allowed ranges of flux densities for blue and red peak-up
point sources are f
blue
¼ 0:8 150 mJy and f
red
¼ 1:4 340 mJy,
respectively. H owever, to avoid ex cessive integration times
and a higher probability of failure, we recommend a flux of at
least 2 mJy for the blue and 5 mJy for the red . The IRS peak-up
algorithm has been op timized for point sources (and indeed
this mode has been the most extensively tested and verified in-
orbit), but an ‘‘exte nded source’’ mode is available for sources
with diameters between 5
00
and 20
00
(e.g., comets). The allowed
flux density range for extended sources, for either the blue
or red mode, is f
ext
¼ 15 340 MJy sr
1
.
The observer can specify two peak-up accuracies: ‘‘High’’
and ‘‘Medium.’’ In the simplest case, when peaking up on the
science target itself, the first move to a slit after peak-up
insures placement to within 0B4(1 rms radial) of the slit
center for High Accuracy and 1B0 for Medium Accuracy. The
High Accuracy value is driven by the 5% radiometric accuracy
requirement for the SH slit (4B7 width). The IRS section of the
SOM provides details on how these accuracy options apply to
multiple slit positioning after peak-up (e.g., for targets in IRS
cluster mode).
In the event that the science target cannot also serve as the
peak-up target (e.g., if its flux is too low or too high), the
‘‘offset peak-up’’ mode giv es the option of peaking-up usin g
a nearby source. The observer is free to use any suitable
source within 30
0
of the science target, but the Spitzer plan-
ning tool (SPOT) also provid es a list of recommended ca n-
didates. Note that this mode requires accurate coordinates for
both the offset and science target. An IRS peak-up returns the
image of the peak-up target.
The PCRS extends the IRS peak-up capability to optical
point sources with visual magnitudes ranging between m
V
¼ 7
and 10. The only mode for PCRS currently available provides
pointing equivalent to the High Accuracy mode of the IRS .
The PCRS peak-up does no t produc e or re turn an image.
4.3. Imagginggwith the IRS Peak-up Arrays
The peak-up fie lds in the SL module provide a means of
obtaining images at 16 and 22 m (5.4 and 7.5 mFWHM,
respectively). The 16 m window in particular is interesting
because the o ther cameras on Spitzer do not cover this
13
See http://ssc.spitzer.caltech.edu /tools/specpet/.
INFRARED SPECTROGRAPH ON SPITZER 21No. 1, 2004
wavelength region , a nd because this wave length region was
used by the Infrared Space O bser vatory (ISO) for most of its
deep extragalactic surve ys. Since these surveys probed the
properties of galactic evolution only up to z 1, IRS can
provide the link between these past results and the new dis-
coveries that will be made by Spitzer at z > 2.
Imaging with the IRS peak-up camer as has n ot been fully
supported for the first year of operations, but the SSC will
provide the ability to obtain large mosaics in a manner similar
to IRAC (F azio et al. 2 004) be fore the end of 2005. However,
an interim method called CHEAP (Cornell High-Efficiency
Advanced Peak-up) has been used extensively to obtain mid-
infrared images. CHEAP is based on an IRS SL staring ob-
servation and uses our knowle dge of the offsets between the
center of the peak-up win dows and the SL slits in spacecraft
coordinates, along with the standard two-position nodding
along the slit, to place a source target in various positions on
the two peak-up windows. The user may select any combi-
nation of exposure times and cycles of the SL to obtain an
image. The d ata are processed by the standa rd SSC pipeline as
though they are standard SL spectroscopic observations. Th e
software that combines and calibrates the produced images is
already available.
As an example, taking a CHEAP image using a standard
60 s SL staring e xposure will produce two images at 16 an d
22 m with an rms noise of 30 Jy in only 720 s of total
time. It is interesting to note that the rms noise of the ISO
observations of th e Hubble Dee p Field (HDF) was just
13 Jy at 15 m and that the faintest source detected in the
HDF at 15 m was 50% brighte r than the sens itivity in the
above example (Aussel et al. 1999). The use of CHEAP to
obtain 16 and 22 m images of a sample of high-redshift
(z > 1:5) submillimeter galaxies is presented by Charmandaris
et al. (2 004 ), and the first 16 m imaging of a Lyman Break
Galaxy at z ¼ 2:79 is discussed by Teplitz et al. (2004).
4.4. PlanninggIRS Observvations
A number of issues sho uld be considere d in the design of
spectroscopic observations u sing IRS. These are addressed i n
detail in the SOM, but we briefly mention a few of the more
important ones in this section.
Although the infrared background is fainter than the ground-
based background by a factor o f 10
6
, it may still be necessary
to take an ‘‘off-source’’ me asure ment of the background. The
low-resolution modules do this automatically in the sense that
both subslits are exposed at the same time so the spectrum
from the ‘‘off’’ slit can be subtracted from the ‘‘on’’ slit to
remove the background. Alternately, the low-resolution slits
are long enough to do the background subtra ction on the slit
itself. However, the high-resolution slits are too short to
subtract the background by differencing the nod positions. If
the con tinuum level is important, off-source integrations using
the same module are required fo r the purpose of backgroun d
subtraction. SPOT provides an estimate of the background that
is helpful in planning for background subtractio n. As a general
rule, the observers should use the low-resolution modules to
measure continuum an d broadba nd fe atures, and the high-
resolution modules to meas ure unresolved lin es.
The peak-up system works extremely well over its range of
parameters. However, the observe r needs to ca refully check
that the flux of the peak-up candidate falls within the specified
limits of the cameras and correctly enter it into SPOT. Fur-
thermore, as already mentioned in x 4.2 the candidate must be
the brightest object within 120
00
,toensurethatitwillbethe
one selected by the peak-up algorithm. There have been
several peak-up failures due to violations of one or the o ther
of the above requirements.
Observing with Spitzer is unlike mos t ground-based infrared
observing. On the ground th e background signal overwhelms
the source by many orders of magnitude. Therefore, the sys-
tem noise is set b y the shot noise from the background. On
Spitzer, the noise (N) is dominated by detector noise for small
signals (S), so N constant. However, as the signal level in-
creases the shot noise in the signal eventua lly dominates: N /
S
1=2
. At the highest signal levels fixed pattern no ise (flat-
fielding errors, etc.) are dominant and consequently N/ S.
At the present time it appears that the flat-fielding term is 1%
to 2%, effectively limiting the maximum S/ N to 50–100.
Similarly, at very low signals the flat-field uncertainty limits
our 1 detection to 0.3 mJy at 16 m for low backgrounds
(see Higdon e t al. 2004a). The online S/ N estimator includes
all but this last term.
5. PIPELINE DATA PROCESSING
As described in detail in the SOM, IRS observations, either
in DCS or Raw mode, are stored as FITS files of what is called
a Data Collection Event (DCE). A DCE contains all the data
obtained by an IRS module since the most re cent destructive
read. The IRS Scien ce Pipeline at the SSC treats each DCE
independently. The processing of DCS data, such as tho se
collected during an IRS peak-up, b y the pipeline is minimal
and all processing is performed on board the spacecraft.
However, for data obtained in Raw mode the pipeline removes
basic instrumental signatures and corrects for variations
in spectral re sponse within a nd b etween spectral orders. A
number of th ese steps, such as corrections/checks for satura-
tion, cosmic rays, dark current subtraction, as well as linear-
ization, are performed in the cube level prior to fi tting a slope
to the sampled up the detector charge integration ramp. Others,
such as correction for drifts in the dark current, and stray light
or crosstalk between the orders a re applied on ce the slope
image has been cre ated.
TheIRSarraydatasuppliedinFITSfilestotheobserverare
organized into four categories: Engineering Pipeline Data,
Basic Calibrated Data (BCD), Browse-Quality Data (BQD),
and Calibration Data. Additional files are also available of the
masks used in the pipeline processing a s well as p rocessing
log and quality assurance files. The Engineering Pipeline data
consist of the raw detector sample images, with some de-
scriptive information in the FITS file headers. The BCD files
are the fundamenta l basis for science analysis, w ith the pri-
mary pro duct being a two-dime nsional slope image of each
DCE in units of electrons s
1
pixel
1
, accompanied by a
header co ntaining the e ssential programmatic information and
the processing, calibration, and pointing history. Additional
BCD data files include uncertainty images that accompany
each slope image and various intermediate images that do not
have all of t he pipeline processing corrections applied. The
BCD also include co-added image data for those observations
with more than on e ramp at a sky location. It is expected that
the BCD files will be used by the IRS obse rvers as the stan-
dard data input for s ubseq uent p ublishable science data anal-
ysis. As an example, i n a stan dard IRS s taring observation in
which the observer h as requested three cycles of 30 s inte-
grations on a s cience target with the SH module will result in
six DCEs in Raw mode. The Engineering Data pipeline will
process the da ta, which will have the form of six data cubes
(three integration cycles at each of the two nod positions along
HOUCK ET AL.22 Vol. 154
