where 
is the FWHM
resolution. The spectral resolution modes are described below.
A. Danks
Hughes STX/GSFC, Code 683.0, Greenbelt, MD 20771
B. Woodgate, R. Kimble, C. Bowers, J. Grady
GSFC, Code 681.0, Greenbelt, MD 20771
S. Kraemer
CSC/GSFC, Code 683.0, Greenbelt, MD 20771
M. E. Kaiser
JHU/GSFC, Code 680.1, Greenbelt, MD 20771
W. Meyer, D. Hood, C. van Houten
Ball Aerospace Systems Div., Boulder, CO 80306
Keywords: spectrograph, MAMA detectors, CCD's
The STIS instrument was selected by NASA in order to maintain a long term
UV spectrograph capability on-board the Hubble Space Telescope (HST),
replacing the currently aging instruments. STIS is
scheduled to be placed in the GHRS (Goddard High Resolution Spectrograph) slot
during the February 1997
refurbishment mission. The spectrograph employs two-dimensional detectors,
two Multi Anode Microchannel plate Arrays (MAMAs) for the UV and a CCD for
the visible. All three detectors have (1024 x 1024) pixel formats. The
MAMAs are true photon counting devices and cover the wavelength regions
from 115nm to 170nm (Band 1) and 165nm to 310nm (Band 2). The MAMAs are
photocathode devices, the Band 1 detector has a CsI photocathode and the
Band 2 a CsTe. The use of photocathodes provides a high degree of solar
blindness. The CCD is used to cover the
wavelength range from 305 to 1000nm.
The spectrograph offers a number of advantages over the current
instrumentation which are described briefly below.
1) Two-dimensional detectors (1024 x 1024 pixels), allowing both long slit spectroscopy and, in the UV, an echelle format, which gives wide simultaneous wavelength coverage.
2) The UV detectors have a 10 to 20 times lower background, (per resolution element) than GHRS, which combined with the 2-D detector capability enables a good estimate of the background, to be made for background subtraction, which in turn provides improved sensitivity.
3) Improved visible efficiency and wavelength coverage (out to 1000nm vs. 700nm) compared with FOS (Faint Object Spectrograph), thanks to the use of a CCD detector.
4) Solar blind imaging in the UV.
5) A coronographic capability for spectroscopy and imaging.
Four spectral bands have been defined in which STIS will carry out
spectroscopy and imaging.
Band 1: 115 to 170nm CsI MAMA detector
Band 2: 165 to 310nm CsTe MAMA detector
Band 3: 305 to 555nm CCD
Band 4: 550 to 1000nm. CCD (same)
Within these bands observations can be carried out at different
spectroscopic resolutions, where
is the FWHM
resolution. The spectral resolution modes are described below.
1) Low resolution R 
500 to 1,000, spectral imaging can be
carried out in
all 4 bands using long slits.
2) Medium resolution R 
5,000 to 10,000, spectral imaging can
be carried out
also in all 4 bands with long slits; however, in this mode more than one
exposure is required per band.
3) Medium resolution echelle spectroscopy R 
24,000 using
short slits
is available in the UV only and will require two exposures in Band 2, to
cover the whole band.
4) High resolution echelle spectroscopy R 
100,000 using short
slits is
available in the UV only and both bands require multiple exposures.
5) Objective spectroscopy R 
26 (at 300 nm) to 930 ( at Lyman
) using
the band 2 detector only.
6) Broadband wide field imaging, in all four bands using a modest
filter
complement.
Each instrument mode is then described by two numbers, e.g., (w,r), where w refers to the wavelength range, and r to the resolving power. A detailed description of the instrument modes of operation are given in Table 1.
Table 1: STIS Spectroscopic Modes, each mode is described as by (w,r) where
the first letter denotes wavelength Band and the second resolution.
The instrument is shown schematically in Fig. 1.
Figure: A schematic of the Space Telescope Imaging Spectrograph.
The instrument carries its
own corrector mirrors to
compensate for the HST aberrations. These will provide sharp imagery at the
field limiting slits. The mirror positions are adjustable to optimize
performance after installation in HST. The instrument has two primary
mechanisms: the slit and grating wheels. The slit wheel has a selection of
slits or diaphragms, short slits for the echelle modes, longer slits for
the single first order modes and a number of special slits. After the
corrected image passes through the slit it is collimated and falls onto
one of the elements of the grating wheel. Mounted on the grating wheel are
12 first order gratings, four are order sorting
gratings, an objective prism,
and a plane mirror.
Selection of the wavelength and resolution determines which grating and
detector should be used. The grating wheel nutates and positions the grating
to point to the appropriate detector path. In the echelle modes the order
sorting gratings direct the light to one of the 4 echelles, and from there the
light passes to a camera mirror and then is imaged onto the
detector.
The slit wheel allows selection of the slit appropriate to the desired science observation. The first order spectral imaging modes can select from 50 arcseconds long slits ranging in width from 0.05 arcsecs to 6 arcsecs, three of which have internal occulting bars for bright object exclusion and four of which are rolled at 45 degrees for planetary use. Examples of these slits are shown in Fig. 2.
Figure: Examples of STIS slits.
Sixteen slits are used among the six echelle spectroscopy modes. They range in length from 0.10 to 10 arcsecs with widths of 0.05, 0.10, 0.20 and 0.5 arcsecs. Camera apertures of 50 x 50, 25 x 25, 6 x 6, and 2 x 2 arcsecs are provided. Coronographic camera apertures of 50 x 50 and 25 x 25 arcsecs with internal occulting bars are also provided. There are also a number of special slits, an example of an occulting mask is shown in Fig. 3.
Figure: A schematic showing a STIS occulting mask.
In
the UV spectral modes the full STIS spectral resolving power is matched to
0.05 arcsecs, and in the visible/IR modes it is matched to 0.1 arcsecs.
The STIS spectrograph utilizes 16 diffraction gratings, 12 of which are used in first order and four of which are echelle gratings used in higher diffraction orders. The first order gratings are all located on the grating wheel and four of the 12 are order sorting gratings for support of the echelles. There is also an objective prism. These 12 gratings support 12 of the 13 spectral modes, serving as the dispersing element in the first order spectral imaging modes and as a cross-disperser's in the four echelle spectroscopy modes. The other spectral mode, is the objective spectroscopy mode of Band 2, light is dispersed by a prism also located on the grating wheel. The prism performance is shown in Fig. 4.
Figure: Objective prism resolution vs. wavelength for Band 2 only.
These gratings have grating constants ranging from 67 to 3600 grooves per mm and blaze angles ranging from 0.67 to 14.7 degrees. Two of these gratings are replicated on slow off-axis segments of parabolas using manufacturing techniques developed for the GHRS cross-dispersers. The other 10 of the first order gratings are replicated on plane grating blanks. All elements on the grating wheel (gratings, prisms, and mirrors) are about 25 mm square, to cover the 19 mm pupil diameter. Each of the four echelle spectroscopy modes has a dedicated echelle grating which is mounted on the optical bench bulkhead (see Fig. 1) supporting the STIS detectors. These plane echelle gratings are the primary dispersing element for their modes are used in conjunction with the appropriate grating wheel mounted cross-disperser grating The echelles range in grating constant from 26 to 99 grooves per mm, in blaze angle from 26 to 68 degrees, and are used in orders ranging from 51 to 432 among the four modes. An example of the echelle formate is shown in Fig. 5 for the mode 1.4.
Figure: Pt-Ne Hollow Cathode Lamp Mode 1.4.
Two of the echelles are 56 x 30 mm and two are 56 x 76 mm in size. Each of the 16 gratings are optically coated to optimize their performance over their intended spectral region. The dispersions and plate scales for each primary mode are given in Table 2.
Table 2: Dispersion and spectral coverage per exposure.
STIS uses two detector technologies, MAMAs and CCDs, in order to optimize
performance over the instrument wavelength range. The STIS CCD will be used
in the visible and near IR, where its (DQE) is a maximum. The STIS chip has
been designed to provide sensitivity in the Band 2 wavelength range, and
it could be used as a back up detector for Band 2. However, operating CCDs in
the UV as primary detectors is difficult. First, the high visible
sensitivity must be suppressed to achieve solar blindness. This can be
done by using Woods filters to suppress the visible. But, they are very
inefficient, bringing the DQE of the combined CCD and Woods filter below
that of a photocathode device and, to date, the Woods filters have proven
unstable. Secondly, the
CCD requires cooling, and contaminants stick to the cool
surface reducing the UV efficiency very quickly. The photocathodes of the
MAMA detectors provide solar blindness and do not require cooling,
and they have a higher DQE in the UV than the CCD and Woods filter combination.
They are photon counting
devices.
A Band 1 MAMA is shown schematically in Fig. 6.
Figure: Schematic of a Band 1 MAMA detector.
Both detectors Band 1 and 2
employ 
entrance windows, and it is this window which is responsible
for
the short wavelength cut-off at 115nm. In the Band 1 case, the opaque
photocathode is CsI which is deposited directly onto the
microchannel plate (MCP). Electrons are liberated from the cathode and
under an applied potential (2000V) are accelerated through the MCP
pores. The MCP consists of a honeycomb-like structure made from glass. The
pores are 12 microns in diameter on 14 micron centers. The walls of the
pores have secondary emitting properties, and as the electrons are
accelerated through the pores, they suffer collisions with the walls and
generate more electrons. A single event can generate typically 4 x 
to
electrons, which emerge from the plate as a cloud and fall onto the
anode array. Usually some positive ions are also generated in the collision
process and these migrate back toward the cathode. They can reduce the
cathode efficiency over time and in order to reduce this ion feedback, the
pores are curved into a C shape. In the Band 1 case, the DQE can be
increased by using a repeller field above the cathode. The repeller pushes
additional electrons emitted from the cathode, which may normally escape
collection, into the nearest pores. Although the DQE is increased by this
process, the resolution is slightly degraded and these effects are
illustrated in the performance Table 3. The Band 2 design differs
slightly, the cathode is semi-transparent CsTe deposited on the inside of
the window, and the window is positioned closely to the MCP for proximity
focusing. In all other respects the two designs are identical. The anode
array is placed in proximity focus behind the microchannel plate. The array
design consists of two layers of anodes (x and y) in an encoded pattern,
which allows
the charge cloud position in x and y to be determined using only 132 charge
amplifiers. A decoder interprets the output of the charge amplifiers and
assigns events to a position (x, y) and time of arrival t. The anode array
element separation is 12.5 microns, and the pixel size is 25 x 25 microns.
The decode chip can be used to event center to 12.5 microns. The MAMA will
be used in this latter mode providing 12.5 x 12.5 micron pixels, these are
known as ``high res'' pixels and the detector then has 2048 x 2048 ``high
res''
pixels. The 12.5 x 12.5 pixels can be summed to recover the 25 x 25 pixel
formate. Experience shows that the ``low res'' image has better photometric
quality but that the ``high res'' image provides higher spatial or spectral
resolution.
The use of photon
counters has some implications for the user. There is an overall
brightness or global count rate limit; this is set by the on-board computer
and is 300,000 counts per second. There is also an individual pixel
brightness limit which for flight is set at 50 counts per second per pixel.
This limit is due to the electron recharge time of the individual pores or
dead time. If events arrive closely spaced in time, at the same pixel
position, then the pore does not have sufficient time to recover to produce a
charge
cloud of sufficient number to overcome the charge amplifier threshold and the
event
is not recorded. These events are not recorded and the detector departs
from linearity. The 10% roll over point for each detector is different, but
typically is measured in the laboratory at 200 to 300 counts per second per
pixel. The STIS specification
of 50 counts per second per pixel it therefore easily met. Typically MAMA
observations are carried out in the ``accumulate mode'', in which a 2048 x
2048 image array is incremented as photons arrive. Note the pixel counters
in buffer memory are 16 bits (65,535 counts) deep. The instrument software
provides on-board Doppler compensation to re-register detected photon
events to the proper half-pixel bin before they are accumulated in memory.
This is necessary only for the echelle modes. MAMA data can also be
recorded in ``time-tag'' mode. Here each individual event is stored with its
time of arrival. The time resolution is 150 microseconds. Limitations are
imposed on the maximum rates at which data can be taken.
These limits are imposed by the tape recorder memory size and by the speed
of the data link. Finally, a sub-array can be defined which can not be
smaller than 8 x 8 pixels. Some typical detector performance specifications
are given in Table 3.
Table 3: MAMA detector measured performance capabilities and flight
requirements.
The CCD detector extends the STIS wavelength range through the visible
portion of the spectrum, covering 305 to 1000nm. The CCD detector has 21 x 21
micron pixels, and 1024 x
1024 pixels format, supplied by SITe (Scientific Imaging Technologies). It is
backside-illuminated, buried channel device operating in the inverted mode,
with an overscan region included for bias removal. The design includes the
Multi-Phase Pinned (MPP) technology, which minimizes the surface state
contribution to dark current.
The CCD is cooled to -80C by a Thermo-Electric Cooler (TEC) in order to
achieve a low dark current. The cooled CCD and preamplifiers are
enclosed in a vacuum housing which is manufactured from a high density
material to provide radiation shielding from the South Atlantic Anomaly
(SAA) protons. The housing has a 
window which is close to 20C
during
operation. This should minimize the condensation of UV absorbing contaminants.
The CCD detector has a wide range of operating modes. During
non-integration periods, basic ``flush'' operations will be used to condition
the CCD, with continual flushing the normal mode. A number of bias frames
may be combined in memory, providing data for background or fixed pattern
removal. Twenty pixels of overscanned data per quadrant also may be used
for background removal.
Integration periods are selectable, in increments of 0.01 seconds up to 60
mins. for long total integration times, several shorter exposures will
generally be taken, to permit the removal (through vetoing) of cosmic ray
hits.
There are four readout amplifiers, one for each quadrant of the detector.
Readout of one quadrant, two quadrants, or the full detector may occur
through any given amplifier. Data from rectangular portions of the image or
one sub-array (size from 16 x 16 pixels upwards) may be read out and
stored. There are provisions for one programmable sub-array.
The read-out electronics have selectable gain to provide for differences in
image intensity beyond the dynamic range of 65,535 (16 bits). The
gain is selectable from 1 electron/bit to 8 electrons/bit. In addition, by
combining or ``binning'' pixels on-chip before readout, extremely low surface
brightness image data can be accommodated. This will be used, for example,
in spectroscopy of faint galaxies.
Doppler correction for spacecraft motion is provided primarily by timely
dumps of images or sub-array data to memory or to the spacecraft. As the
highest resolution offered by STIS in the visible is only 10,000, Doppler
correction is less critical than for the high resolution MAMA modes.
The mini channel implants in the parallel channels, and inverted MPP
operation will minimize the effects of both lattice dislocations and
ionization damage. To offset residual effects of long term degradation, the
design provides the capability to vary clock timing and phasing, to
operate at lower temperatures and to change critical voltage levels.
The performance of the three top ranked CCD flight candidates are shown in
Table 4.
Table 4: CCD performance data.
Normally the object is acquired in mode 4.6, using the CCD camera with a 50
x 50 arcsec field. Two short exposures are taken to enable vetoing of cosmic
rays. An acquisition image will usually be provided with the data. The FGS
has a pointing accuracy of 
2 arcsecs, and the target is usually easily
identifiable in the field. Once identified the object is positioned via
small angle maneuvers to the
center of the chosen science mode slit position. Two more exposures are
made and then the calibration lamp is flashed through the slit using the HITM
(Hole in the
Mirror) system to confirm the exact slit position. A further peak up on the
image is then performed. The peak up procedure depends on the detector
selected, for instance for the MAMA peak up is carried out in dispersed
light to avoid overexposure. However, acquisition can be expected to take
approximately 20 mins.
STIS incorporates a small number of filters for use with its imaging modes. The filters are in hand and the measured quantities are given in Table 5 below.
Figure: Filter specifications.
As illustrated in Fig 1, STIS has a number of on-board calibration lamps:
for wavelength, two Pt/Cr-Ne to be used with the HITM, one Pt/Cr-Ne, and
for flat fielding, one Tungsten lamp assembly with four bulbs, one Kr and
one 
lamp.
Each STIS grating has been measured for efficiency and scattered light properties in the DGEF (Diffraction Grating Evaluation Facility) at GSFC. The grating performance has been measured at its appropriated wavelength and operating angles. Similarly, the performance of the remaining optical components have been measured, allowing the implementation of a realistic computer simulation of STIS to be written. STIS sensitivities are given below in Table 6, for a 1-hour observation. These calculations do not include slit loses.
Table 5: STIS sensitivity in a 1-hour observation.
The instrument is in assembly, and alignment, all optical elements are mounted on the optical bench, and alignment of most of the modes is complete. The flight CCD has been selected and is under test. The flight MAMAs are soon to be selected. In parallel data analysis software is being written to ensure wavelength identification and calibration. Finally, an example of a stellar spectrum is shown in Fig. 7 calculated by Derck Massa, illustrating the STIS performance in Mode 1.4 (the same mode illustrated in Fig. 5).
Figure: Spectrum of a B0Ib star, V=9, E(B-V) = 0.1, Mode 1.4, S/N =100:1,
Integration time 1.9 hours.