VIRUS-P: Camera Design and Performance
Joseph R. Tuftsb, a, Phillip J. MacQueena,
Michael P. Smitha, Pedro R. Seguraa , Gary J. Hilla, Robert D. Edmonstona
a
McDonald Observatory, The University of Texas at Austin, 1 University Station C1402, Austin,
TX, USA 78712;
b LCOGT, 6740 Cortona Dr., Ste. 102, Goleta, CA 93117
ABSTRACT
We present the design and performance of the prototype Visible Integral-field Replicable Unit Spectrograph
(VIRUS-P) camera. Commissioned in 2007, VIRUS-P is the prototype for 150+ identical fiber-fed integral field
spectrographs for the Hobby-Eberly Telescope Dark Energy Experiment. With minimal complexity, the gimbal
mounted, double-Schmidt design achieves high on-sky throughput, image quality, contrast, and stability with novel
optics, coatings, baffling, and minimization of obscuration. The system corrector working for both the collimator
and f / 1.33 vacuum Schmidt camera serves as the cryostat window while a 49 mm square aspheric field flattener sets
the central obscuration. The mount, electronics, and cooling of the 2k ⨉ 2k, Fairchild Imaging CCD3041-BI fit in
the field-flattener footprint. Ultra-black knife edge baffles at the corrector, spider, and adjustable mirror, and a
detector mask, match the optical footprints at each location and help maximize the 94% contrast between 245
spectra. An optimally stiff and light symmetric four vane stainless steel spider supports the CCD which is thermally
isolated with an equally stiff Ultem-1000 structure. The detector/field flattener spacing is maintained to 1 µm for all
camera orientations and repeatably reassembled to 12 µm. Invar rods in tension hold the camera focus to ±4 µm
over a -5-25 ºC temperature range. Delivering a read noise of 4.2 e- RMS, sCTE of 1 – 10-5 , and pCTE of 1 – 10-6 at
100 kpix/s, the McDonald V2 controller also helps to achieve a 38 hr hold time with 3 L of LN2 while maintaining
the detector temperature setpoint to 150 µK (5σ RMS).
Keywords: Cryo-camera, VIRUS, VIRUS-P, CCD, HETDEX, Spectrograph, Hobby-Eberly Telescope
1. INTRODUCTION
Traditional astronomical instruments are purpose built as monolithic prototypes for unique telescopes. As such, a
large fraction of the instrument cost is expended on engineering effort. With a seeing limited telescope, increases in
aperture demand faster focal ratios, larger focal planes, and larger instrument optics. The current generation of 10 m
class telescopes demand instrumentation that is approaching a limit to what is reasonably achievable with traditional
design.
The Hobby-Eberly Telescope1 (HET) is an existing 9.2 m aperture telescope located at the McDonald Observatory
in West Texas. We are beginning an ambitious scientific program using the HET to map, in three dimensions, the
distribution of ~106 Ly-α emitting galaxies (LAE) over a volume of ~9 Gpc3. This program, the HET Dark Energy
Experiment2, 3 (HETDEX), will constrain the expansion history of the Universe to 1% and provide significant
constraints on the evolution of dark energy. It requires a new efficient highly multiplexed spectrograph, the Visible
Integral-Field Replicable Unit Spectrograph4, 5 (VIRUS), as well as a major increase in HET field of view6, 7, from 4
to 20 arcmin, to support it.
VIRUS is an instrument comprising some 150 unit spectrographs each fed with a separate ~250 channel fiber
integral field unit (IFU) and having its own collimator, VPH grating, and camera. These spectrographs are mounted
in 2 – 4 thermally stabilized enclosures situated in a gravity neutral location on the telescope structure.
Unlike the vast majority of astronomical instrumentation, VIRUS contains many smaller, individually realizable,
identical units. The quantity of unit spectrographs allows the development of VIRUS to follow a standard
production model. As such, we have constructed a prototype unit, VIRUS-P, and it is in regular use on McDonald
Observatory's 2.7 m Smith reflector dominating the dark-time usage of the telescope. It has also been demonstrated
on the HET. The primary motivations for VIRUS-P are to understand the trades between material, design, and
assembly cost, and to identify technical and scientific risks. This paper presents the design and performance for the
VIRUS-P camera. The overview of the VIRUS-P system, details of system mechanical design, the HET wide field
upgrade required to support it, and fiber and grating testing are left to other works.
1.1 Design Constraints
The design of VIRUS flows strongly from the requirements for HETDEX, to maximize the number of LAEs
detected in a set observing time, and to span sufficient redshift range to survey the required volume. These science
requirements flow down to the following technical requirements for VIRUS-P: Coverage of Δz~2 and coverage into
the ultraviolet to detect LAEs at the lowest possible redshift.
VIRUS-P is designed for 340 < λ < 570 nm (1.8 < z < 3.7 for the galaxies of interest). The resolution, R~800, is
chosen to match the line width of LAEs to maximize detectability. The minimum acceptable throughput on sky
(including atmosphere) ranging between 5% at 350 nm and 15% at 450 nm to reach sensitivity of 3-4e-17 erg/cm2/s
in 1200 s on HET. It is to use a low read noise detector, <3 e-, to achieve sky-background dominated observations in
300 seconds. It must be insensitive to ambient temperature variations, but because of the installation location on
HET, it need not be insensitive to gravity variations. Finally, the cost savings can only be realized through design
simplification.
OPTICAL DESIGN
The collimator-grating-camera design is strongly based on a traditional Schmidt camera with the interface between
the camera and collimator assemblies at the aspheric corrector plate (surface A in the right panel). The f/3.65
collimator is a Schmidt camera used in reverse to project the system pupil at the plane of the VPH grating8, 9, which,
in the standard configuration, has 831 fringes/mm and is sandwiched between fused silica plates. A fold mirror is
introduced for compactness and to allow flexibility in the design of the integral field unit interface. The extremely
large ratio of field to stop size is achieved in the conventional manner by making all optical surfaces concentric
about the intersection of the chief ray and pupil. The resulting field curvature existing at the fiber feed is fit by a
concentric cylindrical fan of fiber outputs immersion coupled to a single anti-reflection (AR) coated cylindrical lens
effectively creating a slit-like input to the spectrograph. All optics in the system use high throughput dielectric
coatings specified for 370 – 640 nm. All collimator optics are fused silica and edged slightly larger than the
illumination shape at their respective surfaces. The spherical primary is further light weighted on the back for
minimum system mass.
Figure 2: Optical design of the VIRUS-P spectrograph. The standard setup is shown, covering 340 – 570 nm at R=850.
The linear fiber input is arrayed out of the page. The physical layout of the optics is shown right.
Owing to the fact that the VPH grating is not at the camera stop, the f/1.33 camera accepts a 125 mm beam, and is
slightly faster than the 115 mm pupil would otherwise require. It is also a concentric Schmidt design with all fused
silica elements. It uses a light weighted spherical primary, an aspheric corrector at the pupil and a spherical front
aspheric back lens to flatten the field for the 43.44 mm 2k2 15 µm pixel CCD diagonal. Each fiber is reimaged to
~4.9 pix at the CCD. Dispersion in the standard configuration is 0.11 nm/pix and the oversampling affords the
possibility of 2 X 2 or 2 X 1 binned operation for faster readout. It was found that the correction for both the
collimator and camera could be achieved with a single aspheric surface on the camera corrector. The throughput
losses are dominated by CCD QE and obscuration, and the image quality is superb and higher than could be
achieved with the more complex dioptric designs investigated.
The camera's fast focal ratio dictates the tightest system tolerances are in the elements with the most power. That is,
the tip, tilt, and spacing of the camera primary and field flattener relative to the CCD. The root-sum-square (RSS)
tolerances for these spacings are of order 15 µm. The VIRUS-P camera was designed to a fairly tight (1.5 µm)
stability requirement for detector positioning.
2.1 Baffles and Masks
Figure 3: While baffle profiles for VIRUS-P were modeled using the footprint analysis feature of Zemax and shapes set
by the sum of profiles from three principle dispersion modes for the instrument, the focal plane mask form was
derived by simple extrusion. Steps were then cut into the focal plane mask with dimensions designed to prevent
all high angle of incidence scattered light from directly reaching the detector.
Critical to the reduction of scattered light induced photon shot noise, certain baffles and a detector mask are key in
this design. All baffles are milled using conventional CNC techniques with 10 – 15 µm thick knife edges, and
blackened. The critical baffles are located at the aperture stops at the grating, cryostat window, and camera primary,
in addition to the detector plane on the outer rim of the spider. A mask is within 1 mm of the detector surface
masking the highly polished mounting structure underneath.
All baffles, and in fact all interior camera surfaces are painted in a special clean filtered spray booth with Alion
MH2200 and then vacuum baked for 4 h at 150 ºC to cure paint and remove xylene based contaminants which
would otherwise degrade the vacuum and put the detector at risk of contamination. Many samples that went through
this process were tested for contaminants with a Stanford Research Systems RGA200 mass spectrometer. No
detectable contaminants were found at the 10-9 mbar L/s level that is the noise floor of the instrument used.
2.2 Camera Alignment
To establish the optical axis, the camera setup relies on accurate positioning of the detector and field flattener in the
detector spider assembly.5 Although the field flattener lens met specifications, both the Steward-Orbit and Fairchild
detector packages showed significant wedge between detector surface and package back. The thermal isolator was
trimmed to compensate as described below.
Alignment of the spectrum of the central fiber with rows on the detector is done by rotating the camera in its mount
to the spectrograph10 with a simple tangential adjustment screw. Alignment of wavelengths to the columns requires
a small amount of rotation of the grating in its cell.
Figure 4: Image quality of complete VIRUS-P instrument. The left panel shows a full field exposure of a HgCd line
source. The right panel shows a zoomed in view of the upper left corner where the image quality is worst.
The final alignment of VIRUS-P is in the tip, tilt, and piston of the camera mirror using imaging feedback from
observing spectral emission line lamps illuminating the IFU. The image quality is analyzed as a function of fiber
position and wavelength to provide the best overall focus and alignment of the spectra on the detector. On VIRUSP this is done with a cold detector using a special camera back described below. Although this is a lengthy and
interactive procedure, it provides exceptional image quality while maintaining all the fibers on the detector.
Production VIRUS units will require each camera, collimator, and IFU be independent, allowing interchange of
subassemblies without requiring realignment. This will be achieved using fiducial subassemblies, test jigs, and more
automated analysis software.
3. MECHANICAL DESIGN
The camera is an evacuated Schmidt design with the aspheric corrector acting as the cryostat window, and the
detector located at the camera's internal focus suspended on a stainless steel spider assembly. We did explore a
design similar to the HET/LRS camera, SF1, which uses a detector sized cryostat, for VIRUS-P, but we opted for
the reduced complexity of the larger cryostat and increased throughput provided by the more compact detector
arrangement.
3.1 Camera Housing
Although we explored the idea of housing the instrument in a temperature controlled environment, an option
afforded by the installation location of the VIRUS system on HET, we opted for the more conservative approach of
requiring VIRUS-P camera stability over the useful temperature range, -5 < T < 25 ºC. This constraint required a
low CTE spacer between camera primary and detector.
In an effort to reduce part count, we initially chose a conceptually simple Invar cryostat housing, but found this
exceedingly difficult to requisition and fabricate. The only feasible fabrication method for small quantities involved
roll forming two half cylinders from Invar plate, seam welding the two half cylinders together, annealing the
cylinder to remove material variations in the heat-affected zone, and then post machining all the critical features.
This method failed because the rolled cylinder did not have enough uniformity to allow post machining and the post
machined welds were not vacuum tight.
Shortly into the manufacturing process, we abandoned the low CTE housing in favor of a significantly less
expensive aluminum housing with Invar spacing rods between the detector assembly and camera primary. The inner
surfaces are masked at the various mounting points of sub assemblies and painted in the same manner as the baffles
previously described.
This same housing constrains the spacing between the fused silica window and spider assembly but the effect of
CTE based spacing error results in a second order negligible effect on image quality, however it will result in a very
small change in aberrations with temperature possibly manifested as an apparent shift in magnification. The
differential thermal expansion between the housing and window is absorbed in it's mount which is effectively the
Viton O-ring which acts as a cushion and seals the vacuum environment.
Figure 5: Sections of the assembled VIRUS-P camera. Note the Invar spacing rods, thermal link to the Dewar,
penthouse electronics module, mirror assembly, spider assembly, and cryostat window. All baffles except the
window baffle are shown as well.
The cryostat front flange serves as the interface between the camera and the collimator. This cylindrical flange
mates to a machined cylindrical pocket on the collimator base plate providing the radial constraint. Rotation is
constrained at this same interface during optical alignment with imaging feedback. Other interfaces include three
back bored pockets for internally mounted Detoronics vacuum rated PT connectors, a surface mounted seal for the
LN2 Dewar, and a sealing back cap.
For optical alignment at operating temperature, the sealing back cap can be replaced with an actuated version which
uses three axial ferrofluidics feedthroughs for adjustment and three radial feedthroughs for locks. Although
ferrofluidics couplings are not necessary in this case, they perform well and we had them on hand. The six
feedthroughs have shafts coupled to a hex key inside the vacuum. Each is mounted on a bellows with enough range
to allow the hex key to engage and disengage from the adjustment and clamp screws. Once the length of the hex
keys is correctly set the lid can come off and on freely, as the spring force of the bellow disengages the hex keys,
and when the housing is evacuated the bellows are compressed, inserting the hex keys into the heads of the adjuster
screws.
Alignment with a cold detector was deemed necessary as dark current was too high to provide imaging feedback
and align the camera properly while warm. The symmetry of the Ultem spacer, which flexurally absorbs the cold
block thermal expansion, should conceptually allow the instrument to remain in focus as the detector cools to
operating temperature. That said, without careful material matching of all cold components and flexural isolation of
components at different temperatures, it will prove impossible to maintain the desired constraint on detector
position over a wide range of ambient temperatures.
3.2 Primary Mirror Support Assembly
An Invar puck is bonded to the back of the fused silica camera primary with Epotek 301-2 providing a mounting
surface for the mirror. The puck is fastened to a triangular adjustment plate with one shoulder screw in the center to
constrain position, and another in a radial slot to constrain rotation. Three adjustment screws on this adjustment
plate are held in tension by a stack of Bellville spring washers supported off of a reference mounting ring which is
in turn supported on four radial flexures around its circumference. The radial flexures allow differential thermal
expansion between the housing and ring, constrain the assembly's x/y position, and serve as spring preload holding
the Invar metering rods which support the entire mirror assembly in tension.
This design was a retrofit falling out of our original choice of using an Invar housing. As such, it is more complex
than necessary and the details will be simplified in VIRUS while maintaining the functionality described.
Figure 6: Spider and camera mirror assembly. The left panel shows details of the tip-tilt adjusters. The adjuster plate is
fitted with precision studs mounted on integral flexures, the studs are inserted into the bearings and three springloaded screw tip-tilt adjusters are used to align the mirror. Since the mirror is spherical, only tip, tilt and piston
adjustments are necessary. Once aligned, the clamp on the split-spherical bearing provides an orthogonal clamp.
The center panel shows the primary mirror support assembly. The right image shows details of the flexure
constraining the mirror mount plate radially to the camera housing
3.3 Spider Assembly
The detector and field flattener are suspended in the beam with a stainless steel spider that carries the copper cold
finger and rigid flex printed circuit along the same arm. Initially the spider was fabricated from Aluminum for CTE
matching to the housing. This spider was designed to support an available but obsolete Steward-Orbit device which,
due to aging, had significantly poorer performance than required. In early 2007, the spider design was slightly
modified to support a Fairchild Imaging CCD 3041 BI, and at that time we opted to fabricate the spider out of
stainless steel with differential CTE absorbed in integral flexures designed into the outer edge of the spider. With the
rigidly coupled Al spider, we observed an unrepeatable 1.5 pixel image shift in the first few large temperature
swings in a given night. The flexures on the stainless spider reduce that motion below 0.03 pix/ºC despite the larger
CTE mismatch between housing and spider. A novel radial construction in these flexures allows the slightly
oversized radial constraint to be reduced for installation and preload in operation.
The spider assembly has been optimized to reduce the obscuration to 23% roughly independent of field angle to
maximize throughput and minimize unwanted diffraction. The low level of focal ratio degradation in the fibers
results, to a large extent, in the preservation of the central obstruction from the telescope and hence reduces the
effect of the detector obstruction. Each of the vanes is identical, 1 mm thick by 20 mm high by 59 mm long. To
minimize off-axis obscuration, angular sections were removed from the top and bottom of the vanes with minimal
consequence to stiffness. Although the stainless spider could achieve performance goals with only 0.3 mm thick
vanes, we no longer needed to minimize thermal conductivity, and were uncomfortable with the fragility of the
spider during rough assembly handling. During assembly spider stiffness was measured (0.05 µm/N) and compared,
with excellent agreement, to the finite element analysis used in the design.
The spider locates the detector assembly and field flattener lens near the beam center and camera focus. The field
flattener lens is edged at the corners to produce four coplanar flats. These flats rest on protruding spider lands which
constrain the lens and is pulled down onto the hub with four spring loaded spherical clamps. The detector assembly
is manually trimmed to position it perpendicular to the field flattener's optical axis. Normal machine tolerances
position the detector laterally to within the centration tolerance.
The OFHC copper cold finger creates a larger obstruction than the remaining vanes combined. It's cross section is a
truncated diamond with four sides 6 deg into vertical which allows allows all field angles to see the same
obscuration of approximately 5 mm. This shape creates a dead region above and below the cold finger which carries
the stiffened flex on the sky side and one of the vanes on the mirror side. The copper is highly polished and gold
plated and runs through an electropolished stainless steel tube. The Stainless steel tube acts as a radiation shield on
the ~80 K Cu part and is painted flat black on the outside in the manner previously described. It is supported at only
one end by a complicated under constrained G10 thermal isolator. The complexity of the isolator and cold finger fell
out of the chosen Dewar location. The cold finger could be decreased in length, complexity, and substantially in cost
with a more suitable Dewar location.
Figure 7: Left is the spider and focal plane assembly. On the right is the finished spider. Although it looks complex,
excepting the radial holes used during installation, the spider can be machined in only three setups. Two on a
conventional 3-axis CNC mill and a third on a wire EDM machine.
3.4 Detector Assembly
The Fairchild CCD 3041 BI device is inserted into a custom socket constructed from two banks of Mill-Max 1031
low insertion force sockets soldered to a four layer polyimide circuit board. Each bank of sockets is constrained and
electrically isolated by a G10-CR block rigidly mounted to the header board. The large number of socket contacts
provides a repeatable, reliable alignment in x/y between the detector and header board. The header board is rigidly
mounted to a Au plated OFHC Cu cold block on four lands. The detector is then constrained to the cold block on
four lands each of which is at the corner of the thermally conductive region on the back of the detector package.
Flexural clamps hold the detector down. The cold block assembly is mounted to an Ultem part which thermally
isolates and mechanically constrains it to the spider assembly. A final G10 component simply ensures that the
Hirose connectors between the header board and vacuum flex circuit remain mated.
Figure 8: The left photo shows the various parts of the detector assembly. In the sectioned view in the center, the
sprung clips that hold the detector to the cold block are shown. The right photo shows the completed detector
assembly ready to be installed in the spider.
Given a material and geometry, there is a direct trade between stiffness and thermal conductivity. Materials with a
high ratio of Young's modulus to thermal conductivity are typically exotic plastics or composites. With the Al spider
and Steward-Orbit detector we used a G10-CR spacer component. This design was tuned for a 250 mW load and
resulted in a 0.32 µm/N z stiffness, and a 0.08 µm/N x/y stiffness. The severe geometrical constraints on this design
led to acceptable but relatively poor performance.
With the Fairchild device and the Stainless steel spider, the geometry changed and the same performance could not
safely be achieved with a G10 component, we experimented with and ultimately settled on the design of Ultem part
pictured. It has a thermal conductance of 241 mW, a z stiffness of 0.18 µm/N, an x/y (dispersion/spatial) stiffness
of 0.08 µm/N. As the detector assembly weighs less than 5 N, the total motion in a non gravity neutral situation is
about 2 µm worst case in z, and about 1/20 of a pixel in translation.
CCD surface flatness was measured to be 25 µm on our device however this is independent of the reference back
surface which showed a significant wedge. The full assembly was built up and the wedge measured with an
Olympus measuring microscope. The measured wedge was fit and extrapolated to the outer corners of the Ultem
part where a surprisingly deterministic process of sanding and measuring was employed. After two iterations the
CCD active surface was measured parallel to the field flattener lands on the spider to 27 um. The authors feel
experience and a faster measuring device based on a high angle of incidence position sensor could be effectively
employed on each VIRUS camera and would be substantially less complicated than an adjustable mechanism. It
might also be possible to specify an acceptable wedge tolerance from the CCD manufacturer.
Figure 9: Finite element analysis results showing from left to right, the 48 µm shift due to the temperature difference
between the Cu cold block and the ambient spider, the model used to balance and measure a total thermal
conductance of 241 mW (60 mW per corner), and the 0.08 µm/N spring constant in the spatial direction. The
diagonal connecting lines are modeled as Cu and a temperature gradient is applied across the Ultem part.
Additionally maximum stress, z displacement, and x displacement were also modeled. The principal gains of the
Ultem part over the G10 part are due to geometry.
4. THERMAL DESIGN
VIRUS-P is designed to be a low background spectrograph. Cooling the detector to -110 ºC achieves an extremely
low dark current (~1 me-/s/pix) and minimizes the cosmetic effects without significantly degrading the CCD's
charge transfer efficiency. Typically this temperature is chosen because it also represents a compromise between red
QE and dark current, but in VIRUS-P the red cutoff of the dielectric coatings used on the optics is blue enough that
other undesirable effects dominate before we see decreased QE due to cold operation. Consequently, we chose a
temperature of -110 ºC (instead of -100 ºC) because it represented the operating point with the minimum required
heater current, and best thermal performance.
Thermal isolation of the CCD from the ambient environment is with high vacuum and by standing the cold block
off from the spider assembly with an Ultem spacer. The dominant thermal loads are an estimated 750 mW radiative
on the detector surface, 300 mW conductive through the Ultem spacer, and a small radiative load on the remainder
of the polished and gold plated detector package. The aspect diameter of the Ultem part (i.e. its thermal section if it
were a cylindrical rod of diameter D and length l) øa = (π D2 / 4l) = 1.46 mm. In addition to these loads, the
temperature controller nominally supplies a total of 250 mW to four identical Vishay G-2 resistors potted in the cold
block to maintain the temperature at set point. The 1.3 W total load is removed through a tuned cold link
constructed from copper braid manufactured for ribbon cable shielding. The heat is ultimately dumped into a 3 L IR
Labs LN2 dewar. The cryogen hold time is > 36 h and therefore dominated by radiative load on the cryogen
container itself. One of the Dewar ends also supports the KF25 flanged vacuum valve, the Pfeiffer compact full
range gauge, and the Varian noble diode ion pump.
Cold block temperature is sensed with an Omega 500 ohm thin film Pt RTD with a four wire Kelvin connection to
the two wire device as close to the recommended 1 mm lead length as possible. The sense and excite traces for this
circuit are mirrored on adjacent planes through all of the vacuum circuitry to minimize spurious differential signal.
The thermal system is controlled with the McDonald V2 temperature controller which servos the heater power in
response to the RTD. Ultimately the RTD is referenced to a very precise Thaler voltage reference in the controller
which has 1 ppm/ºC drift. The servo error signal, typically 70 µK 3σ RMS, is reported to the user during exposure
and in FITS headers.
Figure 10: The McDonald Observatory V2 temperature controller and the gold plated Cu cold block with four heater
resistors and one Pt RTD.
During the design phase we experimented with two other methods of thermal isolation. Initially we had hoped the
relatively low thermal conductivity of the vanes in the stainless steel spider would be enough. Although larger than
desired, it would have worked, except the thermal load was dominated by radiation between vacuum compatible
optically black paint used on the housing walls and spider vanes. A different optical design, affording a low
emissivity coating on the housing and vanes may have made the metal insulator acceptable. Thermal isolation via a
kinematic mount of sapphire balls on carbide faces was also explored, but was complicated and did not achieve the
desired level of thermal isolation.
4.1 Vacuum design
At operating temperature, and in fact any sensible CCD operating temperature, the VIRUS-P camera requires a
vacuum to operate properly. The assembly's compact design, in the worst case, puts ambient temperature
components 0.5 mm from the cold detector components. With such a tight spacing, the only suitable insulator is
vacuum. Although a modest pressure of 10-3 mbar would put the vacuum system into the molecular flow regime for
such tight spacing, it was deemed necessary to operate at least three orders of magnitude lower in pressure to ensure
the thermal system would be dominated by radiative loading.
In order to minimize the obscuration as much as possible, the entire VIRUS-P camera is within an evacuated
cryostat. Where possible through holes are used instead of blind holes with vented fasteners. However, it is noted
that vented fasteners are often substantially cheaper than the extra machining setups required to vent otherwise blind
holes. In all, vacuum performance is excellent, <10-7 mbar, and is maintained in operation with a combination of a
LN2 cooled activated charcoal cryopump for water, CO2, N2, and O2, and a Varian noble diode ion pump for Ar.
Although VIRUS-P is thermally cycled to ambient when it is not being used, it is expected that a similar design will
allow VIRUS to maintain vacuum as long as it is maintained at operating temperature.11
Proper cleaning is necessary to achieve a clean vacuum. Prior to assembly all parts, including painted and plated
ones, were cleaned ultrasonically in a 5% solution of Brulin 815 GD at 60 ºC for 5 m and rinsed with distilled water.
Excess moisture was removed with pressurized UHP N2. Excepting the vacuum housing because of its size, parts
were then ultrasonically cleaned in Hexanes for 5 m, followed by 2-propanol (IPA). Finally all parts were vacuum
baked at a the maximum temperature suitable for their material for 2 – 4 h. Viton O-rings were also cleaned in IPA
and vacuum baked separately from the other components. They were then lubricated with an extremely thin film of
Pfeiffer ultra low vapor pressure vacuum grease.
5. ELECTRONICS
VIRUS and by consequence VIRUS-P is designed for a 2k x 2k x 15 μm pixel CCD. Present yields on this
relatively small format detector are very high, and they represent the lowest cost per pixel for science-grade devices.
Working into the near-UV, low readout noise, and good UV-blue QE are important considerations. Available
detectors from Fairchild (CCD3041-BI) and E2V (CCD230-42) meet the needs of the instrument, as could a custom
foundry run. The Steward-Orbit CCD originally chosen for VIRUS-P because of availability had higher readout
noise (around 6 electrons at 100 kpxl/s) than chips being considered for the final design of VIRUS. Subsequently in
summer 2007 the camera was upgraded to use a Fairchild Imaging CCD3041-BI detector.
5.1 Controller
Although the scale and architecture of the full VIRUS instrument dictates a purpose built controller, with the
prototype, we decided the benefits of such a controller were not necessary. Consequently, we use the more general
McDonald Observatory V2 controller (with minor modifications) for CCD clocking, readout, temperature control,
and shutter triggering. Modifications include a new shutter driver for the Vincent and Associates Uniblitz shutter,
adapting the temperature controller for the 500 Ohm RTD (normally we use 4 X 100 ohm RTDs in series), and
adjusting the preamplifier gain for the ~4.5 µV/e- output of the 3041 device.
We characterized the system in the standard manner for gain, read noise, and CTE with an Fe55 X-ray source. Read
noise with the FI 3041 CCD is higher than advertised, but acceptable at 4.2 electrons at 100 kpxl/s and 3.6 electrons
at 25 kpxl/s. We will specify the VIRUS CCDs to be below 3.5 e- RMS. Parallel CTE is exceptional at more than 1–
1e-6 per pixel. Serial CTE however is a disappointingly low at 1–1.5e-5. We have tried to no avail to tune this out,
and we intend to specify a higher number for the VIRUS devices.
5.2 Vacuum Electronics
Figure 11: On the left is the four layer polyimide flex circuit with coverlay and stiffeners manufactured by Flex
Interconnect Technologies. On the right is the completed VIRUS-P camera, Dewar, and V2 controller. The camera
is shown connected to the turbo pump in the McDonald Observatory CCD Lab.
Lessons learned with the SF1, MM1, and LRS-J instruments on HET quickly pointed us in the direction of a
vacuum compatible stiffened flex circuit for VIRUS-P. Although only a one of prototype, we decided that the cost
and design overhead of a custom flex was mitigated by the significantly reduced assembly time. With VIRUS there
is no question this is the right decision.
Traditional systems based on the V2 controller have separate vacuum connectors for temperature control, clock
driver signals, and bias and video. The V2 penthouse (bias and preamplifier module) plugs directly onto the video
connector without a cable, but the clock driver and video require separate cables. VIRUS-P is no exception, but we
anticipate using a single connector for VIRUS. Inside the vacuum these three connectors are all on the same flex
circuit.
In this circuit the flex contains all circuit layers, but two were removed in the flexible regions of the circuit. The
rigid parts of the board have clearance holes, but no traces, solder pads, or vias. The stiffened flex is significantly
less expensive than a rigid flex, especially in small quantities, but it requires attention to mechanical layout as solder
can only be applied to one side of the board in a stiffened area. We use a mixture of surface mount and through hole
components to mitigate the problem.
The vacuum connectors are Detoronics nickel plated PT style connectors, and there is a Hirose board to board
connector between the flex and header electronics boards. In our design the 2 layer flexible portions are 0.2 mm
thick, the stiffened regions around the vacuum connectors use a 1.3 mm stiffener, the stiffened region along the cold
finger uses two 0.6 mm stiffeners, and in the region of the Hirose connector a single 0.6 mm stiffener. In the flex
regions the ground plane is a grid to limit the thermal conductance of the part. All installed components are passive
with the critical output drain load and reset drain bypass capacitance located within 1 mm of the CCD contact.
No special processing was done to the components, but the finished circuit was ultrasonically cleaned for 5 minutes
each in a mixed bath of Hexanes/IPA followed by a bath of virgin IPA.
6. CONCLUSIONS
The VIRUS-P camera is a robust starting point and test platform for the VIRUS pre production camera currently in
design. Although the entire VIRUS-P camera is in an evacuated cryostat, it is worth reconsidering this design
relative to something similar to the LRS/SF1 camera on the production VIRUS units. If the modestly larger central
obscuration could be tolerated, design, assembly, and maintenance of the camera may be eased at the expense of
detector head complexity. Of course there are other issues to consider as well, and we have an existing, wellunderstood prototype.
Independent of the final instrument concept, a significant improvement in assembly, complexity, and obscuration
could be achieved if a VIRUS specific detector package were engineered. This package could be designed to
eliminate the need for a cold block, complex spider hub, header electronics board and possibly even the thermal
isolator component.
The flexure based spider assembly performs exceedingly well and all measurable image motion is dominated by
movement in the Ultem part. In fact a similar design is recommended for the mirror mount assembly as well. The
required radial constraints could be cut into the mirror mounting ring in such a way that they could also serve as the
spring preload to the axial adjusters which could be directly attached to the Invar spacing rods. Furthermore, if a
repeatable method of focusing the camera while warm could be developed, it would be possible to shim the mirror
into position without any actuators. Such a solution would substantially reduce the cost of the mirror assembly. In
any case, excepting the puck glued to the back of the camera primary, Invar need not be used for this assembly.
Although it has a relatively high CTE, and an inexpensive cryostat housing could easily be fabricated from stainless
steel, the substantially higher thermal conductivity of aluminum affords a more uniform temperature at its surface.
An insulating layer surrounding the housing might also improve temperature uniformity which could help improve
image stability. Regardless, the quantity of black paint used adsorbs a large amount of water; the pumpdown time
could be significantly reduced if it were possible to bake the entire camera to 50 ºC while using the turbo pump.
That said, the major annoyance with the current housing design stems from the fact that with VIRUS-P, the optics
were procured before the housing design had been completed, and the outer diameter of the corrector ended up
between two standard O-ring sizes. This resulted in an unnecessarily large stretch (~4%) in the seal making
assembly risky, difficult, time-consuming, and somewhat problematic. For future designs we recommend standard
O-ring sizes with grooves designed for 1 – 2% stretch.
Regarding the camera's thermal design, it is important to note that any excess heat needing to be removed from the
inner detector assembly ultimately results in a larger required cold link adding to obscuration, diffraction, and
potential cross-talk problems. Similarly, using a simpler cold finger shape, perhaps cylindrical, in the beam will
either add to the previous problem or result in a reduced ability to remove heat. However, the extremely
complicated bend in the cold finger, is completely unnecessary and we would use a straight connection in the future.
Although VIRUS-P was designed to a tight stability requirement, the software reduction pipeline was not yet
available and we have since learned that accurate rejection of false positives requires a much higher degree of
camera stability, presently believed to be about 0.3 µm. As VIRUS-P actually achieves stability at the 0.8 µm level
this is not an impossibility, but a challenge that needs careful consideration in the next stage of VIRUS design. We
believe that a linear trade in thermal load for increased stiffness could be applied, and we feel that kinematic
improvements to the mounting surfaces, continued care in maintaining symmetry, proper temperature control, and
stress relieving of all materials could achieve this specification without active structural control.
Finally, a system with 150 CCDs requires a highly parallelized readout in order to avoid unacceptable overheads.
The VIRUS configuration lends itself to a modularized client/server controller design. However, synchronous or
highly repeatable clocking of all devices simultaneously must be used to minimize readout noise. Giving up some of
the flexibility afforded by the general purpose V2 controller with a purpose built design, using modern components
in the A-D chain, and centralizing control in a hierarchical tree should allow the VIRUS controller to be small,
lightweight, robust and efficient.
ACKNOWLEDGMENTS
We thank the staffs of McDonald Observatory, HET, USM, MPE, and AIP for their help with the construction and
deployment of VIRUS-P. VIRUS-P was funded by a gift from the George and Cynthia Mitchell Foundation. The
HETDEX pilot survey is funded by the Texas Advanced Research Program under grants 003658-0005-2006 and
003658-0295-2007. HETDEX is led by the University of Texas at Austin with participation from the UniversitätsSternwarte of the Ludwig-Maximilians-Universität München, the Max-Planck-Institut für Extraterrestriche-Physik
(MPE), Pennsylvania State University, the Astrophysikalisches Institut Potsdam (AIP), and the HET consortium.
REFERENCES
1. J.A. Booth, M.J. Wolf, J.R. Fowler, M.T. Adams, J.M. Good, P.W. Kelton, E.S. Barker, P. Palunas, F.N. Bash,
L.W. Ramsey, G.J. Hill, P.J. MacQueen, M.E. Cornell, & E.L. Robinson, “The Hobby-Eberly Telescope
Completion Project”, in Large Ground-Based Telescopes, Proc SPIE 4837, 919, 2003.
2. G.J. Hill, K. Gebhardt, E. Komatsu, & P.J. MacQueen, “The Hobby-Eberly Telescope Dark Energy
Experiment”, in The Mitchell Symposium on Observational Cosmology, AIP Conf. Proc., 743, 224, 2004.
3. G.J. Hill, K.Gebhardt, E. Komatsu, N. Drory, P.J. MacQueen, J.J. Adams, G.A. Blanc, R.Koehler, M. Rafal,
M.M. Roth, A.Kelz, C.Gronwall, R.Ciardullo, D.P. Schneider, “The Hobby-Eberly Telescope Dark Energy
Experiment (HETDEX): Description and Early Pilot Survey Results”, in “Panoramic Surveys of the Universe”,
ASP Conf. Series, in press.
4. G.J. Hill, P.J. MacQueen, “VIRUS: an ultracheap 1000-object IFU spectrograph”, Proc. SPIE, 4836, 306, 2002
5. G.J. Hill, P.J. MacQueen, M.P. Smith, J.R. Tufts, M.M. Roth, A. Kelz, J. J. Adams, N. Drory, S.I. Barnes, G.A.
Blanc, J.D. Murphy, K. Gebhardt, W. Altmann, G.L. Wesley, P.R. Segura, J.M. Good, J.A. Booth, S.-M. Bauer,
J.A. Goertz, R.D. Edmonston, C.P. Wilkinson, “Design, construction, and performance of VIRUS-P: the
prototype of a highly replicated integral-field spectrograph for HET”, Proc. SPIE, 7014-257, 2008.
6. J.A. Booth, P.J. MacQueen, J.G. Good, G.L. Wesley, P.R. Segura, P. Palunas, G.J. Hill, R.E. Calder, "The Wide
Field Upgrade for the Hobby-Eberly Telescope", Proc. SPIE, 6267, paper 6267-97, 2006.
7. R.D. Savage, J.A. Booth, K. Gebhardt, J.M. Good, G.J. Hill, P.J. MacQueen, M.D. Rafal, M.P. Smith, B.L.
Vattiat, “Current Status of the Hobby-Eberly Telescope Wide Field Upgrade and VIRUS”, Proc. SPIE, 7012-10,
2008.
8. G.J. Hill, M.J. Wolf, J.R. Tufts, & E.C. Smith, “Volume Phase Holographic (VPH) Grisms for Optical and
Infrared Spectrographs”, in Specialized Optical Developments in Astronomy, Proc. SPIE 4842, 1, 2002.
9. J.J. Adams, G.J. Hill, and P.J. MacQueen, “Volume Phase Holographic Grating Performance on the VIRUS-P
Instrument”, Proc. SPIE, 7014-258, 2008.
10. M. P. Smith, G. J. Hill, Phillip J. MacQueen, W. Altman, John A. Goertz, John M. Good, Pedro R. Segura, and
Gordon L. Wesley, “Mechanical Design of VIRUS-P for the McDonald 2.7m Harlan J Smith Telescope,” Proc.
SPIE, 7014, 2008.
11. M.P. Smith, G.T. Mulholland, J.A. Booth, J.M. Good, G.J. Hill, P.J. MacQueen, M.D. Rafal, R.D. Savage, B.L.
Vattiat, “The cryogenic system for the VIRUS array of spectrographs on the Hobby-Eberly Telescope”, Proc.
SPIE, 7018-117, 2008.