Realization and application of a 111 million pixel backside-illuminated detector and camera
Norbert Zacharias, Bryan Dorland, Richard Bredthauer, Kasey Boggs, Greg Bredthauer, Mike Lesser
aa r X i v : . [ a s t r o - ph ] S e p Realization and application of a 111 million pixelbackside-illuminated detector and camera
Norbert Zacharias a , Bryan Dorland a , Richard Bredthauer b , Kasey Boggs b , Greg Bredthauer b ,and Mike Lesser ca United States Naval Observatory, 3450 Massachusetts Avenue NW, Washington DC, 20392; b Semiconductor Technology Associates, San Juan Capistrano, CA 92675; c Steward Observatory, University of Arizona, Tucson, AZ 85721
ABSTRACT
A full-wafer, 10,580 × ×
95 mm) CCD was designed and tested at Semiconductor TechnologyAssociates (STA) with 9 µ m square pixels and 16 outputs. The chip was successfully fabricated in 2006 at DALSAand some performance results are presented here. This program was funded by the Office of Naval Researchthrough a Small Business Innovation in Research (SBIR) program requested by the U.S. Naval Observatory forits next generation astrometric sky survey programs. Using Leach electronics, low read-noise output of the 111million pixels requires 16 seconds at 0.9 MHz. Alternative electronics developed at STA allow readout at 20MHz. Some modifications of the design to include anti-blooming features, a larger number of outputs, and useof p-channel material for space applications are discussed. Keywords:
Astrometry, large-format CCD, all-sky-survey, star tracker, Space Situational Awareness
1. INTRODUCTION
This paper descibes the motivations and requirements which led to the development of the world’s largest-formatCCD detector. The history and realization of the entire camera around this device for the U.S. Naval Observatory(USNO) is presented. Some performance results obtained from a thinned, backside-illuminated detector of thiskind are given. The paper concludes with plans for the future and explains applications in ground-based andspace-based programs.
2. REQUIREMENTS
In recent years the requirements for ever larger focal plane assemblies at astronomical telescopes have mostly beensatisfied by assembling numerous smaller devices into large focal plane mosaics. The advantage of this mosaicapproach is a larger yield in producing high quality detectors, which results in cost savings for the instrumentdevelopment. Although larger, monolithic detectors are desirable, there has not been a real driver requiring thebold step to go beyond the current typical 2k ×
4k scale devices.
In planning projects beyond the successful USNO CCD Astrograph Catalog (UCAC) program it was realizedthat a large-format detector is needed. Using the technique of photographic astrometry the positions of starsare determined with respect to several reference stars with known positions by direct imaging of the sky usingdedicated telescopes. The large-format photographic plates (up to about 17 inches on a side) traditionallyused were later replaced by CCD detectors, providing higher quantum efficiency and more accurate centroidingresults as compared to the photographic process. Unfortunately, CCD detectors are very small compared tophotographic plates. For high accuracy astrometric measurements many reference stars need to be on the samedetector or at least many well exposed anonymous stars are required to tile together overlapping fields with asfew mapping parameters as possible. We needed to advance beyond the existing CCD formats in order to make Further author information: (Send correspondence to N.Z.)E-mail: [email protected], Telephone: 1 202 762 1423 igure 1. The U.S. Naval Observatory Twin Astrograph, currently at the Flagstaff Station, with 10k camera dewarattached to the “red lens.” This setup will be used for astrometric test observations on the sky in preparation for theURAT program. significant progress in this area and to fully utilize the existing large focal plane of our astrograph as well asthose of future dedicated astrometric telescopes.At that same meeting where the need for larger-format CCD detectors for astrometric mapping was pre-sented, we learned that designers of a Large Binocular Telescope spectroscopic instrument were also looking atsimilar types of CCD detectors to improve calibrations and lower systematic errors. Following discussions, bothgroups agreed to share risks in development of a large-format detector which could be used for both projects. Apixel size of 9 µm was agreed upon, but independent funding avenues had to be pursued.Of particular importance for astrometric applications are a high charge transfer efficiency and relatively fastreadout, with a goal of about 10 sec for the full frame. This required the use of a large number of paralleloutputs. Standard high-quality materials and designs used for science grade CCD detectors were sufficient tosatisfy all other requirements, providing the yield issue could be addressed successfully. Figure 1 shows the USNO Twin Astrograph, which was used for the UCAC program (1997 to 2004) and beforethat for astrometry using photographic plates (24 cm square and 8 ×
10 inch). The original “blue” lens wasreplaced by a 5-lens “red lens” objective of extremely high astrometric performance, which has been in operationsince 1990. The new 10k camera dewar is attached to the red lens now, while the second telescope tube featuresa visual bandpass corrected lens which is used for guiding.2he following table presents the main characteristics of the existing USNO astrograph and the planned USNORobotic Astrometric Telescope (URAT).
4, 5
Both feature an available focal plane area of about 30 cm in diameter.Design work on the URAT began in 2000 and concluded in 2005. A contract was signed with EOST to producethe primary mirror which will be delivered by the end of 2007. Funding of the URAT telescope is uncertainbeyond that time. However, the focal plane development is progressing well with major purchases anticipated in2008.
Table 1. Comparison of the existing USNO Twin astrograph (red lens) telescope with the planned USNO Robotic Astro-metric Telescope (URAT). property astrograph URATaperture [meter] 0.20 0.85focal length [meter] 2.00 3.60scale [arcsec/pixel] 0.90 0.50diameter field of view [degree] 9.00 4.50diameter focal plane [mm] 320 283bandpass [nm] 550 −
750 600 −
3. REALIZATION3.1 Research Program
A sponsor for the development of a general, large-format, monolithic detector was found at the Office of NavalResearch. A Small Business Innovation in Research (SBIR) topic proposed by the USNO Astometry Departmentwas accepted for a phase I study in 2004. Originally an 8-inch full-wafer CMOS or CMOS/hybrid device wasconsidered, but this was quickly rejected as unrealistic at the time. Instead, 2 companies were funded in phase Ito develop a 6-inch wafer full-frame CCD detector. In 2005 the main research phase II funding was awarded toone of the phase I participants, Semiconductor Technology Associates (STA), of San Juan Capistrano, CA.
This 10,580 × igure 2. The STA1600 chip after packaging. The top and bottom areas connect to 8 outputs each. The photo-sensitivearea is 95 mm ×
95 mm.
The STA1600 full-frame detector was successfully manufactured by DALSA in June 2006 (see press release).The yield was sufficient to produce several engineering and science grade chips. Initial characterization by STAof the full-wafer device confirms acceptable parameters. Figure 2 shows the packaged device with 4 connectorsfor 4 outputs each.
The backside processing of the 10k CCD is similar to processing smaller devices. The wafers are first mechani-cally lapped to 250 µm . Gold stub bumps are applied to each bond pad and then the wafer is diced. The die ishybridized to a 1.4 mm thick silicon substrate with indium bumps matching the CCD bond pads. Epoxy is usedas an underfill material. Backside thinning is accomplished in a selective etch which stops at the epitaxial layer.A final etch polishes the surface. Backside coatings are applied using the University of Arizona ChemisorptionCharging process. A custom invar package and circuit board set has been designed and fabricated for thebackside parts. After packaging and wire bonding the device is ready for testing. A mechanical flattness isachieved to support an f/4.5 beam of the URAT instrument.
4. CAMERA AND PERFORMANCE
In addition to the thinned, backside-illuminated, science-grade STA1600 CCD detector, the 10k camera consistof a custom dewar, filter, shutter, electronics and required interfaces. The filter, made by Andover Corporation,is 12 mm thick, ultra-flat, and has a diameter of 160 mm for a 683 to 747 nm bandpass. The bandpass has beenchosen to be as red as the astrograph lens supports but to exclude the H α region of the spectrum, in order toavoid photons from emission nebulae on exposures taken for high accuracy centroiding of stellar images. Thefilter is fixed mounted as dewar window and the separation of the backside of the filter to the focal plane is4 igure 3. The STA1600 chip inside the 10k camera dewar. Here an operational front-side chip is used. only 5 mm. In addition to the interference layers and coatings of that filter a small (1.2 mm diameter) neutraldensity spot with a factor of about 200 attenuation has been added near the center of the filter. This will allowastrometric observations of bright stars in reference to much fainter stars in the same field of view.Figure 3 shows the dewar with an engineering grade STA1600 chip and a clear glass window for testing.This 10k camera is currently limited by the Leach electronics. A complete readout of the full 111 megapixelimage requires 16 seconds. We are using an Astronomical Research GenIII camera with 2 ARC48 8-channel A-Dboards. The camera is running the CCD with a 912 kHz serial clock and a 30 kHz parallel clock. An ARC42fiber optic timing board relays output data from the ARC controller to the PC. We are digitizing 16 bits foreach of the 16 outputs. This data rate of 912 kHz is limited by the capacity of the current fiber optic card. At912 kHz we achieve a read noise of 6 electrons RMS on a thinned STA1600 CCD. Alternative electronics hasbeen developed at STA and a readout at 20 MHz has been demonstrated on a frontside device.A 150 mm aperture shutter was custom built for the 10k camera by the Bonn instrumentation group (Fig. 4). The camera is run by a Linux PC. Operation of the shutter has been integrated and a completelynew interface to an upgraded astrograph is in preparation, all controlled by the same PC from a command-lineinterface suitable for robotic operation. Image data files will be stored in a compressed FITS format, about 120MB per full-frame.
5. APPLICATIONS5.1 Ground-based Star Catalogs
The main application for this detector and camera is to support DoD needs and requirements for star positions.This will also serve the general astronomical community by providing highly accurate positions and propermotions of millions of stars.For star tracker applications (bright stars) the goal is to improve upon the Hipparcos Catalog positions,which have steadily degraded due to accumulation of proper motion errors since their mean observing epoch in1991. This improvement can be accomplished by observing bright targets with the USNO astrograph and the5 igure 4. The “Bonn” shutter system for the 10k camera. This custom made shutter (black device) comes with controlelectronics to fit into a standard rack. new 10k camera through the neutral density spot on its filter. Tycho-2 stars in the same 2.5 by 2.5 degree fieldof view will serve as reference frame.For Space Situational Awareness research the 10k camera can be used at either the astrograph or the URAT todetermine accurate positions of faint stars (down to R magnitude 18 and 21, respectively). For this applicationmaximal sky coverage per exposure is needed. Figure 5 shows a focal plane layout with 4 of the 10k CCDdetectors in their current packaging. The circle is 333 mm in diameter, close to the limit of the astrograph focalplane area. Attached to the astrograph this would provide 27 square degrees sky coverage in a single exposure.This layout would need to be modified to be able to mount the 10k chips closer together for the slightly smallerURAT focal plane.We plan to purchase such a “4-shooter” camera in 2008. This would allow us to construct a URAT focalplane and use it at the astrograph. After only 2 years of observing time from the Cerro Tololo Inter-AmericanObservatory (CTIO), positions and parallaxes of stars in the 11 to 16 mag range on the 5 to 10 mas level couldbe produced, significantly improving current star catalog data.If the USNO-lead Milli-Arcsecond Pathfinder Survey (MAPS) mission
12, 13 is approved, the tie of the resultingnew celestial reference frame to fixed, extragalactic sources would be performed by URAT, and funding is expectedfor the new, dedicated, ground-based telescope utilizing the 4-shooter camera based on an anti-blooming modifiedversion of the STA1600 chip.
The initial design for the STA1600 called for 16 2-stage outputs that could run at up to 15 MHz, resulting in amaximum frame rate of approximately 2 frames per second (fps) with a resultant post-CDS read noise of 40 to50 e − RMS. Because it was designed to be operated at ground-based observatories, the current design has nobuilt-in radiation mitigation capabilities. 6 igure 5. Layout of a 4-shooter focal plane based on the existing STA1600 chip design. This design would barely fit theastrograph field of view and is slightly too large for the URAT focal plane. A modification of the packaging is planned.
A space-based implementation of this CCD could include an increase in the frame rate and improved radiationhardening. In order to increase frame rate, the number of readout amplifiers would be increased to 32 and 3-stageamps would be used rather than 2, increasing the speed to 40 MHz per channel. This approach would allowthe frame rate to be increased from 2 to 10 fps, with an increase in read noise to around 60 e − RMS. In orderto improve radiation hardness, perhaps the most straightforward approach would be to use p-channel ratherthan n-channel material. Numerous results have shown an increase of approximately an order of magnitudein hardness vs. displacement damage is achieved when using p-channel material. Other solutions, such as activecircuitry, could also be considered, although these methods are less attractive due to their added complexity andpotential negative impact on yield.
6. DISCLAIMER
Although some manufacturers are identified for the purpose of scientific clarity, the USNO does not endorseany commercial product nor does the USNO permit any use of this document for marketing or advertising.We further caution the reader that the equipment quality described here may not be characteristic of similarequipment maintained at other laboratories, nor of equipment currently marketed by any commercial vendor.7
CKNOWLEDGMENTS
We wish to thank the Office of Naval Research (ONR) for funding this large, monolithic detector research programthrough a Navy SBIR program and Sean Urban, Head of the Nautical Almanac Office, for finding that sponsor.
REFERENCES
1. N. Zacharias, S. E. Urban, M. I. Zacharias, G. L. Wycoff, D. M. Hall, D. G. Monet, and T. J. Rafferty, “TheSecond US Naval Observatory CCD Astrograph Catalog (UCAC2),” AJ , , p.3043, 20042. D. Vukobratovich, T. M. Valente, R. R. Shannon, R. A. Hooker, and R. E. Sumner, SPIE , , p.245, 19913. N. Zacharias, “Astrometric reference stars: from UCAC to URAT,” in , AN , , p.631, 20044. U. Laux, and N. Zacharias, “URAT optical design options and astrometric performance,” in Astrometry inthe Age of the Next Generation of Large Telescopes , K. P. Seidelman and A. K. B. Monet eds.,
APS Conf.Ser., p.106, 20055. N. Zacharias, “The URAT Project,” in
Astrometry in the Age of the Next Generation of Large Telescopes ,K. P. Seidelman and A. K. B. Monet eds.,
APS Conf.Ser., p.98, 20056. R. Bredthauer, K. Boggs, G. Bredthauer, “A monolithic 111-M pixel high speed, high resolution CCD,” in , p.170, 20077. M. P. Lesser, “Very Large Format Back Illuminated CCDs,” in
Scientific Detectors for Astronomy , P. Amico,J. W. Beletic, and J. Beletic eds, p.137, Kluwer Academic Publishers, 20048. M. P. Lesser, V. Iyer, “Enhancing Back Illuminated Performance of Astronomical CCDs,”
SPIE , , p.23,19989. M. P. Lesser, D. Ouellette, “Development of hybridized focal plane technologies,” SPIE , p.03, 200610. K. Reif, “USNO150 a Bonn Shutter for the USNO Robotic Astrometric Telescope,” user manual , BonnInstrumentation Group, 200611. ESA, “The Hipparcos Catalgoue,”
European Space Agency
SP-1200 , 199712. B. Dorland, N. Zacharias, R. Gaume, K. J. Johnston, G. Hennessy, V. Makarov, and C. Rollins, “TheMilli-Arcsecond Pathfinder Survey (MAPS) mission,”
BAAS , , abstract 117.04, 200513. N. Zacharias and B. Dorland, “Concept of a stare-mode astrometric mission,” PASP , , p.141914. C. J. Marshall, et al., SPIE ,5499