GMOS CCD Upgrade Options S. Kleinman, J. Jensen 26Sep08 Background We are planning to upgrade the scientific capability of GMOS-N by upgrading its roughly 10 year old E2V CCDs to newer CCDs with enhanced red response. The GMOS instruments each feature a mosaic of three ~2k 4k E2V CCDs. The GMOS-N CCDs have better red response compared to those in GMOS-S which were coated for more blue efficiency. The state of the art has moved significantly beyond the current GMOS CCDs and scientific demands for improved red response in our GMOS instruments are quite high. Here, we report on replacement CCD options for GMOS-N that dramatically improve upon the current red response. We are only considering CCDs that are either currently available, or substantially far along in the development process that availability on the timescale of less than a year is not a large risk. Under this constraint, there are now several attractive options. Four CCDs from the Massachusetts Institute of Technology/Lincoln Labs (MIT/LL) are now available to Gemini, originally purchased as part of a consortium led by the University of Hawaii (UH). Improved red-sensitive versions of our current CCDs are now being made by E2V. Both the MIT/LL and the new E2V CCDs would improve upon our current red response and have the additional attraction that mounting them is fairly straightforward. Hamamatsu and the Lawrence Berkeley National Lab (LBNL) both have devices with still better red sensitivities that we can expect within a few months of ordering, although both would require new focal plane hardware to be built for them before they could be mounted within GMOS. E2V is also working on devices with characteristics similar to the LBNL and Hamamatsu CCDs, but available on a later timescale. The San Diego State University (SDSU) Leach Generation 2 controllers we are currently using for GMOS have been performing quite well and our goal is to keep using them, if possible, after the upgrade. All options presented here should allow us to do so. Rick Murowinski (HIA), who assembled the original GMOS focal plane, informs us he can be available for CCD mounting/ alignment support in mid 2009, if needed. We expect he will be a valuable resource to us in this project. Current CCDs Each GMOS currently uses a set of three E2V 42-90 CCDs. The GMOS-N and GMOS-S CCDs are coated differently from each other, with more red performance in the north and more blue performance in the south. Each of the three 16μm thick devices contains 2048 4608 13.5μm pixels which combine to form a 6144 4608 detector plane (Figure 1). Spectroscopic dispersion runs along the short axis of the individual CCDs. At 900nm, there is significant fringing of order 20-30%. Figure 2 shows the quantum efficiencies (QEs) of the current GMOS-N and GMOS-S CCDs, along with the reflectivity of the Gemini multi-layer mirror coating and the GMOS-N optics transmission. One of GMOS's strengths is its nod and shuffle mode. Nod and shuffle is a spectroscopic technique to both improve night-sky subtraction from the data and decrease the slit size used in multi-object masks. It is used about 40% of the time GMOS is in operation. Nod and shuffle involves pointing the telescope on and off source ( nodding ) while simultaneously shuffling the charge so that object and sky are interleaved at a particular detector location. The same optical path and CCD pixels are used to measure the target and background sky spectra quasi-simultaneously. It is a powerful technique and particularly useful in the red where there are many strong night sky emission lines. This operating mode is more sensitive to bad detector cosmetics than normal operations since all charge is being shuffled back and forth over several CCD rows many times before readout. All options presented here are capable of charge shuffling along the long axis of the detector, as needed for nod and shuffle.
Figure 1: GMOS CCDs in imaging (top) and spectroscopic (bottom) modes. Note that in imaging mode, a part of the CCDs on each edge is vignetted.
Figure 2: Current GMOS-N and GMOS-S CCD quantum efficiency curves. The curves labeled Mirror and GMOS-N optics show the reflectivity of the Gemini four-layer silver mirror coating and the transmission of the GMOS-N optics, respectively. MIT/LL The UH-led consortium to purchase MIT/LL CCDs has now packaged and made available to us four devices which offer a significant improvement over our current CCDs (Figure 3). They are 43μm thick, 2048 4096 15μm pixel CCDs and are similar to those currently in use in VLT's FORS2 instrument. The CCDs we have been offered are three Boron Implant (BIV) devices, coated and optimized for good red response, and one blue-optimized Molecular Beam Epitaxy (MBE) CCD. We therefore have the option of replacing the three existing GMOS-N CCDs with either three BIV CCDs or two BIV CCDs and one MBE CCD. The latter option would enhance the blue sensitivity in spectroscopic mode while still improving our overall red performance. In imaging mode, the mixture of different CCD types would mean some extra calibration and a loss of red QE for ~25% of the total imaging field (see Figure 1), but in spectroscopic mode, it would more closely optimize each CCD for the spectral region it covers. This option would give GMOS increased capabilities in the near ultraviolet (UV) and blue wavelength regions without sacrificing any red spectroscopic performance. The cosmetics of the MBE CCD are a bit worse than those of the BIV CCDs, so we may have to spend some time characterizing it in the lab and testing its usefulness with nod and shuffle before making the decision to assemble a mixed MBE/BIV or pure BIV focal plane. As a result of earlier procurement efforts, we already have a test dewar along with mounting plate hardware and cables that will integrate the MIT/LL CCDs with GMOS and its existing controller. The additional work necessary for completion (installation, alignment, and testing) could be provided by the Hertzberg Institute of Astrophysics (HIA) and GL Scientific. SDSU controller software for these CCDs has already been developed at HIA. This option would require very little additional funding since the CCDs and hardware have already been purchased.
Figure 3: Quantum efficiency curves for the MIT/LL MBE and BIV CCDs along with the current GMOS-N and GMOS-S CCDs. The gray box on the left represents the wavelengths below which the combined efficiency of the mirror and GMOS-N optics drops below 20%. Item CCD Cost Availability Read noise Dark current Full Well Capacity CTE Linearity Surface defects Red Fringing Controller upgrade? Resources for upgrade MIT/LL CCD Specifications Value $0 for CCDs; ~$30k for installation Now ~2e- at 100kpix/s <10e-/hr/pixel at 160K ~150,000 e-(initial specification) CTE >0.999997 for 1600 electron signal CTE >0.99999 for 100 electron signal (initial specification) <1% linearity deviation from 100 to 60,000 e-/pixel(initial specification) Grade A devices <3% at 900 nm in 2 nm bandpass filter (initial specification) Not required UH, GL Scientific, Rick Murowinski (HIA)
E2V The current GMOS-N CCDs are E2V 42-90 CCDs. E2V has now developed several new varieties of the 42-90 CCDs with improved red response. The new detectors have identical 2048 4612 13.5μm pixel packaging as our current CCDs. They could be placed directly into the current focal plane using the existing hardware, cables and electronics, but since this would eliminate being able to keep the existing focal plane as a backup, we recommend building new focal plane mounting hardware from existing drawings if we decide on this option. The deep depletion (dd) variety of the 42-90 is a 40μm thick device with improved red response and better fringing characteristics than our standard 16μm thick 42-90s. E2V is also testing thicker, 80μm, bulk versions of the 42-90 which provide even better red response and less fringing than the dd devices. Like the dd CCDs, they would be a direct replacement for our existing standard CCDs, using the same mounting plate, electronics, and controller. Available in 12-18 months are 200μm thick high resistivity (or high-rho ) versions of the CCD, with even better response and even less fringing in the red. These hi-rho devices would also be direct replacements for our current CCDs, package-wise, but would need some controller modifications to work with the higher required bias voltages. Since neither the bulk nor high-rho devices will be available during our target one year timescale, we do not consider them further. E2V offers several possible anti-reflection coatings. The QE curves plotted below all feature E2V's astro-midband coating. This coating gives us the best overall response and is our recommended option. E2V's dd CCD with the astro-broadband coating, however, has a QE curve nearly identical that of the MIT/LL MBE detector, so we could also consider one astrobroadband coated CCD for the blue end of the focal plane in combination with two astro-midband coated devices much like the MIT/LL MBE + 2 BIV option. Figure 4: Quantum efficiency curves for the current GMOS-N CCDs and the three E2V42-90 variants. The dd CCD is the only one available in less than a year.
CCD Cost Availability Read noise Dark current Item Full Well Capacity E2V CCD Specifications ~$300-400K for three CCDs Value <9 mos. dd, ~12 mos. bulk, 12+ mos. high-rho <3e-, typical ~1 e-/pixel/hr ~150,000e- CTE ~99.9995% Linearity Surface defects <~0.2% as currently used Grade 0: <2 column defects, 15 traps, 300 white spots Red Fringing standard dd bulk hi-rho onset: 720 800 900 strength: 1 ½ ¼ 1/12 Controller upgrade? Resources for upgrade Hamamatsu Required only for high-rho option New mounting plate recommended, HIA? Hamamatsu has developed fully-depleted CCDs that have been installed in Subaru's SuprimeCam and are planned for their Hyper SuprimeCam instrument as well. These are 4-side buttable, 200μm thick, 2080 4224 (2048 4096 active) 15μm pixel CCDs with excellent red response and a modest blue response that matches the current GMOS-N E2V CCDs shortward of 650nm (see Figure 5). The Hamamatsu CCDs use higher bias voltages than our current CCDs, requiring some change to our controllers. LBNL makes a bias board for their CCDs that would work with our current SDSU controllers and which Subaru suggests should work with the Hamamatsu CCDs as well. Figure 5: Quantum efficiency curves for the new Hamamatsu CCDs and the current GMOS-N CCDs.
There is also an additional anti-reflection coating option for the Hamamatsu CCDs that improves QE by up to 20% between 380nm to 700nm. At wavelengths below 380nm, the QE drops more sharply than the nominal coating, thereby sacrificing some of the UV shelf seen in Figure 5 (although there is little telescope/instrument throughput in this region). The coating options and their resultant QEs are illustrated in Figure 6, taken from a Hamamatsu presentation. Figure 6: Different coating options for the Hamamatsu CCD. Process A is also plotted in Figure 5. Hamamatsu CCD Specifications Item Value CCD Cost ~$300K for three CCDs Availability Two months after receipt of order Read noise ~4 e- rms Dark current 5 e-/pixel/hr max Full Well Capacity >90,000, >~150,000 typical CTE >0.999995 Linearity <~0.5% up to half full well Surface defects no current information Red Fringing Should be excellent Controller upgrade? Voltage/bias board required Resources for upgrade Mounting plane hardware and elec. required.
LBNL The Lawrence Berkeley National Lab is developing new high-resistivity CCDs with exceptional red response for the Supernova Acceleration Probe (SNAP) and the Dark Energy Survey (DES) projects. These CCDs offer the best red response of the options under consideration, but are still somewhat unproven in that they are not currently in use by any astronomical instrument. They are 3-side buttable, 2048 4096 15μm pixel, 250μm thick devices. There is a run of devices being made now that may yield three detectors which could potentially be available to us by the end of October, 2008. LBNL has no plans to do any further 2k 4k device runs, so if there are no science-grade devices available in October, there may never be. As with the Hamamatsu and E2V high-rho CCDs, the LBNL CCDs would require a new bias voltage board to work with our current controllers. LBNL already has just such a board, proven to work with our current generation of SDSU controllers. LBNL is also developing a single focal-plane mounted Application-Specific Integrated Controller (ASIC) for use with their CCDS. This controller represents a future upgrade option, but we would also investigate its applicability now were we to choose the LBNL CCDs. Figure 7: Quantum efficiency curves for the new LBNL and Hamamatsu CCDs along with the current GMOS-N CCDs. The dashed Hamamatsu line represents the Process B coating option. Note that the LBNL devices have little sensitivity below 360nm.
CCD Cost Availability Item LBNL CCD Specifications Read noise ~4 e- Dark current Value ~$270K for three CCDs End of October or not at all <~7 e-/pixel/hr Full Well Capacity >~250,000 CTE >0.999995 Nonlinearity <~1% Surface defects Red Fringing Controller upgrade? Resources for upgrade no current information None blueward of 900nm; ~5% at 1050nm with 10nm wide filter. LBNL board for SDSU, LBNL ASIC? Need new mounting/focal plane Summary There are two classes of GMOS-N upgrade options presented here: CCDs that require manufacturing a new mounting plate with associated hardware and electronics, and those that don't. In terms of cost, the MIT/LL options are the least expensive, involving only marginal costs to finish the CCD integration into GMOS. Their immediate availability is also an attractive plus. The E2V, Hamamatsu, and LBNL CCDs all cost roughly the same, but the E2V option would save some money over LBNL and Hamamatsu in that a new focal mounting system would not need to be designed, just manufactured. The MIT/LL and E2V (see Figure 8) options may be easiest to install, but they also provide the least performance improvement. Due to their identical packaging as our current CCDs, the E2V 42-90 dd detectors represent our lowest risk, but also our lowest performance enhancement option. The MIT/LL CCDs are attractive due to their low additional cost, already-manufactured mounting plate, and enhanced performance compared to both our current CCDs and the E2V 42-90 dd detectors. The E2V CCDs are known to have extremely accurate packaging, making it easier to ensure that all three CCDs are co-planar in their mounts. We do not know the packaging accuracy of the MIT/LL CCDs, but expect planarization to require significant effort by either GL Scientific or Rick Murowinski. CCDs already paid for Pros: MIT/LL Mounting plate hardware and electronics in hand Better red response MBE + 2 BIV option for enhanced blue spectroscopy Available immediately MIT/LL vs. E2V dd Pros: E2V Good support available from E2V Can install directly into current focal plane or reproduce current hardware/electronics Better blue and mid-range response astro-midband + 2 astro-broadband option for enhanced blue spectroscopy Lower dark current
Figure 8: A comparison of the two potentially quickest upgrade options. The E2V dd CCDs can be ordered now and a mounting plate could be reproduced relatively quickly from existing drawings. The MIT/LL CCDs and corresponding mounting plate hardware and electronics are available now. The Hamamatsu and LBNL choices offer the most improvement in red QE of all the options (Figures 7 and 9), but both require new focal plane mounting hardware and electronics. They also require controller modifications which the MIT/LL and E2V options do not. We expect, however, that the mounting plate and controller work can be accomplished within our one year time frame. Although the LBNL CCDs have better red QE, with the exception of full well capacity, there are only minor differences in the other operating specifications of these and the Hamamatsu CCDs. Available now Pros (cons): Hamamatsu In use now at Subaru. Subaru becomes potential local technical resource (Controller upgrade needed) Best UV/blue response Process B coating offers enhanced 350 700nm response Hamamatsu vs. LBNL Pros and Cons Pros (cons): LBNL Available now (or never) (unproven) Proven controller upgrade available Best red response Highest full well capacity
Figure 9: Quantum efficiency curves for the primary CCD options discussed here. Recommendations: 1) Wait until the end of November, 2008 for LBNL to have characterized three scientific devices that could be available for purchase. If they live up to their current specifications, order them. If not, place an order with Hamamatsu (with the Process B coating) and see if the LBNL SDSU controller bias board will work with them (or otherwise upgrade the controller). 2) Proceed immediately and assemble a MIT/LL focal plane (three BIV, or two BIV/one MBE CCD) as a backup to the Hamamatsu or LBNL procurement. The MIT/LL focal plane could be installed in GMOS- S if not needed in GMOS-N. If configured with one MBE and two BIV CCDs, could provide even better spectroscopic UV/blue capabilities than is currently available with GMOS-S while simultaneously performing much better in the red than the current CCDs (Figure 3). 3) Depending on scientific interest and the final quality of the MBE CCD we receive, decide on whether we ultimately populate the MIT/LL focal plane with three BIV CCDs or one MBE and two BIV detectors.