Integrated system tests of the LSST raft tower modules
    P. O'Connor
    , P. Antilogus
    , P. Doherty
    , J. Haupt
    , S. Herrmann
    , M. Hu?er
    , C.
    , J. Kuczewski
    , S. Russo
    , C. Stubbs
    , and R. Van Berg
    Brookhaven National Laboratory, Upton, NY USA
    LPNHE, Paris FR
    Harvard University, Dept. of Physics, Cambridge, MA USA
    SLAC National Laboratory, Menlo Park, CA USA
    Univerisity of Pennsylvania, Philadelphia, PA USA
    The science focal plane of the LSST camera is made up of 21 fully autonomous 144 Mpixel imager units designated
    raft tower modules (RTM). These imagers incorporate nine 4K x 4K fully-depleted CCDs and 144 channels of
    readout electronics, including a dedicated CMOS video processing ASIC and components that provide CCD
    biasing and clocking, video digitization, thermal stabilization, and a high degree of monitoring and telemetry.
    The RTM achieves its performance goals for readout speed, read noise, linearity, and crosstalk with a power
    budget of less than 400mW/channel. Series production is underway on the ?rst units and the production will
    run until 2018. We will present the RTM ?nal design, tests of the integrated signal chain, and performance
    results for the fully-integrated module with pre-production CCDs.
    Keywords: LSST, CCD, Readout Electronics
    The Large Synoptic Survey Telescope (LSST) is a large aperture, wide ?eld survey telescope with a 3 Gpixel
    camera which will perform a 10-year all-sky survey in the optical and near IR
    To maximize survey e?ciency,
    the readout time for the focal plane is set at 2 seconds. The fast framing, together with the constrained location
    of the camera at the 3-mirror focus of LSST, lead to challenging requirements for integration density and power
    dissipation. Electronics to control and process an aggregate of 1.5Gpixels/s must ?t into a volume of 1/6m
    behind the focal plane, inclusive of thermal and mechanical support for the sensor array.
    To achieve these goals, the large science focal plane is divided into 21 modules, denoted rafts (Figure
    Each raft consists of a mosaic assembly of nine 4K x 4K CCDs together with an integrated package of elec-
    tronic, mechanical, and thermal support components, forming an autonomous and fully testable 12K x 12Kpix
    imager (Raft Tower Module, RTM).
    Compactness ( < .008m
    /RTM) and low power ( < 60W/RTM) are realized
    by minimizing CCD-electronics interconnect length and using analog CDS video processing implemented in a
    multichannel CMOS ASIC.
    The remaining sections will discuss the design of the key components, test results on the fully-integrated
    readout chain, and plans for the remaining development.
    2. SENSORS
    LSST will use thick, fully-depleted CCD (FDCCD) sensors to cover the spectral range from 320 to 1050nm. To
    meet the 2 second readout time while simultaneously achieving low read noise, a conservative pixel rate was
    chosen. As a consequence, the array needs to be highly parallelized with each 4K x 4K CCD having sixteen
    output ports. Additional stringent constraints stemming from image quality are also important; the sensor
    design, optimization, and prototype characterization is described in Refs.
    , and
    . Key sensor requirements
    summarized in Table 1.
    Further author information: (Send correspondence to P.O'C.)
    P.O'C.: E-mail:

    Figure 1. LSST science focal plane, readout electronics envelope, and rendering with RTMs.
    Table 1. Key sensor requirements.
    CCD type
    n-channel, full-frame
    High-resistivity, fully-depleted silicon, 100 ? m thickness
    Pixel size
    10 ? m
    Number of outputs 16
    Frame read time
    2 seconds
    Implied pixel rate
    550 kpix/s
    Read noise
    < 8 electrons
    < 2%, 1000 - 90000 e-
    0.999995 (serial), 0.999997 (parallel)
    QE (band)
    41% (u), 78% (g), 83% (r), 82% (i), 75% (z), 21% (y)
    < 5% rms
    PSF (di?usion)
    < 5 ? m rms (0.235 arcsec FWHM)
    Image height
    95% of image surface within ? 9 ? m of target
    To minimize noise, obscuration, and vacuum penetrations the electronics is housed in a compact enclosure
    occupying a volume of 8000 cm
    , located directly behind the sensor array within the camera cryostat. Three
    Readout Electronics Boards (REBs), each serving three CCDs, contain all circuitry needed to control and read
    out the CCD array. Table
    lists the key requirements for the raft electronics. In LSST it will be possible
    to send independent clock sequences to any individual CCD, providing versatility e.g. to perform high speed
    region-of-interest acquisition during science exposures. The read noise requirement is set such that 95% of survey
    exposures will be sky noise limited. The crosstalk performance is based on planned crosstalk correction in the
    image processing pipeline; if uncorrected, this level of crosstalk would result in 100 detectable crosstalk ghosts
    per CCD in every exposure. The allowed power dissipation is lower, by a large factor, than other contemporary
    mosaic cameras (PanSTARRS, DECam, HyperSuprimeCAM).
    The raft electronics includes provisions for handling sensors from either of two manufacturers. The 2 sensor
    types are mechanically and electronically plug-compatible, but di?er slightly in the readout section. One model

    Table 2. Readout electronics requirements.
    CCD Bias
    Provide programmable CCD biases (except back depletion voltage).
    Range -10 to +40V, 1 to 3.5mA, 12-bit.
    CCD Clocking
    Provide switches, drivers, and programmable rail voltages for all CCD clocks
    Timing generator for CCD clocks, CDS switches, ADCs.
    Independent state machines for each CCD.
    Video processing
    CDS by dual slope integration (ASPIC chip
    Read noise (REB contribution) < 3e-
    Crosstalk < 0.2%
    Max. signal > 180000 e
    Power dissipation
    < 55W per RTM (380mW per channel).
    Data acquisition
    18b sampling, 48 channels @550kpix/s
    Serialize with ECC. No frame bu?er on the REB
    3.125Gbps Cu link (per REB) to optical TX/RX
    Slow controls and monitoring Con?gure ASICs and DACs for bias and clock rails
    Monitor board temperatures, supply currents and voltages, CCD biases
    Measure CCD temperature to 0.5C accuracy, 24-bit precision
    Provide up to 4W power to raft makeup heaters.
    Read serial ID chip
    Hardware protection for CCDs
    Vacuum operation; -40C cold sink; low outgassing
    has a higher-impedance output ampli?er, and requires active bu?ers installed on the ?ex cable that connects the
    CCD to the REB.
    shows the physical arrangement of the sensors and electronics within the RTM. A functional block
    diagram of the RTM is shown in Figure
    . The clock/bias, video processing, and video digitization functions
    are replicated three times on each REB and each stripe serves one CCD, while common functions are located
    towards the output end of the board. The REB is a double-sided, 16-layer board with multiple heavy ground
    planes for heat conduction (Figure
    Fully-automated test stands have been constructed and provide acceptance testing for the CCDs and REBs; they
    are described in References
    , respectively, in these Proceedings. For integrated tests of the full signal
    chain, two development test benches were fabricated. The ?rst includes a small cryostat holding an individual
    CCD cooled to -100C with a potted ?ex cable and interface board connecting it to the REB under test. This
    arrangement permits us to operate the signal chain while having access to all test points on the REB for probing.
    A compact optical train allows multiwavelength ?at?elds and image projection. The interconnect between the
    room-temperature REB and the CCD requires a series of boards and cables totaling about 10" in length; the
    resulting capacitance degrades system performance somewhat. A prototype of the LSST data acquisition system
    communicates with the REB via optical ?ber. FPGA ?rmware con?gures the board components, de?nes the
    timing states and sequences, serializes and transmits the 18-bit data, and communicates with the host via a
    custom protocol having virtual channel capability. Host software provides a GUI-based sequence editor, real-
    time image display and diagnostics, FITS ?le image formatting, and handles safe power-up of the board and

    Figure 2. Raft tower module. Left: The raft contains three readout electronics boards (REBs), each connected to three
    CCDs by a set of ?ex cables. The CCD mosaic on its silicon carbide baseplate is cooled to -100C, and the heat dissipated
    in the REBs is sunk to a cold plate at -40C. Center: RTM showing mechanical and thermal hardware. Right: mechanical
    The second test bench uses a large cryostat which holds a single RTM and provides cooling to both the CCD
    mosaic at -100C and the REBs at -40C. Both test benches include deployable, in-vacuum xray sources in addition
    to optical projection systems to illuminate the CCDs. Cooling is provided by closed-cycle Joule-Thompson
    cryocoolers allowing uninterrupted operation. A Python-based software library allows scripted acquisition of
    images needed for gain, noise, linearity, and dynamic range measurements. Monitoring and telemetry, including
    CCD temperature, is logged at 10Hz. Image display with statistics is provided in real time. Thanks to the fast
    2s readout, we get rapid feedback during setup and performance tuning, and can e?ciently acquire large image
    stacks for o?ine processing. An upgraded version of the RTM test stand is being constructed and will be used
    in the production cleanroom for RTM acceptance testing. The two system test benches are shown in Figure
    The unique architecture of LSST led us to develop several novel test methods. First, the readout speed,
    power budget, and compactness of the focal plane lead to multiple electronic crosstalk paths with signi?cant
    coupling. To evaluate this, we use a multi-spot mask to project sixteen "aggressor" spots on the CCD, one in each
    segment, to simultaneously acquire all 256 elements of the crosstalk matrix of a CCD .
    Second, since the entire
    electronic chain is in an inaccessible location in the Camera cryostat, access to analog signals will be impossible.
    However, by operating the video digitizers in "sampling scope" mode we can indirectly acquire waveforms with
    ?ne amplitude and time granularity. For dark or ?at?eld images with repetitive pixel waveforms, we time the
    ADC convert signal to acquire one point per pixel period, then increment the sampling time on successive pixels
    to build up a representation of the waveform at the output of the dual-slope integrator. Moreover, the video
    ASIC can be set to "transparent mode", whereby the DSI processing is bypassed and the chip acts as a unity-
    gain preampli?er; in this mode we can visualize the CCD output source waveform itself, a valuable diagnostic.
    In e?ect, the normal data path can be reprogrammed to provide the equivalent of a 100 Msample/sec, 18-bit
    oscilloscope. Examples of the sampled waveforms are shown in Figure
    5. RESULTS
    5.1 Single-CCD
    5.1.1 Basic functionality
    On the smaller test bench, we veri?ed basic performance of the signal chain. For these tests, we operated at
    about 90% of the ?nal LSST pixel rate, to allow for CCD transients to settle while driving the high interconnect
    capacitance. Waveforms from the CCD, ASPIC, and clock drivers are shown in Figure

    Figure 3. Functional block diagram of the RTM. The clock/bias, video processing, and video digitization functions are
    replicated in three identical stripes on each REB, see Figure
    Figure 4. Layout of the REB (105 x 420mm).
    5.1.2 Multi-vendor compatibility
    shows USAF targets imaged onto sensors from each of LSST's candidate suppliers. Signal level is 8ke
    5.1.3 Noise, gain, and crosstalk
    (a) shows photon transfer curves (variance vs. mean) for all 16 ampli?ers. Figure
    (b) shows the
    Fe spectra for the 16 ampli?ers. The K ? and K ? peaks are resolved and the gain dispersion is about 1.5%.
    Also shown in the ?gure are the read noise pe rchannel and a target image at 12e- signal level.

    Figure 5. Left: single-CCD test bench with REB at room temperature. Right: Large cryostat for full-RTM testing.
    Optical projection cameras can be seen in each photo.
    Figure 6. Examples of sampling-scope mode waveform acquisition. Left:CCD output source waveform (inverted) after
    ASPIC with DSI disabled. Right: with DSI enabled.
    In the single-CCD con?guration, intra-CCD crosstalk was found to reach a level of about 0.4% for nearest-
    neighbor channels. The crosstalk pattern and magnitude wasI consistent with a SPICE simulation model that
    included parasitic capacitances present in the test stand.
    5.2 RTM
    5.2.1 Basic imaging performance
    We tested a partial RTM populated with an array of three CCDs and read out by one REB (Figure
    ). This
    con?guration represents one of the 63 basic electro-optic cells of the LSST focal plane. 4K x 12K ?at?eld and
    target images are shown in Figure
    The 48-channel photon transfer curves are shown in Figure
    . The REB used in this test had each half of
    the board populated with a di?erent value gain-setting resistor, resulting in 2 gain families for each CCD. Note
    that the PTC shows quadratic, rather than linear behavior due to the well-known correlations between pixels
    found in thick CCDs.
    5.2.2 Noise and crosstalk vs. pixel rate
    With the reduced interconnect parasitics in the RTM con?guration, noise and crosstalk were improved and
    met speci?cations when running at the full 550kpix/s rate. Figures
    illustrate the RTM intra-
    and inter-CCD crosstalk. The images were obtained by illuminating the center CCD with the multi-aggressor
    crosstalk mask. In Figure
    , the upper images show the aggressor and victim patterns. The lower graphs

    Figure 7. Waveforms from the single-CCD integrated readout test.
    Figure 8. Images from each of LSST's candidate sensor suppliers. (Missing segments on right image are due to an
    intermittent interconnect line/).
    show the crosstalk matrix (rows and columns correspond to aggressor and victim channels respectively) and the
    distribution of positive and negative elements. Figure
    shows that there is no measurable crosstalk between
    CCDs at the level of 10
    To understand the e?ect of pixel rate on performance, we made a series of runs with pixel rate varying from
    100 to 600kpix/s. The results are shown in Figure
    . Noise decreases linearly with decreasing pixel rate until
    reaching a ?oor at around 3e-. Crosstalk is about 5 times smaller in the RTM test bench, where the interconnect
    capacitance is signi?cantly less than in the single-CCD con?guration; there, it was found to drop exponentially
    with decreasing pixel rate, again consistent with SPICE modeling. These results may be used to inform future
    system-level optimization.
    5.2.3 Temperature dependence
    The electronic gain of the RTM video channel is required to be stable to 1% over a 12-hour period and 0.1% over
    a 1-hour period; temperature variation of the REB analog components is expected to be the dominant source of
    gain ?uctuation (CCD temperature is stabilized to 0.1 degrees C while REB temperature will roughly follow the
    variation in cold plate temperature). To measure the susceptibility to REB temperature variation, we measured
    the gain and noise of the RTM while varying the cold plate temperature from -60 to +10 degrees C. The results
    are shown in Figure
    . Noise stays within speci?cation over this range while the gain temperature coe?cient
    sets a limit on the allowable REB temperature ?uctuation of 1.6C over 1 hour and 16C over 12 hours.
    A 160m
    ISO 7 cleanroom has been commissioned at Brookhaven National Laboratory to fabricate and test 22
    production RTMs. Production equipment includes two CCD electro-optic test stands, an RTM electro-optic test
    stand similar to the development station described in Section
    , non-contact metrology stations for measuring
    sensors, baseplates, and assembled rafts at room and cryogenic temperature, a REB test station, and handling

    Figure 9. Performance with single-CCD integrated readout con?guration. (a) photon transfer curves; (b)
    (c) read noise; (d) target image at 12 e- signal level.
    and assembly tooling. All test stands have fully automated acquisition, analysis, and report generation with
    results entered into an electronic traveler database. Production RTM construction will begin in 4Q2016 and is
    planned to be completed by 1Q2019.

    Figure 10. Partially-populated RTM showing 3 CCDs connected to one readout electronics board. Gold heater resistor
    mounted on top surface of baseplate.
    Figure 11. Upper: ?at?eld image, signal level 140ke-. Lower: Target image.

    Figure 12. Photon transfer curves for all 48 channels in the RTM. Due to di?erent gain resistors used on the top and
    bottom of this REB, there are 2 distinct gain families on each CCD. Quadratic ?t to the variance vs. mean is shown
    (solid lines).
    Figure 13. (a): Crosstalk mask image, scaled to show aggressors alone; (b), scaled to show aggressors (red), positive
    (green) and negative (blue) victims; (c) crosstalk matrix (absolute values); (d) distribution of positive and negative
    crosstalk matrix elements.

    Figure 14. Crosstalk mask illuminates center CCD. No measurable inter-crosstalk observed.
    Figure 15. Left: read noise vs. pixel rate for the RTM con?guration. Error bars are rms of 24 channels. Blue arrow
    is required pixel rate for 2 second readout. Right: Nearest-neighbor crosstalk vs. pixel rate. Circles: measured in
    single-CCD test bench. Triangle: RTM con?guration at nominal pixel rate.
    Figure 16. Left: read noise vs. cold plate temperature. Right: video gain vs. cold plate temperature.

    Of the many participants in the LSST Science Raft collaboration, the following individuals are particularly
    recognized for their contributions to this work: N. Dressandt, J. Frank, G. Fraser, D. Huang, I. Kotov, P.
    Kuczewski, H. Lebbolo, W. Lu, G. Mayers, J. Mead, M. Newcomer, J. Panetta, V. Radeka, M. Reilly, P. Takacs,
    J. Triolo, V. Tocut, W. Wahl.
    This manuscript has been co-authored by employees of Brookhaven Science Associates, LLC. Portions of
    this work are supported by the Department of Energy under contract DE-AC02-98CH10886 with Brookhaven
    National Laboratory. LSST project activities are supported in part by the National Science Foundation through
    Governing Cooperative Agreement 0809409 managed by the Association of Universities for Research in Astron-
    omy (AURA), and the Department of Energy under contract DE-AC02-76-SFO0515 with the SLAC National
    Accelerator Laboratory. Additional LSST funding comes from private donations, grants to universities, and
    in-kind support from LSSTC Institutional Members.
    [1] Ivezi?c, v., Tyson, J. A., Acosta, E., Allsman, R., Anderson, S. F., Andrew, J., Angel, J. R. P., Axelrod,
    T. S., Barr, J. D., Becker, A. C., et al., \Lsst: from science drivers to reference design and anticipated data
    products," (2008).
    [2] Kahn, S. M., Kurita, N., Gilmore, K., Nordby, M., O'Connor, P., Schindler, R., Oliver, J., Van Berg, R.,
    Olivier, S., Riot, V., Antilogus, P., Schalk, T., Hu?er, M., Bowden, G., Singal, J., and Foss, M., \Design and
    development of the 3.2 gigapixel camera for the Large Synoptic Survey Telescope," in [Ground-based and
    Airborne Instrumentation for Astronomy III], McLean, I. S., Ramsay, S. K., and Takami, H., eds., Society
    of Photo-Optical Instrumentation Engineers (SPIE) Conference Series 7735 , 0 (2010).
    [3] O'Connor, P., Kotov, I., Takacs, P., Frank, J., Plate, S., Van Berg, R., Newcomer, M., Antilogus, P., Lebbolo,
    H., Tocut, V., et al., \Development of the lsst raft tower modules," in [SPIE Astronomical Telescopes+
    Instrumentation ], 84530L{84530L, International Society for Optics and Photonics (2012).
    [4] O'Connor, P., Radeka, V., Figer, D., Geary, J., Gilmore, D., Oliver, J., Stubbs, C., Takacs, P., and Tyson,
    J., \Study of silicon thickness optimization for lsst," in [SPIE Astronomical Telescopes+ Instrumentation ],
    62761W{62761W, International Society for Optics and Photonics (2006).
    [5] Radeka, V., Frank, J., Geary, J., Gilmore, D., Kotov, I., O'Connor, P., Takacs, P., and Tyson, J., \Lsst
    sensor requirements and characterization of the prototype lsst ccds," Journal of Instrumentation 4 (03),
    P03002 (2009).
    [6] O'Connor, P., Frank, J., Geary, J., Gilmore, D., Kotov, I., Radeka, V., Takacs, P., and Tyson, J., \Char-
    acterization of prototype lsst ccds," in [SPIE Astronomical Telescopes+ Instrumentation ], 702106{702106,
    International Society for Optics and Photonics (2008).
    [7] Antilogus, P., Bailey, S., Bailly, P., Lebbolo, H., Martin, D., Sefri, R., De La Taille, C., Jeglot, J., Moniez,
    M., Tocut, V., et al., \Aspic: Lsst camera readout chip. comparison between dsi and c&s," in [Topical
    Workshop on Electronics for Particle Physics (TWEPP-09) ], (2009).
    [8] Antilogus, P., Bailly, P., Jeglot, J., Juramy, C., Lebbolo, H., Martin, D., Moniez, M., Tocut, V., and Wicek,
    F., \Lsst camera readout chip aspic: test tools," Journal of Instrumentation 7 (02), C02044 (2012).
    [9] Juramy, C., Antilogus, P., Bailly, P., Baumont, S., Dhellot, M., El Berni, M., Jeglot, J., Lebbolo, H., Martin,
    D., Qureshi, A., et al., \Driving a ccd with two asics: Cabac and aspic," in [SPIE Astronomical Telescopes+
    Instrumentation ], 91541P{91541P, International Society for Optics and Photonics (2014).
    [10] Kotov, I. e. a., \Characterization and acceptance testing of fully depleted thick ccds for the large synoptic
    survey telescope," these Proceedings (7 2016).
    [11] Lu, W. e. a., \Ccd emulator design for lsst camera," these Proceedings (7 2016).
    [12] O'Connor, P., \Crosstalk in multi-output ccds for lsst," Journal of Instrumentation 10 (05), C05010 (2015).

    Back to top