Here are Fred Harris' notes from SDW2005. Please refer to the
conference proceedings for more definitive information.

SDW2005 Notes for CCD-world
Fred Harris
U.S. Naval Observatory Flagstaff Station
fhh@nofs.navy.mil

Scribe's Disclaimers:

- Many speakers talk faster than I can type. Please check the content for
accidental errors of transcription and understanding.
- This content is NOT summaries of the talks. Rather, I have tried to record what
was spoken that has not been presented visually.
- The content and views expressed herein are not the views of the U.S. Naval
Observatory, and have not been approved by the Editorial Board of the U.S. Naval
Observatory.

SDW2005 Day 1 2005-Jun-20 FHH

0. Dopodomani = The day after tomorrow

1. Roberto Gilmozzi: Science, Technology and Detectors for Extremely Large
Telescope:
- ELT = Extremely Large Telescope
- ELD = Extremely Large Detectors
- Second generation instruments for "mature" large telescopes
- Telescope apertures grow by x2 roughly every 50 years, but proposed ELTs will
grow faster than this "rule"
- The era of "improving" telescopes by improving detector QE has reached its limit
- Spectroscopy of objects seen by upcoming space scopes requires minimum of 30m
aperture
- AO for such large scopes requires 8K-9K actuators
- Dealing with wond loading a challenge for such ELTs
- Choice of ELT design driven by science
- Providing sufficient pixel counts in focal plane also challenging
- Radial velocities good to errors near 1 meter/sec, proper motions determine that
suspected exoplanets are tied to their parent stars
- 30m scope can yield 3-sigma detection of terrestrial exoplanet
- Simultaneous differential imaging: speckle in two simultaneous adjacent bands
allows subtraction of the speckle
- Thermal IR best place to look for terrestrial exoplanets so must be in space
- FOWL = frozen OWL (Antarctica)
- SN studies with 100m scopes gives look-back to pin down cosmology
- Pixel sizes down to 1.6 mas means 5Gpxl to cover 2x2 arc-min field, so detector
costs as much as telescope, or need cheaper detectors
- Scopes cost diam^2.6, so budgets dictate aperture
- Flat secondaries relaxes the mechanical tolerance for centration
- No Q&A.

2. Host Italian Agencies:
- Observatorio Catania
- Italian Institute of Astronomy and Astrophysics
- University of Catania

3. Sandro D'Odorico: Requirements on Array Detectors from OWL Instrument Studies:
- Hiring Beletec leads to better meetings
- Hiring Gilmozzi leads to spending money on better (bigger) telescopes
- OWL studies range from blue to sub-mm spectral range
- COsmic Dynamics EXperiment = CODEX: high-res spectrograph (150000) => 1cm/sec
radial-velocity stability, array of 5 echelle spectrographs, must be separate
from scope linked by fibers; detectors = 60 of 2Kx2K 15-micron CCDs with read
noise 1-2 e- rms
- Earth-like Planet Imaging Camera Spectrograph = EPICS: pixcel sizes on order of
1/2 mas, order of 8Kx8K pxl IR and CCD; AO sensor 3Kx3K at 1KHz
- QUANTEYE: ultra-fast photometer down to 10^-9 sec time resolution; SPAD detectors
(Single Photon Avalanch Photodiode) with test structures of diameters 25-100
microns diameter
- Large-Field NIR Camera
- MOMFIS: 18 of 2Kx2K IR detectors
- T-OWL
- SCOWL = SCUBA at OWL: 350 and 450-micron bands, 850-micron desired
- Q&A:
Q: Large focal planes doing mas astrometry: do believe sub-mas corrections can
be reached with ELTs?
A: Thinks can be done, Strehl ratios 20-30% reachable with reasonable
extrapolation of today's technology; achieving 70% Strehl will be a 2nd gen
of instruments
Q: So many focal-plane instruments: what is the plan?
A: Studies go back to the science cases, not an instrumentation plan; needs
priorities to be set.
Q: Instruments will be on line?
A: 6 focal stations, 1 for engineering, five for science.

4. Doug Simons & Paola Amico: Ground-based Observatories -- Instrumentation and
Detector Systems: Overview
- Study 31 different telescopes at least 3.5-m aperture & "snap-shot" of their
instruments
- No one observatory is large enough to "dominate" instrumentation
- bimodal wavelength "break" at 1-micron driven by technology
- 2 our ot 3 instruments are optical (CCDs)
- Break between Near-IR and Mid-IR is 5-microns
- 60% instruemnts are spectrometers, 25% imagers
- E2V wins on market share counted once/instrument, also for absolute number of
detectors (nearly half)
- Few instruments with pixels as large as 1/2 arc-sec, given large apertures which
give longer focal lengths
- In IR, 18.5-micron and 27-micron pixel sizes dominate
- In CCDs, dominated by 15-micron pixels with 13.5-microns in second
- 75% of future IR instruments will use 2Kx2K
- Current CCDs dominated by 2Kx4K, gong to 4Kx4K
- SDSU controllers dominate (1 in 4). Do we want to standardize within the
community?
- Future trends in 10-20 years: IR will be key, looking for earliest stars
- 1-2 micron regime due to red shifting
- 25 years from strip charts to current IR array detectors
- Want ELTs to catalog spectral types of exoplanets, study dark energy
- Want IR arrays to cost-per-pixel as CCDs do now
- Q&A:
Q: Present detectors driven by scopes built more than 20 years ago?
A: Older scaope might receive bigger focal planes, driven by economics.
Q: Survey of IR detectors and pixel sizes: context driven by technology that
A: Have "standardized" rather than having found optimal sizes, and detector
support that is easy to make.

5. Kirk Gilmore: The Large Synoptic Survey Telescope (and its Gpxl camera):
- Just starting R&D phase
- 65 cm diameter focal plane, 3.5-degree Field of View, want 4Kx4K detectors
- 2 exposures per pointing
- SDSS filter set
- Coming out of the KIPAC group; LSST Corporation in Tucson
- A*omega greater than 250
- Detectors:
- slicon greater than 75-micons thick
- high resistivity
- 10-micron pixel size with full well greater than 90K e-
- Segmented readout, 32 output ports
- moved individual on-chip amps about 300-miconrns apart to attempt to minimize
crosstalk
- 4-sude buttable package, 3x3 on a "raft", 5x5 grid of rafts
- want focal plane to be really really flat (given fast beam)
- Looking to pursue both CCD and CMOS technologies, also interested in Tonry's OTAs
- PSF better with focus **within** the silicon 10 micons, want flatness globally to
10 microns
- 20-30 terabytes/night
- Q&A:
Q: Clocked antiblooming?
A: No; segmentation of detector to limit extent of blooming
Q: Operating temp? How cooled?
A: -100C desired.
Q: When does LSST show up?
A: LSST shows up in 2012

6. John Tonry: Pan-STARRS and its Gpxl Camera:
- Similar to LSST but delivered sooner (2008 for Pan-STARRS4)
- Conventional RC optics, no new optics technologies involved
- If present state-of-the-art is SDSS, will go magnitudes fainter than SDSS
- 10 terapixels on the sky x 6 colors
- astrometry to less than 5 mas, photometry to better than 0.01 magnitude
- 3-element refractive corrector
- cheap small scopes means needs cheap pixels, need to drop per-pixel cost by x10
compared to current costs
- curved focal plane on SiC
- 16x16-inch final focal plane
- expect to monitor collimation on minute-by-minute basis with deployable WaveFront
Sensor (WFS)
- Calcite within WFS to give paired images each side of focus
- Electronics:
- controller has same footprint as detector
- 2 detectors per board set (preamp bd and DAC3U board)
- Each controller doing OTCCD shifting
- 256MB of RAM per controller, 15GB in entire camera
- refuse to use hermetic connectors, using rigi-flex through slots in camera wall
- AD9826 ADC
- No dual-slope integrator, multi-sampling with fast ADC, do fast summing and
subtraction in FPGA, brings ADC noise down 2 ADUs
- 3.8 e- noise realized at 500 Kpxl/sec
- test cam 1-2 90 Mpxl, test cam 3 360 Mpxl
- Q&A:
Q: peer-to-peer data org for how to archive pixels?
A: Expect a few petaBytes of storage; Air Force not happy with peer-to-peer.
Q: Why 4 scopes?
A: May be that LSST may morph into Pan-STARRS; detectors don't need to be that
expensive, even small scopes not that cheap.
Q: Where analysis of guide star?
A: Wants to take focal-plane info, read small sub-arrays, get into pixel servers,
analyze, take set of X/Y coords, do global analysis, do OT rransfer, iterate
on 100 ms basis.
Q: Guide loop: granulation on OTCCDs?
A: Each "cell" (600x600 pixels?) about isoplanatic patch.
Q: Compare digital CDS with analog?
A: Thinks digital will compare favorably to analog; sampling during "noisy"
transitions, have not explored weighting of samplings.

7. Dietrich Baade: Optical Detector Systems at ESO:
- NGC = New Generation detector Controller:
- prototype has seen first light
- one main board, second board differs depending on CCD or IR
- Multi-conjugated Adaptive Optics Demo: up to 400 frames/sec, 500 frames/sec in
2x2 binning
- >1K actuators, about 1KHz frame rate
- no real cryogeneics so will have dark current
- E2V developing CCD220: split-frame frame-transfer, 8 gain regs per CCD, 240x240
pixels, 24-micron pixels
- OmegaCAM: Safety -- electrical protection of CCDs, concern about continuous-flow
cryogenics system (see Olaf Iwert)
- X-shooter: 0.3 to 2.4-micron sinultaneous coverage
- MUSE: 24 individual spectrographs, each with 4Kx4K CCDs or equivalent mosaics
- Plasma cleaning tests well with big OmegaCAM cryostat: plasma-filter window, body
of cryostat is one of the electrodes
- Removing water from cryostat improves QE (similar to oxygen soak?)
- No Q&A.

8. Martin Roth: PMAS: 2 Years Experience with Nod & Shuffle 3-D spectroscopy:
- Review: move scope, shuffle charge, per SDW2002
- Always pair object and sky for each aperture, pairs are adjacent on CCD
(interlaced)
- Time overhead since object and sky exposures not simultaneous
- Needs telescope pointing to reregister field after a nod
- Q&A:
Q: Trade-off using nod-and-shuffle versus "noiseless" photon-counting system?
A: Not done, no plan to try photon counting.

9. Sebastian Deiries: Plasma Cleaning -- A New Method of Ultra-Cleaning CCD
Cryostats:
- Contamination more of a problem in the blue
- Is "conventional" cleaning sufficient?
- Plasma cleaning used by industry
- Simple, results in a few days
- Conventional Mechanical, ultrasonic, chemical, soap, baking) cleaning should be
done before plasma cleaning (SDW2002)
- After conventional cleaning, still have residual organic contamination which can
transfer to surface of the CCDs
- Plasma good to remove very thin films of oils, oxides and greases
- Use synthetic air, or oxygen; oxygen ions combine with hydrocarbons
- Plasma temperature about 10000 degrees C, 0.2 to 1 mBar
- LFG40 high-Voltage plasma controller from Diener
- Modified vacuum oven (any brand)
- Alcatel vacuum pump
- Gasses:
- Pure oxygen is best but high risk
- Ambient air is sufficient for metal parts and plastic parts, PCBs
- Argon works
- Can clean already-assembled cryostats without removing anything except detector:
- high-voltage anode applied OUTSIDE of cryostat window, cathode = case
- 10 minutes is sufficient
- Exhaust gasses should not be breathed, send outside of lab
- UV radiation stopped by glass (non-quartz) window of oven
- 20K Euros initial investment; 0.20 Euro cost per cleaning
- PCBs should not be exposed to more than 10 minutes in plasma, risk of becoming
black
- After plasma cleaning, must be handled strictly in clean-room conditions
- After cleaning, high AMU species gone
- Q&A:
Q: Cleaning PCBs with active components mounted?
A: Yes, no problems up to now.
Q: Difference with soldermask versus no solder mask on PCBs?
A: No difference.

Poster Pops:
- Derek Ives: UKIRT WFCAM:
- CCD co-mounted with 4 IR arrays, Faraday cage around the CCD including
metal-coated window, no crosstalk problem;
- Special HAWAII-2 clocking scheme removes reset clocking residuals.
- Kirk Gilmore: LSST
- Bruce Atwood: Detector Systems for MODS
- Decentered Schmidt camera
- Looked at Cryotigers, decided to stick with LN2
- Cryostat can handke up to 8Kx4K of CCD(s)
- 8192x2880 CCD new device, 15-cron pixels, 32 outputs
- Phillippe Feautrier: Thermal modeling
- CT1050 cryohead
- Thermal modeling of thermal transient(s)
- I-deas software tool used
- Phillipe Balard: Photon Counting Developments:
- Moving from hardware-based to software-based system to do centroiding and
photon counting
- GaAs photon-counting head
- Francis Beigbeder: Camera for large scale of solar surface
- high resolution and fast temporal cadence needed
- need 12Kx12K pixels; starting with CMOS 3950x4090 pixels, QE x Fill Factor
about 40%, 4 outputs x 10MHz = 2.5 frames/second
- Rainer Kramm: DISR camera on Huygens
- 1.1 Watts total power consumption, 512x256 pixel CCD
- Jean-Louis Lizon: Cooling system for the OmegaCAM
- 30 Watts direct radiation from the window, total heat load 60 Watts
- 155 K operting temperature
- circulated LN2, uses 90$ of filling capacity regardless of telescope position
- external LN2 storage tank
- Klaus Reif: Bonn Shutters
- 480x480 mm square for Pan-STARRS
- a good shutter catches every photon, is absolutely photometric
- Roland Reiss: CCDs for MUSE
- low-frequency pulse tube cooler (closed cycle cooler)
- Cascode CCD buffer state with differential output
- Martin Riopel: On-chip guiding Hawaii-2RG
- one star per each quadrant (detector)
- Master SDSU controller for science, slave SDSU controller for guiding
- Roberto Speziali: LBC
- SKYTECH CCD controller
- Binxun Ye: System design of detector systems
- should have 32 detectors
- mixed-gasses Joule-Thomson refrigerator, no maintenance for 5 years

Roundtable: Observatory Detector Needs (and dreams)

Pixel sizes and formats for IR arrays (or any detectors) -- make pixels smaller
as a means of lowering cost/pxl. CCDs cost per unit area, not per pixel. If
pixels are made smaller, problems with inter-pixel capacitance making crosstalk.
We have historically designed around what devices are already available, and new
instruments have not pushed for "custom" sizes. Discussion oriented to telesocpes
with AO.

What influence can an institute buying only a few detectors do?

Mfrs are coming to come to these conferences to find out customers' preferences.

E2V doing 12-micron/pxl, 10-micron/pxl coming up, 8-micron/pxl after.

What custom devices are available?

If an E2V custom divice is in production, it is on the web pages. Sometimes what
is being made for someone else, can be replicated. E2V does a lot of custom work
for space missions. ELTs should be counted as space missions in terms of their
budgets.

One cannot expect to go much (? larger / smaller ?) than 15-micron pixels. The
instruments for ELTs require many IR arrays, need a breakthrough in lowering the
cost of the detectors. Easier to sell to funders a large structure (such as a
telescope) rather than a detector system.

At Glasgow, question asked if halving pixel dimensions makes sense to the
instrument designers? How far does this push the optical tolerances used to
couple to the detectors?

In IR, with AO systems, forcing to smaller pixels?

No quantative answers to smaller pixels versus optical tolerances.

10-micron/pxl IR devices?

High-order adaptive optics, need to make sure that it can be done with smaller
pixels. Red response of CCDs with deep depletion? Does lateral resolution of such
thick devices suffer with smaller pixels?

What are the (electrical) effects of going to smaller IR pixels? 18-micron/pxl
IR devices have some errors, smaller-pixel CMOS devices have greater errors. Need
to think through these potential changes in device characteristics before making a
change to smaller pixels.

With respect to inter-pixel capacitance, the crosstalk can be corrected
numerically.

If this crosstalk is accurately know, it can be removed.

Optical CCDs are going to smaller pixels. In IR, undersampled due to living with
(larger) pixels.

What does this community want? Some of these sizes are determined by reticle
and/or stitching techniques. Developing stitching technologies are a lot of work.
Manufacturers want us to TELL them what we want. For HgCdTe, max size of the
substrate may dictate smaller pixels to be able to go from 2Kx2K to 4Kx4K, and by
going to 10-micron/pxl can go to 6Kx6K.

Can Paola poll contributors on wished-for pixel sizes?

Sure.

It is not just pixel size.

What we want to have is smaller pixels, but with LESS crosstalk. Must address all
the technical issues before the users will be happy.

Vote for longer serial registers: nod-and-shuffle has the dispersion along the
serial axis.

IR wavefront-sensing devices: high-speed, low-noise needed. Is 128x128 with 3 e-
rms noise, 1KHz frame rate, possible?

Readout-noise spec: with IR, can't do on-chip binning, so want to aim for lower
than 3 e- rms read noise.

Anyone in the audience wanting to do IR wavefront sensing? Yes.

There is a 1Kx1K for IR wavefront sensing, hoping for 7 e- rms at 1KHz frame rate.
This is discusses in a Glasgow paper by Gert Finger.

Imaging with with E2V L3 CCDs: will this be extended into the IR? Is there a means
of doing electron multiplication in the IR?

If there is interest in pursuing this technology, there is option to joing a group
for collaboration and cost sharing. Desire for IR guiders. Common solution
sought.

Are there other detectors (besides OTA) requested?

If you have an OTCCD, you don't need the long horizontal register, as the nod-and-
shuffle can ve vertical or diagonal.

Interest in arsenic-on-silicon detector that would work out to 27 microns.

Are we happy with read noises of a few electrons? Want to see more of is: the
development of large-format photon-counting technologies. Want some additional
work in reducing the read noise of even the present low-noise detectors (IR and
optical).

Want P-type CCDs off-the-shelf. Smaller field of view easier on flight missions.
Want ASICs supplied with the CCDs.

SDW2005 Day 2 2005-Jun-21 FHH

0. Magari: "I wish it were true"

1. Barry Burke, Paul Jorden, Paul Vu: Review of CCD Technology:
- "A CCD is a very simple device to make." This was not said 20 years ago...
- Because of rather coarse design rules compared to more modern technologies,
the CCD photomasks can be printed onto the entire wafer in a single exposure;
4:1 or 5:1 projection reduction for CMOS to yield feature sizes
- CCD substrate resistivities in range of 1 to 20000 ohm-cm; CMOS has strict
requirements on resistivity of substrate material
- 3Kx6K 10-micron/pxl CCD in same wafer footprint as 2Kx4K 15-micron/pxl "standard"
CCD
- (Anti)Blooming drains within channel stops, more effective than clocked
antiblooming by x100 overexposure level
- Electronic shutter for back-illuminated CCDs require tri-level clocks
- E2V uses ion implantation followed by laser anealing for backside treatment
- Hafnium Oxide is typical AR coating
- For thick-silicon CCDs, trade decreased fringing at long wavelengths for
increased cosmic-ray collection and increased PSF
- A two-stage on-chip output amplifier is now typical
- For charge-multiplying CCDs, gain has a temperature dependence so regulation of
of operating temeprature is more critical than a standard CCD
- E2V 4Kx4K offered with integral dual flex-PC binded direct to the CCD (more
usual for space applications)
- In MTF, the diffusion MTF loss is the most-significant factor
- Ionizing radiation damage mitigated by absence of gate voltage during radiation
exposure
- Both CMOS and CCDs have RTS noise effects and dark-current increase as
radiation effects, but CMOS more immune to radiation-induced voltage shifts
- P-channel CCDs more radiation tolerant (less phosphorus) than N-channel CCDs
- In CMOS, three-transistor (3T) pixel is out-performed by five-transistor (5T)
"pinned" photodiode structure, which has 1/10 the dark current of the 3T pixel
structure
- CCD-CMOS hybrid: CCD does image capture, CMOS does the charge-to-voltage
conversion, allows 5 e- rms readout at MHz pixel rates
- Q&A:
Q: Why not antiblooming channel stops in every CCD?
A: Expected greater interest when presented at prior meetings; almost no
performance penalty (slight decrease in well capacity) but no QE penalty;
can't be used on OTCCDs.

2. Dick Bredthauer: Siliconus Maximus:
- OTA has NMOS logic incorporated with "classic" OTCCD structures
- STA0500A 4Kx4K has 2.8 e- rms at 50KHz with full well >80K e-
- 4Kx8K frame-transfer with 8-mcron pixels
- 10Kx10K on 6-inch wafer with readout of entire CCD in 10 seconds using multiple
outputs, 9 micron/pxl
- Cost estimate of 6.4 Gpxl array = $16M - $22M USD
- Q&A:
Q: What sort of flatness to expect on 10Kx10K?
A: Don't want to speculate yet.
A: (Barry Burke) Depends on the packaging, can be as good as 5 microns for a
wafer-scale device.

3. William Kolbe: CCD Development Progress at LBNL:
- SNAP CCDs 4Kx4K with 10.5-micron/pxl, read time 30 seconds; diffusion 4-microns
rms; version 2
- DALSA now fabbing high-resistivity CCDs on 150-mm wafers
- 650-micron full wafer thickness, gets thinned to 200 microns
- At 650-micron thickness, 207 Volts to fully deplete
- 200-micron-thinned back-illuminated pushed to 206 Volts (power supply limit),
spec cals for 80 Volts
- Dalsa has had "vendor ccontamination" that is "fixable"
- Backside finishing with high-temp steps done before front-side aluminum
- Building own cold-probe station in-house
- For QE, reference photodiode in same dewar as test CCD
- Polysilicon for backside gate, poor UV response
- Q&A:
Q: How do you tell at what voltage fully depleted?
A: Looking at the "puffiness" of cosmic rays.
Q: Hot pixels, what found for radiation testing?
A: For radiation, tested with 88-inch cyclotron, lower CTE damage for given dose
compared to standard CCD. There are sometimes hot pixels produced, these CCDs
don't take infinite doses. SNAP procedure does 3-4 images with median
filtering to deal with cosmic rays.

4. Robert Hartmann: A pnCCD Detector System for High-speed Optical Applications:
- pnCCD: fully depleted 3-phase, no serial register, each column has individual readout
- Substrate n-type 2.5K ohm-cm, p+ registers, p+ backside diode
- very high blue and UV QE
- 450-micron thickness, fringing is negligible
- 5-10 oxide layer on backside approximates measured response, yields about 65% QE
over wide range
- QE curves similar to E2V
- 51-micron/pxl, 256x264 pixel image area, split frame-transfer
- 25 microseconds to move from image area to storage areas
- 70 Mpxl/sec with 8 readout amplifiers: 1.8 e- rms at 500 frames/sec, 2.3 e- rms
at 1000 frames/sec, -55 C operation
- Readout ASIC, nulti-correlated double sampling (MCDS), electroinc noise
contribution less than 1 e- rms
- first line available with latency less than 40 microseconds
- Q&A:
Q: 51-micron pixel is fairly large; how small can be made?
A: 36-micron (Y) x 50-micron (X) dictated by JFET geometry
Q: Entrance window?
A: Very shallow surface, and p+ implantation.

5. Gregory Prigozhin: X-ray CCD With Low-noise Charge-injection Structure:
- CCID-41 with input injection:input diode > input gate > imput serial register
- variation of "classic" fill-and-spill operation: only one input gate, clock
simultaneous with first gate of the serial register
- 0 - 3K e- input range needed, works well at very low charge injection levels
- Capacitance based on size of the input-S3 gate gives an order of magnitude
greater than the measured 0.43 fF
- Charge injection noise 8-12 e-, greater than prediction but similar shape of
curve; about 6 e- RMS for less than 1000 injected e-
- presume not in sub-threshhold mode: ? electron evaporation ?
- with injected charge on 40 MeV proton-irradiated CCD, nearly comparable
performance to non-irradiated CCD
- readnoise less than 2 e- rms at 42 KHz
- Q&A:
Q: Column-to-column uniformity of injected charge?
A: No column-to-column injection non-uniformity;
Q: Input transfer gate structure?
A: Yes, input transfer gate.

6. Simon Tulloch: L3CCD Wavefront Sensor Developments at the ING:
- Rather than seeing Gaussian noise in a bias frame, dominated by Clock-induced
charge (CIC), about 1 in 10 pixels in the image area
- range of output singals in response to single e- input, normal photon statistics
do not apply
- multiplication noise has the effect of halving the QE, disadvantage at high
signal levels; cross-over with conventional CCDs at about 30 e- input signal
levels
- needed for guiding on "natural" star
- looking to get 1-2 magnitudes improvment with L3CCD as compared to conventional
CCD in guiding application
- custom-built multiplicatoin board added to stock SDSU controller
- sub-electroni noise at 180 frames/second
- Q&A:
Q: Comment on problem of spurious charge with current technology?
A: Got it down to roughly 1 e- in every 10 pixels; level measured for a Craig
Mackay chip an order of magnitude lower
A: Important to clock quickly, can reduce CIC to 1 even in 1000 pixels.

7. Mark Downing: A Dedicated L3CCD for Adaptive Optics Applicatinos:
- Looking for exoplanets with luminosity rations > 10^5
- Wavefront sensor (WFS) detector is the critical element
- Next-gen 40x40 Shack-Hartmann system, needs 240x240 pixel L3CCD
- want a device for any type of WFS
- want 1.2 KHz frame rate, hope to get to 0.1 e- rms effective read noise
- lower read noise is better than higher red QE even for read guide star
- "metal buttressed" 2-phase image-area clocks, reduces frame transfer smear
- first time to try L3CCD in high-resitivity
- have a design study to add an electronic shutter to this CCD design
- on-chip power dissipation reduced by the metal buttressing
- Q&A:
Q: HOw much lateral diffusion expected in thick deep-depletion device?
A: May have some problems there.
Q: Apply bias to backside?
A: No, but thought about.

8. Craig Mackay: Near-diffraction-limited Visible Imaging on 10m-class Telescopes
with EMCCDs:
- KSC-1000 sequence generator from Kodak for pixel rates 4-60 MHz
- Very fast serial clocks, rise and fall times 6-8 nanoseconds
- L3CCD camera runing between 20-100 frames/second for lucky imaging, frame time of
80 ms
- if you are doing photon counting, then by threshholding you recover the L3CCD QE
- lucky imaging should deliver x3 improvement in seeing at any facility, propose
that x5 to x7 should be realizable
- Q&A:
Q: Strehl ratios in these images?
A: 0.25 realized; speculate that some frames are compensating for mirror
inaccuracies
Q: When doing recombination with interferometric triples for lucky imaging, so
the different triples need to be simultaneous?
A: No need for differing aperture triplets to be as the same moment.

9. Giovanni Bonanno: Single-photon Avalanche-diode Arrays:
- SPAD in array form is SPADA
- SPADA chip cooled by Peltier, ceramic holder/substrate
- need integrated Active-Quenching circuits (iAQC) one for each SPAD in the array
- iAQC integrated circuit realized
- active quenching takes a maximum of 20 nanoseconds
- in collaboration with ST Microelectronics (STM)
- this device suffers from traps, lads to spurious counts
- dark signal 300 - 1500 counts/second at room temperature
- tester illuminates 10-micron pinhole, 5-micron scanning resolution
- 5x5 40-micron SPAD array manufactured by STM
- Q&A:
Q: With 20 ns quench time, what photon counting rates?
A: 50 MHz/3, or so.

10. John Tonry: The Orthogonal-Transfer Array:
- 3 second read time is for 8 of 600x600 OTCCD cells in a column
- about 50% yield on wafers regardless of wafer variants
- metal over gates appears successful, no yield hit, but can see reflection from
metal at very red wavelengths at 10% level
- logic on the OTA works, non-functional cells do not affect neighboring cells
- expect within 24-hour cycle, load & cool & test 4 devices; need to screen about
300 devices over the whole project
- cell can be active (OT shifting or readout) or standby (integration) or
disconnected (isolate defective cell)
- there is some detectable intrinsic device crosstalk, but less than the channel-
to-channel crostalk from a dual-video Leach board
- less easy to see CCD degredation with these small 600x600 cells
- project test image, shift in square path 12800 times, subtract from original, see
2x10^15 image shift
- 40V backside bias really confines charge diffusion well
- any charge diffusion less than 3 microns is acceptable
- LDD for high-voltage logic inverters and pass transistors may minimize glows
- Q&A:
Q: Two intra-pixel shapes: 2 triangles with surrounding rectangles versus 4
triangles; why pick one or the other?
A: Both yield about the same good devices. Barry Burke prefers 4 triangles,
John Tonry prefers 2 triangles with surrounding rectangles, so the 2 triangles
geometry will be used.

11. Brian McLeod: MMT Megacam Performance:
- less than 10 seconds to read entire array with custom in-house controller
- 2 of IR Labs ND-14 dewars for cooling
- CCDs mounted on Invar plate, plate mounted to ring via 6 flexures
- ZIF connectors for CCDs need to be flush on their flex PCBs
- TVSS protection devices soldered onto flex PCBs
- Test box to connect to PCB that mimics a CCD, to allow to check wiring to each
CCD location
- 2 pieces of 5mm glass in each filter holder
- shutter is 2 blades driven by ball screws from stepper motors
- only fill each dewar 1/2 full, gives 36-hour hold time
- separate CCD control chassis for eah half of the mosaic
- about 4.5 e- rms read noise at 200K pxl/sec
- hardest part of instrument has been understanding sources of stray light in the
instrument
- initially had "bent" focal plane from materials thermal CTE mismatch, now fixed
- instrument is thermally cycled on monthly basis, not kept cold all of the time,
astrometry between runs a little worse than astrometry within an obsedrving run
- Q&A:
Q: Any means used to verify metrology for flatness?
A: Image quality.
Q: Flatness from image quality?
A: Near 100 micron peak-to-valley level. f/5 beam.
Q: Any other baffling around mosaic?
A: Did add black baffle after initial construction, inside the cryostat, 3 mm
in front of the CCD surfaces.
Q: Any contamination with so much in the dewar?
A: See a slight fog on some CCDs, when ou cool down, still puzzling. Did try to
bake everything out before going into the cryostat.

Poster Pops:
- Ricardo Schmidt: CCD Mosaics at CTIO and SOAR
- SOAR telescope designed for high image quality, still being commissioned
- 4 kinds of 2Kx4K CCDs used in different mosaics:
- 8Kx8K with SITe SI002A
- Jim Beletic: A New CCD Optimized for Pulsed Laser Guide Star Wavefront Sensing
- Rockwell works on CCDs
- Try to make AO CCDs that are photon limited
- Planar JFET for lower read noise being tested
- Sample a pulsed laser to look at different times = different altitudes
- Philippe Feautrier: Zero Noise Wavefront Sensors Development
- Paul Jorden: Optimised CCD Antireflection Coating
- 2Kx4K with AR coating with graduated thickness, so one side optimized for blue,
other side optimized for red, good for fixed-format spectrograph
- Martin Roth: Multiple CCD Detector Systems for PMAS
- LED glow in one quadrant of 4Kx4K 2-chip modaic
- Jean-Claude Labiche: A Fast Multi-channel CCD-based FReLoN Camera
- using Atmel 7899M CCD
- 10 Mpxl/channel
- 14-bit dynamic range
- David Darson: CCD CAmeras for Nano Semiconductor Objects Optical Properties
Research
- Sony ICX285AL gives 30 e- rms read noise at 20 frames/second
- Just started development of IR system for InSb
- Peter Poole: Develolpments at E2V
- Secure 8-micron pxl process by early 2006
- Adding 6-inch line to 5-inch line with no shut down of 5-inch line
- CMOS imaging at 350 nm

(13. Satoshi Miyazaki: Next Generation Wide Field Camera for Subaru: HyperSuprime:)
- Not given.

14. Bonner Denton: Advanced Detector Technology for Chemical Analysis:
- 6000 ions/second for conventional mass-spec minimum signal
- not suitable for isotope ratio spectrometry
- want individual gains per pixel
- current detection limit 8 ions with 2.5 e- rms noise
- Imager Labs has fabbed newest drift-tube detector
- detector is essentially a mux for an IR detector fed by a Faraday cup
- no Q&A

12. Olaf Iwert: The OmegaCAM 16Kx16K CCD Detector System for the ESO VST:
- two loops: one for guiding, then switch to image analizer CCDs on the same
controller
- working with 36 detectors, need individual testing before integrating into array
camera
- detector baseplate is aluminum, have had good experience with this, wanted
better thermal conductivity than Invar
- all wiring within vacuum is fles/rigid PC boards
- biggest problem was to achieve flatness of optical baseplae, did diamond turning
with multiple thermal treatments
- parts too large for conventional cleaning ovens and ultrasonic, so in-situ plasma
cleaning
- Flex/rigid boards are glue free
- Pt100 temperature sensors
- 2 deg C temperture homogeniety over the array peak-to-valley
- baffling over bond-wire areas, Kevlar coating on baffles
- Have machine to measure flatness through window, so can be measured cold
- peak-to-valey flatness of 25 microns at operating temperature
- Q&A:
Q: Transient voltage surge suppression inside the vacuum?
A: Olaf wants to talk more about this.
Q: Total cost?
A: Expensive, given constraints, much contracted out, but no total available now.
Q: Compromise between mechanical interfaces for access versus mechanical
tolerances?
A: All parts are hanging from the front flange, have eliminated a number of
thermal levels, Only compromise is moving one type of board to get at central
detectors.

SDW2005 Day 3 2005-Jun-22 FHH

0. Di Piu' = "more or even more"

1. Markus Loose, Alan Hoffman, Vyshnavi Suntharalingham: CMOS Technology
- monolithic CMOS, 4Kx4K with 5-micron pixels
- processes down to 0.13-micron (0.09 and 0.06-micron in industry, has not
transitioned to imagers yet)
- random access (read or reset)
- binning is possible but more tricky than in CCDs
- shuttering: snapshot (all pixels expesed simultaneously, more transistors per
pixel, higher noise) or rolling (less transistors per pixel and lower noise)
- 22x22 mm CMOS dies typical maximum without stitching
- 8-inch or 12-inch wafers available in CMOS technologies, stitched CMOS can in
principle be wafer-scale
- monolithic CMOS inherently means reduced fill factor and lower QE than CCDs:
- silicon layer is thin (reduced red)
- top layers absorb blue
- first time that Rockwell and Raytheon have shared a paper
- 16M indium bumps per Sensor Chip Assembly (SCA) currently realized
- pixel counts are doubling with 18-month time
- CTIA = capacitance trans-impedance amplifier, keeps detector bias constant during
integration, good for fast frame rated, very linear but has FET glow, higher
power
- InSb doesn't have a thermal CTE problem bumping to Si because InSb is "stretchy"
- RTS noise seems to be more prevalent when using finer CMOS processes
- CMOS can get down to 2 e- rms noise but requires multiple sampling
- using first 1/3 and last 1/3 of samples in Fowler sampling realizes the noise
minimum
- CCDs are very sensitive to "slip dislocation" defects, will require future fab
equipment mods to deal with future larger-diameter wafers
- CMOS can offer shaped implant profiles, multiple gate-dielectric thicknesses so
lower operating voltages; complementary doped polysilicon allows matching of P
and N threshhold voltages
- the 4-layer poly of OTCCDs is a process engineer's nightmare
- in CMOS, silicides are blocked from forming in the pixel area to minimize defects
(?)
- stacking faults are sources of dark-current defects, problemmatic for
astronomical applications
- vertical integration = 3-D stacking:
- photodiode layer on top
- readout layer in middle
- optional base layer
- allows back-illuminated
- Geiger-mode avalanche photodiodes:
- photon counting
- time-tagging allows distance determination in terrestrial applicaiotns
- time resolution down to 0.1 nanosecond
- Q&A:
Q: APD photons per second? Blind time?
A: Blind time about 1 nanosecond.
Q: HgCdTe used in the visible in astro, with substarte removal?
A: 0.4-micron flying in aircraft for about 8 years.
Q: but in low-backgrond astronomy?
A: not except with InSb.
Q: Linearity of CMOS versus CCD?
A: Advantages for CCDs. Anticipated that CMOS will improve in the future,
requires more effort.
Q: Cost per pixel for giant hybrid focal plane?
A: Approaching that of high-end CCDs for hybridization with silicon, going to
1 cent per pixel.

2. Bernie Rauscher: The James Webb Space Telescope (and its IR Detectors):
- SCA control by Rockwell SIDECAR ASIC, one per SCA, 16 total on JWST
- 7.1-degree K requirement for MIRI cooling accomplished with cryocooler
- Q&A:
Q: Basic cost for JWST, cost overruns?
A: Article in Space News; $1B "mass challenge", NASA has formed 3 panels for
science, cost and a third panel. A process going on to look at the cost.

3. Gert Finger: Performance and Evaluation of Large Format 2Kx2K MBE-grown HgCdTe
Hawaii-2RG Arrays Operating in 32-channel Mode:
- the promise that MBE would remover persistence has not been confirmed
- linear CMOS cryo-opamps used instead of ASICs, operate at LN2 temperature
- symmetric 2-opamp amplifier for differential signal chain
- 68 cryo-opamps dissipate 1 Watt, have temperture of 150 K (detector - 90 K)
- using MBE gets dark current below 10^-2 e-/sec at temp < 80 K
- operating temperature is determined by the cosmetic quiality of each individual
device
- with MBE arrays the QE doesn't drop with temperature
- using shot noise to calibrate gain, get greater than 100% due to errors from
capacitance
- 66% QE in Z-band, better than CCDs
- glow centers are important for selecting science-grade arrays
- have science-grade arrays with no glow centers at all
- don't know the origin of glow centers
- persistence on all arrays tested, can be seen for many hours
- no persistence at all if you do not saturate the array
- ? traps close to the PN junction ?
- can completely get rid of low-frequency noise by interpolating from start and end
reference pixels
- advantage of reading guide window non-destructively is gain of factor of 2 in
bandwidth, does not disturb the science frame
- must reset the guide window before it saturates
- additional ADC board for guide window, with associated fiber link
- Q&A:
Comment: the diffrence in dark current depends on the detector substrate, not MBE
(Peter Love).

Poster Pops:
- Ludovic Duvet: Active Pixel Sensor Developments for Future ESA Space Science
Missions
- Alan Hoffman: 1Kx1K Si:As IBC Arrays for Mid-IR Astronom
- more details for MIRI instrument
- real QE improvement is at 5-10 microns
- new readout Aquarius 1Kx1K for high-background astronomy
- Alan Hoffman: 2Kx2K InSb Focal Plane Array Performance
- Phoenix detectors have just 4 outputs
- Randy Campbell: The Effect of Charge Persistence in Aladdin III InSb Detectors
on Scientific Observations
- persistence fits nicely to a 1/t plot, have to wait for 10000 seconds for it to
go away
- Matthew Brown (Roger Smith): Optical Characerization of 1.7-micron NIR Detectors
for SNAP
+ Roger Smith: Noise Characterization of 1.7-micron NIR Detectors for SNAP
- SNAP has no cryogens, just radiative cooling
- poster compares Rockwell and Raytheon on the same plots
- optimizing AR coatings on a per-filter basis might allow QEs to approach 100%
- Michael MacDougal: Rockwell Scientific Imaging Technologies
- SIDECAR ASIC: 4-channel flight-based parts, then 32-channel ground-based parts
- Ricardo Schmidt: Focal Plane Detectors for Dark Energy Survey
- need good Z-band QE, plan to use LBNL chips
- focal plane 1/2 Gpxl via 62 of 2Kx4K, 2 read channels per CCD
- FermiLab still moving up CCD learning curve, will do full-scale system test

4. Peter Love: 2Kx2K NIR HgCdTe Detector Arrays for VISTA and Other Applications:
- 20-micron pixels
- 16-output mode allows 1 Hz frame rate, 2.5 microseconds/pixel/sample
- 2 pins for each module on mounting plate register X and Y, 3 pads register Z
- SCA flatness 3.1-microns peak-to-valley, routinely achieving 5-6 microns peak to
valley
- still working on dark curent, getting close to SNAP specificaitons
- Q&A:
Q: What is the rotational accuracy of the SCA in the package?
A: 2-3 microns over 40-mm length of the SCA.
Q: Have you qualified the flatness of the whole mosaic at cryogenic temperatures?
A: Can't do; must be done in the UK since there the focal plane is assembled.
Q: Has the flatness been simulated?
A: Yes, simulations say less than half of 25-micron specification.

5. Fernando Pedichini: ENERGY: A Proposal for a Multi-band CMOS Imaging Photometer:
- investigating Foveon detectors
- color filter bands determined by the silicon construction
- can reduce Foveon color bands to standard Johnson color bands
- rolling shutter, but each band is synchronous to each other band
- 12-bit camera
- Q&A:
Q: Could use microlenses for fixed f/ ratio to improve QE?
A: Yes.
Q: How does this device work, flux or fluence detector?
A: Need to contact the manufacturer
Comment: integrated into camera by Sigma.
Comment: color passbands depend on penetration depth versus wavelength, so don't
hold out much hope of altering passbands.
Comment: Rockwell is producing some duaal-band IR detectors, called dual-band
imagers, produced for government customers, later (?) for astronomy?
Comment: Could be a limit to this technology.
A: More band overlap than Johnson filters, cannot guarantee the interaction
depth of any individual photon.
Comment: Filter-wheel system much less complex than spectometer. But dichroic
camera is best for multi-color photometry.

----------

Roundtable: CCD vs. CMOS

Kodak is up to their own CMOS program, but aimed mostly at mass markets, have
their own fab in Rochester, New York.

Need to distinguish between monolithic and hybrid CMOS. Hybrid allows optimization
of each material and operating parameters, gives more flexibility. But a hybrid will always be more expensive, cost will always be a problem compared to
monolithic.

Astrononmy will not allow the waste of photons. Monolithic CMOS cannot deliver
thick back-illumimanted CMOS sensors. Can this be disputed?

E2V believes that back-illuminated CMOS can be done, and they are starting on a
3Kx4K device; believe there are benefits to having both CMOS and CCD
back-illuminated. E2V will develop their own design capability, particulary for
space applications.

A basic problem in monolithic back-illuminated CMOS. Only way to do the
optimization is by hybrids, or 3-D stacking. To get the cost down, must use
industry-standard tools.

Have to look at commercial market when looking at future of CMOS. The commercial
market is complainig about reduced sensitivity (QE) with shrinking pixel sizes, so
a prediction made that the market will help.

Physical silicon will be too thin in monolithic CMOS for good red QE.

What you see in CMOS is being driven by cell phones. Manufacturers of cell phones
are getting around lack of sensitivity by adding a LED for added illumination, and
are not going the way wanted by astronomers.

Not all obervers have $100M scopes, a lot of imaging being done on smaller scopes.

If have instrument with multiple bands, optical and IR, sometimes a CCD is right to
use, sometimes IR hybrid.

Want best of hybrid CMOS.

Hybrid silicon-on-mux compares to best of CCDs in QE and performance overall. In certain areas CMOS is better, space based CMOS with Si PIN on top is better in rad hardness. There are still areas where CCDs are the better alternative (TDI for
example).

If we could combine both words, what to do to combine CCDs and CMOS for multi-band?

Sometimes better to do hybrid rather than monolithic. Can make CMOS readouts or
interline CCDs or split-frame CCDs; all can be hybridized. Without indium bumps,
can up the density.

What are the limits to making the hybrid equivalent to the 2Kx4K CCD? One in which
a 2Kx4K-style CCD is the "detector" layer, and underneath is a CMOS layer that
provides the clock and bias drives, does the analog signal chain and CDS, does the
A/D conversion, and offers resultant digital data. (Fiber I/O is probably too much
to ask... but what about CMOS LVDS digital I/O?)

Think it can be done right now, in hybrid.

At what point do you transition from CCD to CMOS? CCD in the center, with CMOS
underneath. What is the advantage of a CCD hybrid over a CMOS hybrid? No
non-destructive readout with the CCD/CMOS hybrid.

Will we benefit from vast CMOS economy? That is not what the global community
wants; hybrid is too expensive.

Some things that CCDs can do superbly: TDI, OTCCD. In the orthogonal-transfer
array (OTA), the "pixels" are now CCD (sub-)blocks. Considered addressing each CCD
sub-block, with CMOS bump-bonded on a per-sub-block basis. That has a lot of
appeal; hope to see more marriage of CCD with CMOS rather than CCD only or CMOS
only.

Cost of CMOS hybrids: presently like Santa's workshop (with the elves working
away). For the high-quality production at lower cost, need to get more like a BMW
factory. Need economy of scale. With big projects, there is a possiblity of this
economy. Nice with CMOS to get rid of the shutter in a big imager. ("Sorry, Klaus
Reif.")

If we are building a monolithic CMOS imager, resistivity is picked by the
properties of the transistors rather than for optical properties, so one loses the
red response. Advances that improve QE won't happen via CMOS for cell phones.
There is a fundamental difference between CMOS and CCDs: one does the CDS on the timescale of the exposure for CMOS; on the CCD one does the CDS on the timescale of
the pixel. We saw a diagram today that can break this problem: the transfer gate
within the CMOS pixel (4T-type pixel design). If can realize CMOS that does CDS on
pixel timescales, good opportunity.

It is said that one of the advantages of CMOS in big industry is smaller pixels, lower voltage, so where is the dynamic range going?

As volts get smaller, output voltage swing gets lower, but noise floor then needs
to get lower to maintain dynamic range. A limit will be reached.

Discrepancy between microprocessor, memory and CMOS imager processes will grow with
time. Maybe only use the thin-gate low-voltage transistors for the logic on CMOS,
but not for the pixels.

Analogy with CCDs: divergence from main CMOS development.

Process options: higher-voltage process, versus photon counting with the
small-geometry CMOS.

So many dollars into CMOS, solutions will be found from money point of view.
Not so with CCDs.

? Sceptical about big CMOS development, the public market does not care about
linearity, crosstalk. Think the directions of public-consumable CMOS imager
development will be useless for science detectors.

Think it is not true that that cell phone application of CMOS imagers drives
technology advances; instead digital cameras are the market that will drive
technology advances, so dark current rates matter. People are complaining about
lack of sensitivity in digital cameras. In favor of monolithic CMOS because CMOS
is not that bad. Reference: Cyril Cavadore's work reported on CCD-world a few
months ago. (But, did Cyril get the right conversion gain and therefore the right
read noise and dark current when he used photon statistics to determine his gain?)
Single reads on hybrid CMOS imagers are too noisy, but maybe not true on monolithic
CMOS imagers.

With hybrid CMOS-based imagers, non-bumped (reference or not-connected) pixels have
1/2 the noise of the (successfully) bumped pixels; the excess noise is from the
detector material. Need to find out the source of the excess noise. Can do
hybrids without indium bumps.

Bumps needed for devices to be cooled. In commercial CMOS, use solder bumps to
mount in package. Bumps can come back from the fab house pre-mounted onto wafer.
Bumps may get less expensive.

Noise of CMOS: (random telegraph) RTS noise will be more noticeable with smaller
features.

Indium bumps: people have been predicting the demise of bumps, but it hasn't
happened. There was a prediction that one could not get small pixels, and that
has not happened. Don't see a limit right now. In production, bump costs have
come down. Raytheon does 15K hybrids/year, and have delivered 25K hybridized
parts.

A lot of people are working to solve CMOS problems. Astronomy will be one of the
last camps to fall to CMOS.

CMOS is not where we can buy 250 of these deivces. CMOS in short term can't
replace CCDs.

Nobody (non-astronomy) wants red/IR, commercial CMOS imagers want to match the
response of the human eye.

----------

6. Maurizio Seracini: Multispectral Imaging Technologies for Conservation of
Cultural Heritage:
- Started professionally in Bio Engineering, then Electrical Engineering
- Has always been interested in art
- Did some years of midical school
- In 1977, Editech. www.editech.com
- Restoration techniques require technical knowledge
- Most restorers are artisans, but not scientifically trained
- On masterpieces, few have full range of diagnostics performed
- No scientific companies are developing instrumemtation for this field
- Must modify existing technologies from some other application
- Started his career in 1975 by accident, using thermography
- Applying medical field techniques
- Across spectrum, using from gama rays to radar
- With X-rays: mostly using soft X-rays 10 KeV to 80 KeV, with berylium windows,
10 mA
- With microwaves: long wave 320 mm to 400 mm
- Ideal for before, and after, restoration
- Marble (calcium carbonate) changing to calcium sulphate, which gives a porous
texture appearance (as evidenced by Michelangelo's "David")
- With IR:
- 700 nm to 900 nm using film as the detector
- Out to 1.1 microns with CCDs
- With vidicons to 2.3 microns
- PbS contrast worst than that of CCDs
- Can see under-drawing on ground (base) layer, through varnish and pigment
layers
- IR single-spot scanner with InGaAx photodiode, 14-bit gray levels
- Under-drawings typically done by the acknowledged artist, but over-painting may
have been done by the artist's helpers
- Thermography:
- Great for hidden architecture
- Good for planning earthquake reinforcement
- Geo-radar:
- Earlier have used 1500 MHz down to 16 MHz, now up to 7 GHz
- Can spot voids (crypts) under the floor
- Question: What wanted for a "dream machine"?
Answer: Wants to see if a mural is underneath a wall -- how to do???
- An excellent talk.

SDW2005 Day 4 2005-Jun-23 FHH

0. Uffa: used to express exapsteration or extreme boredom; Usually accompanied by loud sigh.

1. Jim Beletic, Don Figer, Gert Finger, Peter Love, Roger Smith: Detector Testing &
Characterization
- CMOS devices ore complex to test than CCDs
- Constant cadence may be required to "calibrate" out behavior that may be
temperature-transient induced, particularly in IR detectors
- expect users (astronomers) to have a hard time grappling with complex detector
behaviors; expect this to be hard to communicate
- need to identify different populations of pixels in a detector, which may need
different treatments versus population
- parameter optimizatoin often leads back to device theory at the manufacturing
level
- what is the conversion gain in a CMOS device? -- less well understood than in
CCDs
- whether collecting electrons or holes, express internal quantities in terms of
electrons
- for CMOS source followers, gain changes (typically decreases) as full well is
approached, but at least with monotonic behavior
- what is seen in CMOS devices with small (18-micron) pixels, see cross talk to
adjacent pixels from what should be single-pixel cosmic ray events
- conversion gain raises its ugly head when QE exceeds 100%
- conversion gain by capacity (capacitance) comparison:
- replace Vreset-drain with big external capacitance
- look at discharge of Cext as pixels are read, sum, realize node capacitance
- must remove protective zener diodes on reset-drains due to their leakage
current
- use normal data acqisition chain for reading output; must capture both
conversions from before and after reset
- know external capacitance, learn ratio of capacitances, determine sense-node
capacitance
- discrepancies between ratio-capacitance-determined versus shot-noise-determined
can be up to factor of 2
- inter-pixel capacitive coupling, small but adds to a non-negligible effect
- for uniform illumination, no signal charge stored on coupling capacitances
- Andrew Moore's SPIE 5167,204 paper arrives at same conclusion by different method
- variance overestimates nodal capacitance
- coupling smears, and has cross-correlation/coupling between neighboring pixels;
if no capacitive coupling, no correlation between neighbors
- Fe55 works on InSb (tested on Aladdin): ??? who knows how many e-/photon for
K(alpha) in InSb ???
- as array pixcels grow smaller, must increase these considerations
- Q&A:
Comment: lateral diffusion problem contributes more crosstalk in smaller pixels,
AND interpixel capacitance also smears
Comment: capacitance coupling is a completely deterministic process
Comment: Look at power spectrom of fourier-xform of picture, see roll-off at
high frequencies; as long as sampling well compared to PSF, can live
Comment: do auto-correlation function
Q: impact on observations done in the past?
A: use calibration stars, efficiencies due to atmosphere, telescope, detector,
then nothing changes
Comment: S/N stays the same (?)
Q: does inter-pixel capacitance have a wavelength dependence?
A: No: we were looking at chart with engineering vs science arrays, not true
difference within a single detector.

2. Vincent Lapeyrere: Calibration of Flight Model CCDs for CoRot Mission
- high-accuracy photometer, using 4 CCDs, only light curves sent to ground
- measure fringing emperically with LED illumination
- about 90K e-/pxl for these frame-transfer CCD42-80, limited by non-settled phase
drives in the center of the CCD during frame transfer
- expect dark current to increase x20 over mission life due to radiation damage
- Q&A:
Q: Improve pixel capacity by slowing down parallel transfers?
A: Cannot make slower because additional transfer time cuts into integration time
for fixed frame rate.

Poster Pops:
- Reinhold Dorn: An Ultra-low Photon Background 1-to-5-micron Detector Mosaic Test
Facility
- built ultra-low-background test facility, no window
- can properly test dark current on IR detectors
- Fabrice Christen: Fast Conversion Factor (gain) Measure of a CCD Using Images
With a Vertical Gradient
- prefer method with two CCD images to that with single CCD image
- Fabrice Christen: CTE Measurement of a CCD Based on the Variance of the Signal in
Flat Field Images
- EPER agrees less than this technique, Fe55 agrees better with this technique
- Rob Phibrick: DC Characterization of CCD-based Detectors for Use in Multi-chip
Focal Plane Arrays
- fast DC measurements give much information to determine if packaging should
proceed
- measure every single input pin on every detector that comes in the door
- some vendor-recommended operating points non-optimal
- Peter Moore:Detector Testing Methodologies for Large Focal Planes
- want to involve detector testing right at the beginning of the detector
selection process for multi-detector focal planes
- detector testing becomes part of the critical process for buidling the new
instrument
- some CCD and CMOS measurements are not directly comparable
- Jose Acosto Pulido: On-the-sky Experience with the Hawaii-1 Detector at the
Camera/spectrograph LIRIS
- Jose Javier Dias Garcia: EMIR Detector Characterization
- Tim Hardy: Characterization of a Hawaii-1RG for an Under-sampled Guiding
Application
- Wanted to do intra-pixel response of Hawaii-2 detectors (under sampled system),
detector not avialable in time
- room-temperature tests done on bare mux before its demise at first cool-down

3. Nagaraja Bezawada: Performance Overview of VISTA IR Detectors:
- AR coating on two modules hygroscopic, unstable, lower QE than other 14 modules
- test setup not optimized for low dark-generation tests
- showing some QEs with higher than 100%
- Q&A:
Q: Conversion gain measured by photon statistics?
A: Yes. But detectors not back-filled with epoxy; pixels bigger than 18 microns.
Q: On cosmic rays, intersting to see variatoin between detectors. Contamination
in some packages?
A: Consistency of test results; might be a correlation between high-dark and
cosmic rays (?).

4. Don Hall: Ultra-low Background Characterization of Rockwell Scientific MBE
HgCdTe Arrays:
- KSPEC instrument converted into test facility
- scatter in dark-signal ramps due to read noise
- archiving test results on 1.5 TB external disks
- get consistent cosmic-ray rates on all detectors tested
- dark-current shot noise is insignificant contribution to total noise
- spikes in dark-current histograms due to software program
- good agreement between spatial averaging and temperal averaging techniques
- 1/2 to 1 e-/1000sec dark-current measurement limit
- digitizing temperature to 1 ADU/milliKelvin
- if can get below 0.06 e-/mK thermal drift tmeplate correction, might eliminate
need for thermal control
- Q&A:
Q: REf pixels factor of 2 lower noise than image pixels in 5-micron devices.
A: These are 2.5-micron detectors for space, not the same as ground-based
2.5-micron cut-off detectors.
Q: are these better than detectors released for ground work?
A: Have to ask Rockwell.
Q: Operating temp?
A: 37 Kelvin nominal.
Q: ref pixel reponse measured dependent on time
A: Ramp at 8 mK/minute, so about 1 mK/frame

Roundtable: CCD vs. CMOS

NOTE: the scribe could not keep up with the pace (flux) of this conversation,
leading to saturation and blooming with associated loss of information. This
roundtable discussion NEEDS TO BE REVIEWED for accuracy!

Gain problem with CMOS: difference between adjacent frames, see banding on small
spatial scale, take power spectrum, fundamentally loks kind of flat. Binned power
spectrum is not flat, less power at Nyquist, PSF has been convolved with something
that makes pixels correlated with one-another. This gets complicated: over-all
response of pixel-to-pixel in detector. What's a model for power spectrum? Don't
know. Inter-pixel capacitance is not same in the vertical axis as in the
horizontal axis. Fall off at high wave number; how to fit this thing? Maybe a
Gaussian? Not a great fit. Fit a parabola? 15% error between Gaussian and
parabola. Spatial variance of inter-pixel capacitance? Maybe falling off faster
than an exponential. Fit with quartic? If one does this, one finds a 22% error,
which is what Gert Finger finds as the capacitance error between his method and as
determined by photon statistics.

Maybe there are two contributions, one more significant than the other. Need to
consider the contribution from lateral diffusion.

A couple of interesting things are going on here. If one is going to use a photon
transfer curve, one can't apply Poisson statistics.

If one looks at inter-pixel crosstalk, about 1.7% - 2% with adjacent (horizontal or vertical) four pixels, diagonal four pixels about 0.07%; majority of effect may be
inter-pixel diffusion of charge.

What is column-to-column vs row-to-row capacitance differences?

Crosstalk is almost identical in horizontal axis versus vertical axis.

Is this device dependent?

When one changes bias volts, might change all of these inter-pixel coupling
capacitances.

Reach back to circuit theory: how does electric field vary in a 5-micron-thick
layer, over millimeter scales? Solve for quasi-infinite sheet of capacitors.
Roger Smith defers to John Tonry.

15% increase in crosstalk in horizontal versus vertical pixel axes, for both
Raython and Rockwell arrays, and for JWST detectors.

If one looks at crosstalk with nearest neighbor, about 6%; twice that amount when
all factored in. In voltage domain, need to calculate Voltage^2 domain.

If given intra-pixel asymmetry (in horizontal versus vertical axes), if one steps
the pixel layout in horizontal and vertical, this might explain the reported
difference of horizontal and vertical charge spreading.

A defective (hot) pixel may not be a good representative single-pixel test case for
charge spreading and/or diffusion; cosmic ray hits could be used as single-pixel
excitations. Try stepping controlled sub-pixel spot across pixel. To do this,
may need improved spot scans in IR.

5. Marco Sirianni, Mark Clampin, David Lumb: Overview of Future Space Missions &
Detector Needs:
- NICMOS cryotiger running well on HST
- NASA has planning panels, decadal panels, prioritize and review > roadmaps
- JWST is current higest priority
- Long lead time dual-path development has worked well in this case
- SAFIR wants operation near 4-degrees K
- will push the technical envelope
- LVO
- don't really have wide-field FUV, NUV detectors > probably won't use CCDs
- Development of Si MCPs for high count rates, less sensitive to contamination
- AlGaN for solar-blind UV detector
- Kepler specified with 42 CCDs, 2200x1024
- TPF-C: contrast ratio of 10^10 is a big technical challenge
- need photon counting with low power and cosmic-ray mitigation
- Spectrograph needs less than 1 e- RMS read noise
- For long observations, need a way of identifying cosmic ray events and removing
them
- ESA slant: serving community x2 of U.S., but budget 1/7 of U.S.
- National agencies have no solid agreements with ESA, so testing and calibration
time come under pressure
- See XMM charge transfer inefficiency (CTI) step changes from solar flares
- had one incident/year/camera from micrometeorites
- Integral Germanium detectors, suspected loss of diodes through thermal cycling?
- Mars Express should have been more careful about testing and calibration before
launch
- GAIA: on-orbit cosmic-ray discrimination
- BepiColombo: solar-electric propulsion, years of flight time
- X-ray detection being spun off into medical imaging
- Darwin: possible problems with vibrations from mechanical coolers?
- No Q&A

6. Marco Sirianni: Radiation Damage in HST Detectors:
- annealing rate doesn't seem to depend on how long you anneal the CCDs (at +20 C)
- Q&A:
Q: Using a direct measurement of magnitude to measure CTE: where reference?
A: Proceedings will contain information.
Q: Effect of regular annealings on CTI?
A: Yes,for both cases.
Q: Why annealing at cretain temperature? Threshhold temperature where annealing
stops working?
A: Perhaps -40 C.

7. Qian Song: An EUV IMaging Detector of Space Solar Telescope:
- If funded, will be first space astronomy mission for China
- EUV's phosphor response tested with photodiode
- Q&A:
Q: How is the flight detector to be cooled?
A: No cooling in laboratory test system; radiative cooling is planned for the
mission.
Q: Have considered direct-electron bombardment of CCD rather than using a
phosphor?
A: No / not yet, this is not avialable in the laboratory system.

Poster Pops:
- Jean-Tristan Buey: The Camera of the CoRoT Space Mission
- Bob Hill: Radiation-induced Emission from the CdZnTe Substrate of the WFC3 IR
Detectors
- Very little increase in dark current, below beginning of life spec; persistence
goes away in 5 minutes
- Beam in original tests 63 MeV, new testing at 14 MeV
- Going to detector without CsZnTe substrate
- Massimo Robberto: The Second Generation of IR Detectors for WFC3

Discussion: The Challengers of Detectors for Space Missions

At one unnamed observatory, did one custom detector run, got one detector that was science grade -- the cost per (the one) detector quite high. This was this
observatory's introduction to CCDs. Employee was fired from non-technical job, and
took a ballpeen hammer to the window of (cold) camera head. The camera-head
window was breached and shattered, and the detector suffered damage from imploding window shards. A great disappointment.

Q: In overview talk, chicken-and-egg problem of new technologies for space. Who
does the (ESA?) space agency look to (defense agencies?) to develop new
technologies?
A: Flyable technology already developed now would apply to missions 2014-2017.
Have to dream up "technology reference missions" to obtain funding for
technology advances. Anticipating the direction of present technology
paths risks a future disconnect with the user community.

NASA has its own set of technology readiness levels; easier with long lead times.
Example: want photon-counting for planet detection, can't fund this by itself
without attaching to someone else's project (funding). The answer is, don't have
to rely on DOD, just easier on bigger programs than smaller programs to do
technology developments.

Regarding radiation: what kind of particles are of interst? Heavy ions? Where can
such tests be conducted?

XMM-Chandra, damage from soft protons: deposit all their energy in the buried
channel, several times more damaging than ionizing paritcles. With higher
energy ionizing particles, damage is more localized than the damage from lower-
energy particles; tends to be worse. Heavy ion testing facility in Netherlands,
and there are facilities in U.S.

Most of testing for missions to L2 are for protons; have fairly significant shielding. Too much shielding promotes secondary particles. Not a lot of
agreement on L2 environment. Depends on kind of science being done.

L2 environment is typically a halo orbit, circle (diameter) is 1 million Km, very
different environments crossed during the halo orbit: in and out of the
magnetosphere, changes of x5 in particle energies and densities at different points
in the halo orbit.

SDW2005 Day 5 2005-Jun-24 FHH

0. Raccomandazione = an inside connection (?); you need a Raccomandazione for
**everything** in Italy

1. Peter Moore: MONSOON Image Acquisition System: Raindrops Keep Falling on our
Heads
- Built to be able to accomodate a wide range of focal planes
- 18-bit dynamic range, saving 16-bit or 18-bit data
- Layers: linked by defined sets of protocols (interface control documents), should
be able to get a large amount of reuse:
- Detectors
- Hardware
- Firmware
- Software
- First pass: getting performance out of the CDS took longer than anticipated
- Each collaboration effort requires pretty explicit support
- Front-end of Acqisition boards redesigned for ODI, cutting 4 amplifiers (per
channel?) out to up density and lower power dissipation
- Q&A:
Q: Do arrays of CCDs need to be in lockstep, or can non-great-circle scans with
different clodk rates in different columns of detectors be accomodated?
A: Differing TDI rates within an array are possible.
Q: Factory for hardware production?
A: NOAO is only institution producing hardware. Currently, NOAO cannot sell
systems, but can help others build their own.
Q: Can others supply the manufacturing need? Mass produce the standard pieces?
A: You take your own ownership of MJONSOON.
Q: Missing piece is the debugging and quality control. Can be supplied outside
of NOAO?
A: Yes (from audience).
A: NOAO has added testing proceedures that dont require connecting a detector.

2. Zhaowang Zhao: Astronomical Array Control & Acquisition System at NAOC:
- connectoin via RJ45 implies no built-in fiber interface
- EP1C20FC400 Nios Processor
- Support 8 vertical clocks, 8 horizontal clocks
- Dual-channel preamp board
- Some CCDs read with 2 simultaneous output amplifiers
- Q&A:
Q: Time resolution for TDI line interval?
A: 20 MHz divided by 40 or 50 states/pixel.
Q: Concern for lightning with copper-connected system?
A: Maybe; might switch to fiber in the future.

3. Manfred Meyer: ESO's Next Generation Controller
- What is shown in this talk is a prototype system.
- 8/10-bit-encoded serial bus
- (Xilinx) Virtex-IV chip supports 10Gbit/sec links
- Virtex chips require no glue logic to the PCI bus
- PCI master does DMA, PCI slave does control, are independent
- Front end prototype is complete 4-channel system
- 14-bit DACs for clock biases
- Telemetry allows monitoring currents to the detector
- 16-bit Analog Devices ADCs, pin compatible with 18-bit ADCs
- Q&A:
Q: Can have several boards in one PC, or several PCs. Provision for
synchronization of multiple channels?
A: Yes, to within the 10-ns clock.
Q: What ADCs?
A: Analog Devices.
Q: Xilinx IP cost?
A: $10K

Poster Pops:
- Bruce Atwood: Zero (almost) Noise Amplifier
- Marc Baril: HIAc: A High-speed Readout Controller for Avalanche-gain CCDs
- for KBO occultations
- fits in PC104 form factor
- can imbed the host inside the camera
- Fabio Bortoletto: The Antarctic NIR/MIR Camera
- telescope placed directly on the ice, which is 2 Km thick
- Fabio Bortoletto: A New Generation of Data and Control Interfaces for Digital
Detectors
- Javier Reyes: NGC Front End for CCDs and AO Applications
- need to develop high-speed high-voltage clock board
- Claudio Cumani: Softwar for the New Generation Detector Controller
- want to use same software in lab and at telesocpe
- Jose Javier Diaz Garcia: EMIR Detector Data Acquisition Electronics
- 2 different fanout board designs, one 4-channel for test, other 32-channel
- Using Leach controller with modificationos
- No cabling between SDSU chassis and cryostat
- Maureen Ellis: Readout Electronics for Detectos at 100 mK
- simultaneous 450-microns and 850-microns sampling on the sky
- hybridized detector layer on SQUID mux layer
- Jean-Luc Gach: A Dedicated Controller for Adaptive Optics L3CCD Developments
- Fernando Gago-Rodriguez: OPA350 Operational Amplifier Charactererization at
Cryogenic Temperatures
- Observed slight increase of 1/F noise when cold
- OPA350 works OK at cryogenic temperatures, using for 3 years
- Cristoph Geimer: Keep Control: PULPO II -- ESO's New Housekeeping Unit
- PULPO 1 single alarm output did not discriminate between thermal alarm and
vacuum alarm, more alarms in this generation
- Jose Gigante: Improved Control Electronics for OSIRIS-GTC Commercial Tunable
Filters
- Enrique Joven: Current Status of the OSIRIS-GC Instrument Control Systems
- modified classical Leach controller
- Leander Mehrgan: 256-channel Data Acquisition System for VISTA Focal Plane to
Read Out Sixteen 2Kx2K VIRGO Detectors
- Gustavo Rahmer: MONSOON Image Acquisition System Configuration Management
- Roland Reiss: We Must be MAD: Pushing IERA to its Limits
- Joseph Tufts: McDonald Obsrevatory nRG Electronics / Waveforms, Reference Pixels
and Other Fun
- cold camera mounting to an existing warm spectrograph
- one spider vane for detector mount thicker than others to pass the cold path

4. Armin Karcher: Integrated Signal Processing and A/D Conversion in One Focal-
Plane-Mounted ASIC:
- 3.3-Volt process
- can't get 16-bit dynmamic range into a 3.3-Volt process
- if integrator signal level reaches a threshhold, switch in a parallel capacitor
to stay within ASIC's dynamic range
- CLIC chip (clock drives, biases, sequencing) on Si-on-insulator process
- Q&A:
Q: What charge-injection-induced offset voltage seen when switching integration
capacitors?
A: Charge injection via back-to-back switches cancels to first order.
Q: Compromise in design from large input capacitor on preamp input?
A: Went with standard process, ESD protection on inputs have relatively large
capacitance (1 pF) so use input cap large compared to this.
Q: Allow capacitorss to switch during input integration?
A: Cannot switch back to the higher gain. Only set DC op point during reset
time.
Q: What size die for signal-chain ASIC?
A: Maybe 3.8 by 5 mm ? Clock chip will be larger.

5. Markus Loose: SIDECAR ASIC: Control Electronics on a Chip:
- 8 gain steps in preamp (3 dB intervals)
- external bandgap reference needed when going down to 30 K operating temperature
- 16-bit ADC is successive-approximation design, non-trimmed
- A little bit higher noise in ADC in trade for lower power; will run higher
power next for lower noise
- spatial and temporal methods agree closely on system noise, also with Fowler 8-8
sampling
- some correlated noise sources at low frequencies
- radiation tolerance harder than 80 K rads
- Q&A:
Q: Will be required to buy Rockwell detector if get an ASIC?
A: Sell separately.
Q: Cost?
A: Depends, refer to program managers.
Q: Total range of pixel rates?
A: Limited by ADCs, 500KHz with 16-bit, 10 MHz with 12-bit.
Q: Lower end of pixel rate?
A: If operated cryogeneically, then leakage not importnt, not lower limit that
makes sense.
Q: Radiation testing with heavy ions?
A: Protons used so far; NASA has asked about heavy ions, but they will damage
the detector, and costly tests.
Q: Worry about single-event upset (SEU) with heavy ions.
A: Depenmds on how much charge is generated. Think can verify good protection
with protons. Memory can recover from SEUs. Also Hamming codes to catch
SEUs.
Q: On analog input to preamps, what
A: Limited by 3.3-Volt power-supply limit due to protection diodes
Q: Bandwidth-limiting flexibility?
A: 60 KHz on low end, can turn off compleltly on high end
Q: Average of fast reads not reducing noise?
A: Noise only went up by a few percent. ADC had to be biased at a slightly
higher bandwidth for lower rate, so doesn't have to be opened up in bandwidth
when going to doubled speed. 30% improvement in noise.
Q: Testing with full 32 channels at 10 MHz?
A: Haven't tested it yet, have some doubts about performance to be seen.
Q: Drift in bias voltages with long exposures?
A: Tested with 1000-second exposures.

Roundtable: Electronics -- Present Day Challenges and the Longer Term Future --
Where is This Going?
- (Beginning of roundtable missing...)
- ASICs: great investment to get started
- can we leverage an ASIC to the greater community, or does it require users to be
collaborators
- What is a break-even number of devices?
- 20 devices; not such a large number.
- What is the price range?
- $50K - $80K for manufacture. Much NRE done, little to add if minor mods made.
Probably less than $100K total.
- Package?
- If FPGAs you have to go to BGAs.
- Issue is, if you want to remove a chip from a valuable board, then you have a
problem. $20K rework machines required, and are these workshops ESD safe?
- If you are going to a house for surface mount, have them mount the BGA.
- If someone was interested in a (? SIDECAR ?) (? other ?) ASIC right now, is it
available?
- Not being management, can't officially say, but probably can be sold.
- Right now, input stage is optimized for P-channel CCDs.
- Why does this matter after the AC-coupling capacitor?
- Now you want to move the DC level of the input voltage, maybe could do this
externally.
- If we had an input stage that was truly differential, then could ground the
appropriate side.
- Hard with 3.3-Volt supply range to do the true differential.
- In the next turn of the ASIC (in th next six months) will have the sequencer.
Will be able to load the timing sequencer, give it start/stop commands.
- Need a way to change the gain of the input stage.
- Next version will probably have a selectable input stage.
- LBL can license the design to a commercial company. Typically the way things
work, establish a collaboration.
- The thing with an ASIC is they can be made again.
- You have to be careful with that. The 0.25-micron technology may have a 10-15
year window of manufacturability.
- These lifetime issues are a concern with ASICs.
- For ground based, a few CCDs, using ASICs is still a ways off.
- The P-channel CCD process has just been licensed to Fairchild. What about
licensing the whole ASIC design?
- ASIC foundry runs are more like $100K.
- Some here could not use this particular ASIC.
- For $100K, how many parts do you get?
- 30 to 50, unpackaged.
- The Kodak sequencer (timing generator) part has full functionality for $25,
quantity 1.
- What is the cost of buying a 4-channel CCD controller? That is the relevant
price comparison.
- SDSU 2-channels is $20K. Maybe not the reference point.
- System cost for using these ASICs may go to $5K - $10K.
- Should we be trying to pool resources and build the same controller?
- Each group is doing their own controller. National observatories have the
resources.
- Bob Leach is doing it without much competition.
- Want reusable technology to minimize costs. To use part of MONSOON pieces
probably needs 10 people.
- If we can get CRIC, CLIC, KSC1000, get a fair bit of MONSOON if it talks the
software layer.
- IF you look around and no controller will do your job, then you have to commit
you and your people to make the resources. If you invest time in what may be
not quite right, then you are not working on what is just right for you.
- How is it quick to start fresh in terms of effort and timescale?
- If you start to build a high-speed controller, what you know about low-speed
controllers helps little. for a gen purpose system, you need to add complexity.
- How much documentation for MONSOON?
- 18 x 8 pages.
- How complicated your software gets depends on your hardware.
- Many of these controllers from bigger institutions can not be purchased.
- Peter Onaka does want to sell his controller. Have an agreement from the
university.
- Advantage of serial links is easy paralleling.
- This was designed with the knowledge of need for high bandwidth.
- Choosing the Virtex chips is a big advantage.
- Is there a chance to interface an ASIC to the new controllers?
- How much of the $100K is NRE?
- Process cost is sort of fixed, depending on the process picked. A wafer lot is a
lot of chips.
- The reason we are rollng our own is that ASICs are not mature and some must get
on the sky sooner. Onaka's controller could be junked sooner rahter than later
if the right ASICs were/are available.
- From the ASIC point of view: how many other people would control enough funds
to join in?
- To get the snowball effect will only occur if there is a critical mass to get it
going.
- 35 IC buyers is not hard to come up with.
- Why the fab run so expensive?
- CMOS is more expensive per lot run, many more layers than used to with CCDs.
- In pictures of the CMOS, multiple devices will need multiple reticles.
- A reticle can do more than one type of CCD simultaneously.
- Wafers runs are 25 wafers/run. Son't know if these ASICs were done on 6-inch
or 8-inch wafers.
- Will Roger Smith be the one to coallesce the snowball?
- What flexibility in icking speeds, bandwidths?
- What is the bandwidth of the signal chain?
- May have to fiddle with the input stage to pick the bandwidth.
- The typical big CCD is 8M pixels with 2 outputs.
- Stamping feet for more CD outputs does no good without money to back the
request.
- 3.3-Volt chips require interface buffers, for bias and clock drivers.
- The ASIC cannot be compared to the controller, and reuiqres integratoin. The
Leach controller starts to liik like a good deal.
- Detectors and these ASICs may not come down in price, once the initial price is
set.
- Support is the thing that nails the companies (such as Rockwell). A customer
buying one costs in support the same as the customer who buys 50.
- What is the positive advantage to (???'s) board versus someone elses?
- The appeal to the ASIC is a turnkey solution (if it is a turnkey solution).