SuperNova Acceleration Probe Research and Development Efforts

1
SuperNova Acceleration ProbeResearch and
Development Efforts

• Chris Bebek
• UC Berkeley Lawrence Berkeley National Laboratory

2
Instrument RD areas

• In the past year we have conducted a technical
  and scientific trade study covering a range of
  options for the SNAP instrumentation suite.
• We have arrived at a coherent instrument working
  concept and observation strategy constrained by
  reliability, satellite, thermal, orbit and
  telemetry issues described Mike Lampton and SNe
  characteristics that optimizes the science reach
  of SNAP.
• We have identified four risk areas and have an
  RD program to address these
• CCDs
• MTF/PSF
• Cold irradiation
• Fabrication
• HgCdTe
• QE
• Readnoise
• Fabrication
• CCD readout
• HgCdTe readout

3
How Science-Driven Requirements map onto
Instrument Concept

• Measurement Program
• 50 Type Ia SNe per 0.03 in z from z0.3 to 1.7
  (2500 total).
• Follow up spectroscopy near peak luminosity.
• Template building (spectra vs époque for subset
  of zlt0.7 SNe).
• Photometry
• U, B, V, (R)-band light curves.
• B-band to 1 at peak.
• BV color evolution.
• Malmquist bias.
• Rise time.
• Peak to tail ratio.
• Spectroscopy
• UV metalicity features strength and location.
• S and Si features
• SII 5350Å line, Dw 200Å
• SII W shape, Dw 75Å

• Instrument
• Use two plate scales to cover the wavelength
  range to obtain time efficient photometry and a
  large FOV.
• Observation cadence commensurate with SNe
  evolution.
• Allocation of time for photometry and follow up
  spectroscopy.
• Imager
• Wavelength coverage from 350 nm to 1700 nm.
• 9 filters.
• Cadence of repetitive exposures matched to SNe
  evolution rate.
• Exposure lengths generate required S/N versus
  magnitude.
• Spectrograph
• Wavelength coverage from 350 nm to 1700 nm.
• S/N 20
• Resolution 140 (l/dl)

4
Photometry illustration
1z spaced B-band filters
5
Spectroscopy illustration
Metalicity
SII W
SiII
6
Instrument working concept
Guiders
Cold plate
Radiator
Cables
Thermal links
Shield
Spectrograph
CCDs/ HgCdTe
Shutter
FE Electronics
Filters
Data/Monitoring/ Command
7
Focal plane concept

• Coalesce all sensors at one focal plane.
• Imager sensors on the front.
• 36 HgCdTe 2kx2k 18 mm
• 36 CCD 3.5kx3.5k 10.5 mm
• Filters
• 1 of 3 per HgCdTe
• 4 of 6 per CCD
• Spectrograph on the back with access ports
  through the focal plane.
• Common 140K operating temperature.
• Dedicated CCDs for guiding from the focal plane.
• Exposure times of 200 s with four/eight exposures
  in CCDs/HgCdTe.
• 20 s readout slow enough for CCD noise and 4 post
  exposure and 4 pre exposure reads of HgCdTe.

rin6.0 mrad rout13.0 mrad rin129.120
mm rout283.564 mm
8
Integral Field Unit Spectrograph Design
Small IFU, 2-arm spectrograph based on image
slicer work performed for NGST.
Image slicer
9
Example scan
10
Establishing detector requirements

• Some sensor requirements are determined by SNe
  characteristics, eg, wavelength coverage.
• Some characteristics are innate to the sensors,
  eg, QE(l).
• Other requirements (eg, read noise, dark current)
  have been bounded by studying the desired S/N as
  a function of SNe epoch, exposure time, and
  number of exposures.

• Example study of achieved S/N with four 200 s
  exposures and
• CCD
• RN 4 e-
• DC 0.02 e-/s/pix
• Npix 4
• HgCdTe
• RN 4 e-
• DC 0.02 e-/s/pix
• Npix 4

Requirement
S/Ngt30
Dmag
S/Ngt30
S/Ngt20
S/Ngt15
S/Ngt10
S/Ngt3
11
Derived Requirements for the Imager
Note Requirements for spectrograph use are
similar.
12
LBNL CCD work
13
LBNL CCD technology
Back-illuminated thick CCD on a high-resistivity
n-type substrate, operated fully depleted.
Advantages 1) Conventional CMOS processes
without super thinning. 2) Full quantum
efficiency to gt1 mm gt no fringing. 3) Good
blue response with suitably designed rear
contact. 4) No field-free regions for charge
diffusion, good PSF. Drawbacks 1) Enhanced
sensitivity to radiation (x-rays, cosmic
rays, radioactive decay). 2) More volume for
dark current generation. 3) Dislocation
generation.
The technology has been transferred to DALSA and
rad hard, p-channel, high resistivity
capability is now listed on their web site.
14
LBNL CCD evolution
100 mm LBNL-fabbed 2kx2k
100 mm DALSA 2kx2k
100 mm LBNL-fabbed 2kx4k
150 mm DALSA PIN diodes
150 mm DALSA 2kx2k 3kx3k
15
Near-IR vs Visible imaging
WIYN 3.5m with LBNL 2048 x 2048 CCD (Dumbbell
Nebula, NGC 6853) Blue H-? at 656 nm Green
SIII at 955 nm Red 1.02 mm
16
LBNL 2k x 2k results
Image 200 x 200 15 ?m LBNL CCD in Lick Nickel
1m. Spectrum 800 x 1980 15 ?m LBNL CCD in NOAO
KPNO spectrograph. Instrument at NOAO KPNO 2nd
semester 2001 (http//www.noao.edu)
17
CCD issues

• Pixel size
• Well depth
• Linearity
• Dark current
• Persistence
• Read noise
• MOSFET operation
• Charge transfer efficiency
• Quantum efficiency
• Diffusion
• Intrapixel response
• Radiation
• Proton damage
• 60Co
• Damage when cold
• Fabrication

18
Performance

• Pixel size
• Well depth
• Linearity
• Dark current
• Persistence
• Read noise
• MOSFET operation
• Charge transfer efficiency
• Quantum efficiency
• Diffusion
• Intrapixel response
• Radiation
• Proton damage
• 60Co
• Damage when cold
• Fabrication

• 10.5 ?m work.
• 130 ke for 10.5 ?m pixel.
• Better than 1.
• 2-5 e/hr/pixel.
• Erase mechanism is effective.
• 2.0-2.5 e.
• Documented at operating temperature.
• CTI 10-6 pre-irradiation.
• Extended red performance realized.
• On-going study.
• On-going study.
• More robust than existing space devices when
  damaged warm.
• No surprises for T300K dosing.(wont discuss
  today).
• An activity during the next 3 months.
• Partially commercialized.

RD areas
19
LBNL 2k x 2k Quantum Efficiency
20
MTF/PSF issues

• We are 2x undersampled.
• Uniformity and simplicity of PSF determines the
  amount of image dithering that will be required.
• Diffusion the fundamental gaussian spread of
  charge as it drifts from the photon conversion
  site to a pixel.
• For conventional CCDs, the rms spread is the
  depletion depth, 10 mm.
• For LBNL CCDs, the rms spread is determined by
  thickness and the sq.rt. of the depletion
  voltage.
• We require 4 mm. E.g., t 200 mm, Vsub 100V, s
  3.2 mm.
• We are working on routine thinning to 200 mm and
  beginning the study of device robustness versus
  Vsub (we have routinely operated in the lab at
  60V with excursion to 140V).
• Intra-pixel response after diffusion drift,
  does the charge hit the correct pixel.
• 2D modeling of conventional CCDs and LBNL CCDs
  with its substrate voltage shows good termination
  of the field lines on pixels.
• Measurements we are commissioning a pinhole
  projector to measure diffusion as a function of
  voltage and thickness and to map intrapixel
  response.

21
Diffusion/intrapixel response measurements
We are commissioning a pinhole projector, 4 ?m
FWZ, to scan the backside of the CCD looking for
charge collection variations in vicinity of pixel
edges. We have scanned the front side so far and
see the polysilicon gate structure. Backside
scans any minute now.
Front-side scans
22
Proton radiation damage
23
Radiation environment

• Integrated for three years, SNAP will be exposed
  to
• A few krad (Si) TID.
• A few ?107 MeV/g NIEL.
• Note
• 1x109 protons/cm2 _at_ 12 MeV is 1.5x107 MeV/g NIEL.
• 1x109 protons/cm2 _at_ 12 MeV is 500 rad.

24
Proton irradiation studies

• We used 12 MeV protons at the LBNL 88 Cyclotron
• Two set of four device were irradiated at room
  temperature.
• Doses were 5x109, 1x1010, 5x1010 and 1x1011
  p/cm2.
• We characterized the devices by measuring their
  CTE and dark current as a function of
  temperature.

25
Dark Current Degradation
Dark current is measured with one thousand or
more second exposures. The gaussian charge
distribution in the active region of the CCD is
compared with the gaussian change distribution in
the overscan region.
SNAP
Fit gives expected Si bandgap, so no new dark
current sources are developing. The plateau at
right is not identified yet, but could be surface
leakage currents.
26
Charge transfer efficiency
CTE is measured using the 55Fe X-ray method at
128 K. The readout speed is 30 kHz, the X-ray
density is 0.015/pixel. Degradation is about
1?10-13 g/MeV.
Comparison to conventional CCDs after converting
dose to NIEL (MeV/g).
SNAP
Caveat. We irradiated parts at 300K and
unpowered. While we have compared apples with
apples, our study will be complete only after
performing damage at operating temperature and
powered. Will try to complete this this summer.
1L.Cawley, C.Hanley, WFC3 Detector
Characterization Report 1 CCD44 Radiation Test
Results, Space Telescope Science Institute
Instrument Science Report WFC3 2000-05,
Oct.2000 2 T. Hardy, R. Murowinski, M.J. Deen,
Charge transfer efficiency in proton damaged
CCDs, IEEE Trans. Nucl. Sci., 45(2), pp.
154-163, April 1998
SNAP
27
CTE vs Temperature at 1x1011 p/cm2
Both serial and parallel CTE exhibit significant
temperature dependence due to interactions with
radiation induced trapping centers.
28
Hole Traps Found in n-Type Si
Trap parameters measured using DLTS
V
V V
VV
Proton Irradiation
CiOi
Sii
Ci
CiCs
29
Fitted trap density versus dose
30
CCD fabrication
31
100 mm wafer fabrication

• LBNL manufactured
• We have fabricated 10.5, 12, and 15 mm devices in
  a variety of formats up to 2kx4k.
• These have ranged from 190 to 300 mm thick.
• Some of these devices are deployed in ground
  telescope.
• Recently, much effort has gone into developing
  careful handling procedures and equipment
  modifications to protect the backside of the
  wafer during manufacture.
• DALSA manufactured
• Our process technology transfer first done here.
• 15 mm devices up to 2kx2k have been successfully
  built.
• Devices as thin as 200 mm have been finished.

32
150 mm wafer fabrication

• DALSA work
• They have converted exclusively to 150 mm wafers.
  These wafers are must be thinned from 675 mm to
  200 mm for our use.
• Unthinned photodiode wafers have been fabricated
  with good results.
• A few thinned wafers have been fabricated. We
  found similar backside damages areas that we
  have already eliminated at LBNL.
• Unthinned CCD wafers have been fabricated that
  were of high quality in front illuminated
  studies.
• We have received one thinned (300 mm) CCD wafer
  that is now under backside illumination tests.
• LBNL work
• We are gradually transferring our backside
  handling knowledge to DALSA but expect this to
  take some time to fully implement. We view this
  as the second phase of our commercialization
  effort.
• In the meantime, we have acquired the one piece
  of 150 mm processing equipment that will allow us
  to perform the last steps of wafer processing
  contacts, metalization, AR coating.
• DALSA will provide 675 mm thick CCDs where the
  front side is complete, device is thinned, and
  backside thin poly is deposited, This includes
  all the conventional CMOS process steps. We will
  continue to work with them on thinning issues
  with fully automated processing equipment.

33
Example of backside damage/particles before
remedies
34
LBNL efforts on backside particles/damage

• Backside processing issues and remedies to be
  transferred to DALSA.
• Back side scratches through ISDP layer fatal for
  fully-depleted operation
• Avoid where possible handlers made from materials
  that can scratch silicon
• Improved wafer carriers for MRC sputtering system
• Manual override of wafer alignment arm on MTI
  resist dispense arm
• Use of sacrificial SiO2 layer on wafer backside
• Not scratch immune but allows for undercut of
  particles during strip
• Particle removal via wafer scrubbing (most
  effective technique to date)
• Use of wear resistant materials on vacuum chucks
  and wafer handlers where possible (DuPont VESPEL
  effective but does shed particles)
• Avoid use of silicone parts (cannot remove with
  scrubbing)
• First wafers through equipment (coater, aligner)
  tend to have significantly higher particle counts
• Photoresist aerosol particles too large to be
  removed with ashing, require addition of solvent
  to scrubbing soap solution
• LBNL experience particles can be removed with
  mechanical action (scrubbing). Main concern is
  damage through thin backside poly layer.

35
HgCdTe
36
Rockwell HgCdTe

• Rockwell HgCdTe devices are our only option at
  the moment.
• WFC3 MBE material with 1.7 mm cutoff is a perfect
  match to SNAP.
• NGST 2k x 2k format being developed is also a
  good match.
• Status (as of March)
• The dark current is OK.
• There is a QE problem in the 900 nm to 1100 nm
  region.
• There is a large read noise, 30 e, not the
  design goal of 10 e.
• Long-term drifts and settling times are seen at
  some test sites.
• Rockwell claims they understand the MBE knobs
  that control QE.
• They are have grown new material (1k x 1k).
• It is presently being bumped and packaged.
• The large read noise is bad for SNAP
• We want to rt-N this down to 5 e, ie, four CDS
  reads take 10 e to 5 e.
• More reads have a big impact on observation time
  budget.
• Rt-N has only got to 17 e so far (there may be
  new info on this).

37
MTF/PSF issues

• We have studied the impact of a gutter around
  each pixel as existed in the PACE devices (this
  is NOT present in the MBE devices). The impact of
  that dead region relative to a device without it
  was to double the number of exposures required to
  obtain equivalent photometry.
• Intrapixel response for the MBE HgCdTe has not
  been measured yet.
• We will acquire a device to measure this
  ourselves. We have ordered a mux to begin setting
  up a measurement system with pin hole projector.
  
• Intrapixel response may be just fine as it is or
  it may not be. If not, Rockwell has posited
• Design changes of the implants near the PN
  junctions. This is essentially tuning up the
  electric fields to better capture the charge.
• Etching microlenses into the CdZnTe substrate to
  focus photons on the pixel sweet spot.

38
CCD readout
39
CCD support electronics

• Goals
• Photons-to-bits focal plane
• Eliminate large cable plant to reduce system
  noise problems.
• Reduce power dramatically relative to
  conventional implementation.
• ASIC Challenges
• Large voltages
• 10 V clock swings
• 20 V MOSFET biases
• 32 V span within CCD, excluding depletion
  voltage
• Large dynamic range from 2 e- readnoise and 130
  ke- well depth.
• Radiation tolerance (borrowed from GLAST for now)
• Total ionizing radiation dose performance
  maintained up to 10 Krad (Si).
• Single-Effect Latch-up (SEL) immune to a minimum
  LET of 40-80 MeV-cm2/mg.
• Singe-Effect Upset (SEU) performance maintained
  for a LET of at least 8 MeV-cm2/mg.
• Operation at 140K to reduce cable plant and
  associated problems requires low power

40
Readout Electronics Concept

• CDS Correlated Double Samples is used for
  readout of the CCDs to achieve the required
  readout noise.
• ADC 16-bit dr, 12-bit res 100 kHz equivalent
  conversion rate per CCD.
• Sequencer Clock pattern generator supporting
• modes of operation erase, expose, readout, idle.
• Clock drivers Programmable amplitudes.
  Supports 4-corner or 2-corner readout.
• Bias and power generation Provide switched,
• programmable large voltages for CCD and local
  power.
• Temperature monitoring Local and remote.
• DAQ and instrument control interface Path to
  data buffer memory, master timing, and
  configuration and control.

41
ASIC roadmap

• We are working with LBNL ASIC designers to
  address CCD clock generation, bias voltage
  generation, and analog signal processing in one
  or more ICs.
• Correlated double sampler
• Starting here since it has the most challenging
  analog issues.
• We have performed a survey of sub-micron CMOS
  processes.
• We have evaluated system noise for different
  technologies and signal processing schemes.
• We have measured pre and post irradiated test
  structures as a function of T.
• We will design a CDS circuit for fabrication over
  the next 4 months.
• ADC
• This could be part of CDS circuit, so we are
  thinking of implementations in parallel with CDS
  development.
• We are exploring a 12-bit pipeline ADC with three
  ranges.
• Clock drivers
• Pattern generator is an easy digital design.
• Amplitude control of large voltages will be
  challenging.
• Study of rad tolerance of 40 V sub-micron CMOS.

42
CCD MOSFET noise
Measured noise spectral density at low
temperature for an LBNL CCD MOSFET.
43
Noise comparison (PMOS)
PMOS noise spectral density for several vendors
derived from their technology models. Model
results have been validated by test data from
others.
44
Differential averager
CCD noise first stage noise

• Pmos Agilent 0.5mmIC 1, Id 100mA
• Nmos TSMC 0.25mm
• IC0.1, Id 100mA

C (100pF)
R
Dt
t
Out
X1
CCDnoisesource
(Noiseless)
t
-X1
Integrator voltage gain 2
Integration time t
Conversion gain 3.5mV/e
Input referred noise (e)
Dt 0
4ms (R20KW)
(1) 2.8 / (2) 2.9
t
t
8ms (R40KW)
(1) 2.1 / (2) 2.2
10ms (R50KW)
(1) 1.92 / (2) 2
The 1/f noise of the input stage is reduced by
the CDS The thermal noise from the input stage is
negligible compared to the CCD
45
TSMC 0.25 mm cold CMOS
Threshold voltage PMOS
Threshold voltage NMOS
1mV/K
1mV/K
Mobility ratio PMOS
Mobility ratio NMOS
After rad means gt10 Mrad.
UTE -1.3
UTE -0.7
46
Sub-micron CMOS comments

• Sub-micron CMOS appears to perform well down to
  100K.
• Vendor BSIM3 SPICE model predicts performance
  down to 150K region.
• Technology is extremely rad hard.
• LBNL has the rad hard by design methodology to
  build robust systems.
• As with CCDs, rad testing at cold temperature
  needs to be explored.
• Sub-micron 40V processes need careful radiation
  study because of the thick oxide used. Our
  default plan is to use external JFETs as voltage
  boosters for clock and bias drivers.

47
HgCdTe readout
48
HgCdTe readout

• Inputs to readout architecture
• Advertised single CDS noise is to be 10 e we
  require 5 e.
• Therefore, we need four pre and four post
  exposure reads.
• To accomplish 8 reads in 20 s at 100 kpixel/s
  rate requires 16 taps per 2kx2k device. H2-RG has
  32 taps.
• Desirable to do pre and post read averaging in
  hardware.
• There is a Rockwell initiative for the NGST mux
  readout for an ASIC operated cold adjacent to the
  sensor
• Five 16-bit ADCs (intended for 4 tap readout)
• Microprocessor based timing sequencer
• Data processing
• Can implement co-adding and averaging at the
  pixel and line level.
• Power including mux is 2 mW per read port.
• RD issues
• Trade study of using existing H2-RG mux or
  developing one with 16 taps.
• Ability to cascade multiple Rockwell ASICs to
  achieve more ADCs per H2-RG.
• Development of our own ASIC.

49
RD Summary

• CCD
• Study operation at high depletion voltage to
  minimize diffusion.
• Measure intrapixel response.
• Radiation damage at 140K.
• Refine fabrication process.
• Establish production yield.
• HgCdTe
• Read noise needs to be reduced.
• Track QE developments.
• Establish production yield.
• CCD electronics
• Radiation measurements at 140K.
• Fabricate demonstration CDS/ADC in 0.25 mm CMOS
  during the next year.
• Radiation study of 40V sub-micron CMOS.
• HgCdTe electronics
• Refine commercial solution, if it exists, to SNAP
  needs.
• Or, develop our own readout.