Spectrograph Detector

1
Spectrograph Detector Hardware
2
Spectrograph Detector
3
Science Requirements

• 2048x2048 frame size
• 1.0-2.5µm wavelength response
• Low dark current lt 0.01e-/sec/pixel
• Low read noise lt 5e-
• Well Depth gt 50,000

4
Spectrograph Detector

• Rockwell HAWAII-2 2048x2048 HgCdTe
• 1.0-2.5µm wavelength cutoff
• Four fully independent 1024x1024 Quadrants
• 18 µm pixel size
• Well depth 100,000e-
• Source follower per Detector (SFD) Input circuit
• Operating temperature 70-80K

5
HAWAII-2

• Rockwell HAWAII-2 2048x2048 HgCdTe
• 1.0-2.5µm wavelength cutoff
• Four fully independent 1024x1024 quadrants
• 18 µm pixel size
• Well depth 100,000e-
• Source Follower per Detector (SFD) input circuit
• Operating temperature 70-80K

6
Focal Plane Array

• Hybrid assembly of HgCdTe detector via indium
  interconnects to CMOS mux.

7
(No Transcript)
8
Two Choices of Technology

• Differ in detector substrate material - lattice
  match HgCdTe detector material
• Choices
• PACE uses Liquid Phase Epitaxy (LPE) to deposit
  detector HgCdTe on sapphire, 2.5µm
• CdZnTe substrate uses Molecular Beam Epitaxy
  (MBE) to deposit detector HgCdTe, 5.0µm

9
PACE Substrate

• Advantages
• tried and proven
• lower risk
• lower Cost
• available
• Disadvantage
• high dark current, Spec. 0.1 e/s/pixel, goal
  0.01 e/s/pixel
• limits observations
• low QE, vary with temperature (especially low
  temperature lt70K)
• persistence problems
• bright OH airglow leaves remnant when switching
  between gratings
• bright standard and reference stars will leave
  remnants

10
Dark Current

• Measured 0.01e/s/pixel with long tail beyond
  0.1e/s/pixel at 78K

Rockwell WWW pages for HAWAII-1
11
Performance Model

• 3600secs, RN 5e
• 0.1 e/s/pixel dark current dominates all
• 0.01 e/s/pixel K - background limited, JH -
  noise sources comparable
• 0.001 e/s/pixel - all background limited, JH -
  random noise less ?1.5 to ?2

12
Quantum Efficiency

• QE 60
• declining to lt50 shortward of 1.3µm

Rockwell WWW pages for HAWAII-1
13
Persistence

• Decay of residual images in 10 seconds darks
• Point source of 5 million photons/sec
• After 60 sec, residual 6000e
• After many hours, still much greater than dark
  current signal.

G. Finger ESO
14
5µm CdZnTe MBE Substrate

• Advantages
• Better lattice match with HgCdTe detector (less
  lattice defects 1/10)
• lower dark current (0.01 ? 0.001 e/s/pixel)
• eliminates persistence. (0.3-0.5 1st read, none
  thereafter)
• higher uniform QE (60 ? 85) (CdZnTe better
  refractive index match)
• Disadvantages
• CdZnTe poor thermal expansion match to detector
  substrate
• stress between detector substrate and mux. gt
  higher risk of debonding.
• schemes to address problem (balanced composite
  structure)
• CdZnTe only available in small wafers, each must
  be produced individually
• inefficient manufacture
• low yield
• expensive

15
Detector Development Status

• PACE 2048x2048 HAWAII-2 HgCdTe
• current development by Rockwell
• consortium of ESO, Uni of Hawaii, Subaru
• first HAWAII-2 device delivered to Klaus Hodapp
  Feb 00
• science detector delivered to consortium mid- to
  late-2000.
• Rockwell taking orders
• CdZnTe 2048x2048 HAWAII-2 HgCdTe
• 5µm cutoff devices are being developed for NGST
• no immediate plan to develop 2.5µm cutoff devices

16
NIFS Schedule

• Bare Mux End-2000
• Engineering device Mar-2001
• Science device Mid-2001
• Completed instrument Mid-2002

17
Recommendation

• 2.5µm PACE device available on instrument
  timescale. (18 month delivery)
• Order HAWAII-2 PACE device as soon as possible to
  get in queue
• Don Hall offer 5µm CdZnTe MBE array from NGST.
• difficult to block out to 5µm (PK50 blocker
  filters)
• will cryostat get cold enough, lt 68K, to stop
  5µm radiation
• 5µm has greater detector capacitance gt read
  noise greater

18
Detector Choice Issue Summary

• Chose 2.5µm PACE from Rockwell or 5µm CdZnTe from
  Don Hall?
• If 2.5µm PACE chosen by how much will the science
  of the instrument be compromised by dark current,
  quantum efficiency and persistence?

19
Spectrograph Detector Wiring
20
Requirements

• Safely mount detector and conduct signals outside
  cryostat
• Allow reading out of detector lt 5 secs
• Minimal heat load on cooling system
• Regulate detector temperature to /- 1 mK over
  60-90K
• Should not add more than 10 to read noise or
  introduce any crosstalk

21
Baseline
22
Drift

• Low drift necessary.
• Long integration times 3600secs
• nature of readout method (no overscan to
  subtract)
• low read noise 5e gt spec. drift read noise
• subtract fixed pattern dark current gt drift of a
  few electrons
• John Barton measured HAWAII-1 variation 1500e/K
• gtcontrol temperature to mK level

23
Detector Carrier

• Ceramic chip carrier, 19x19 pin grid array
• two outer rows for signals
• inner 15x15 pins for cooling

24
Detector Mounting

• Plastic Zero Insertion Force (ZIF) socket
• Soldered into Detector Mounting Board
• Cooling of detector

25
Detector Assembly

• inner 15x15 pins to cool
• equal length wires
• 1mK control
• thermal analysis

26
Detector Mounting Board

• Board material
• FR4 fibreglass
• investigate ceramic
• Protection and filtering discrete components
• SMC
• Indium solder stress relief

27
Protection Devices

• No Electro-static or overvoltage protection
  devices inside cryostat
• minimizing number of components increases
  reliability.
• no guarantee that protection devices have not
  failed and are indeed working properly.
• Electro-static protection provided by
• written handling and connection/disconnection
  procedures.
• provision of shorting plugs.
• when connected SDSU-2 boards provides resistance
  to ground.
• Overvoltage protection provided by
• SDSU-2 detector clock and bias drive boards.

28
Flex Circuit

• Require low wiring capacitance
• Keep settling times of output amplifier short
  (pixel rate is 5µs/pixel)
• Teflon only used once before at RSAA.
• Kapton much more popular and more flexible.
• Light leak with Kapton

29
Flex Circuit

• Detector read out time lt 5secs
• Pixel rate 5µs/pixel
• Minimize output amplifier settling time,
• Output voltage, VS , fixed
• Increase load current, IL , increases
  photoemission from amplifier
• Reduce capacitance, C, by
• minimize wiring length
• select flex circuit material to have low
  capacitance

30
Flex Circuit Material

• Teflon (PTFE)
• low dielectric constant (2.95) gt low capacitance
• low dissipation (0.009 at 1MHz)
• very low outgassing
• applications - high frequency digital equipment
• used at RSAA for CCD instrument
• Kapton (polymide)
• high dielectric constant (4.5)
• dissipation (0.15 at 1MHz)
• acceptable outgassing
• more flexible and more widely used.
• used by RSAA for WFI and AAO for IRIS (2562 PACE
  HgCdTe)

31
Detector Wiring Issues

• Can we control temperature to mK level?
• Should we not put electro-static or overvoltage
  protection devices inside cryostat?
• Should we use Teflon or Kapton flex circuits?

32
Detector Controller
33
Requirements

• Provide clocks, biases and analog output signal
  processor and digitizer to drive HAWAII-2 array
• Communicate with IOC to configure controller,
  initiate readouts, and to transmit image data.
• Read out the array in less than 5 secs
• Controller noise much less (20) than detector
  noise
• 9e DCS
• 4e fowler samples, n8
• Perform standard read noise reduction techniques.
• Readout single rectangular region of interest
• Limit amplifier glow - able to turn off amplifier
• Read out detector in an idle and run mode of
  operation

34
SDSU-2 Controller

• San Diego State University - Bob Leach
• Why
• RSAA has experience.
• interface to VME BUS gt IOC.
• Gemini standard controller.
• no other competitive controller with similar
  performance and cost.

35
Baseline Design
36
Video Processor Boards

• Two Choices
• Dual Channel IR Video Processor Board
• two analog signal video processing chains
• six bias generators
• two boards, four channels, (US10k) for read time
  lt 5µs
• image processing needs to be done in external
  processor.
• Quad Channel Coadder Video Processor Board
  (Coadder)
• four analog signal video processing chains
• no bias generators
• image -processing capability by Motorola DSP56000
  and 1Mword of SRAM.
• four boards (US40k) needed to provide 4Mwords
  SRAM for co-adding
• clock board provides biases

37
Dual Channel Board Chosen

• Video processors channels are the same
• Modest readout rate gt do co-adding in external
  processor
• Coadder cannot do linear fitting
• not enough memory for data and program
• Coadder more expensive
• US40K vs US10K
• longer development time for more complex DSP code
• Coadder more complex gt longer development time
  and higher risk

38
SDSU-2 Performance
39
SDSU-2 Performance Summary

• Sufficient performance
• four channels readout in 5secs, ADC 1µs
• Dynamic Range 50,000 counts at RN4e gt 16 bit,
  65536 levels
• fiber data rate (2.5 Mpixels/sec)
• timing sequencer resolution (40ns)
• DSP memory
• enough clocks and biases
• accommodate both PACE n-on-p and CdZnTe p-on-n

40
Noise

• Equivalent noise of HAWAII-2 is 5.5nV/?Hz.
  Using noise ?4kTR.
• (source follower, 5kohm load, gm1/3kohm)
• Input amplifier major contributor to electronics
  noise. (ADC OK)
• video porcessor schematic op-amp AD846, input
  noise of 11.4nV/?Hz
• voltage noise 2 nV/?Hz
• current noise 11.2 nV/?Hz
• op-amp AD829 used in coadder board and next stage
  of dual video board has equivalent input noise of
  2.8 nV/?Hz. Adds 15 to detector noise.
• by replacing AD846 by AD829, it should be
  possible to use video board pre-amplifier.
• possible to select better op-amp.

41
Drift

• Video Chain
• HAWAII-2 signal conversion gain of 3.4-6.8µV/e
• ADC-937 reference has drift referenced to input
  7.5µV/ºC
• specification drift noise 5e gt 17µV-34µV
• Biases
• 12 bit DAC DAC8420 has drift of 20µV/ºC.
• other devices have choice of different grades.
• example reference AD586, type M - 10µV/ºC, type J
  - 125µV/ºC
• endeavor to get board populated with best grade
  devices

42
Dummy Pixel

• Dummy Pixel - output pin on multiplexer
• Same circuit as output amplifiers, but connected
  to a bias
• Different capacitance gt same drift?
• Additional video channel may be required.
• Wait for Klaus to determine usefulness
• Another possibility is to use clusters of
  unbonded pixel to track and subtract drift

43
Linearity

• Detector load major source of controller
  non-linearity.
• gain of source follower proportional to gm
  (transconductance)
• gm varies with DC drain operating point current
• Two choices
• Resistor load
• DC drain current varies with output voltage gt
  0.5 non-linearity
• simple easy to use, no additional circuitry
• most common (Klaus Hodapp, G Finger)
• non-linearity known and can be corrected if
  problem.
• Current source load
• DC drain current more stable gt 0.1
  non-linearity
• difficult to design programmable, accurate and
  low noise current source
• additional circuitry required

44
SDSU-2 Performance

• Required readout time 5 secs
• 2Kx2K through 4 amplifiers gt pixel rate required
  5µs/pixel
• SDSU-2 fiber optic data rate 2.5 Mpixels/sec.
  For four channel system, data can be sent up to
  1.6 µs/pixel gt gt required
• Timing sequencer resolution is 40ns gt adequate
• DSP5600 Memory
• 32Kx24bits8K P program,8K X data,16K Y
• No pixel data will be stored in this memory
• Looking at HAWAII-1 code gt adequate

45
Clocks and Biases

• HAWAII-2 require
• 10 clocks/quadrant and 5 biases/quadrant
• different quadrant clocks and biases connected
  together
• 24 clocks/board gt 1 clock board adequate
• 6 biases/video processor gt two video boards
  adequate
• 12bit DAC programmable over selectable voltage
  range
• 0 to 5V, -5 to 0V
• gt accommodate both PACE and CdZnTe

46
A/D Converter

• Conversion speed
• 1µs including 300ns internal sample/hold gt
  adequate 5 µs/pixel
• Dynamic Range
• HAWAII-2 potential well depth 100,000
• use 50,000
• read noise 4e- (n4 fowler sampling)
• gt 16 bit, 65536 levels adequate

47
Readout Methods

• Support three readout methods
• Double Correlated Sample
• signal end read - start read
• removes kTC noise, good for quick look display
• Fowler Sampling
• average of n/2 reads at end - average of n/2
  reads at start
• noise improvement ?(n/2)
• simple and easy to implement
• Linear Fitting (Non Destructive Read)
• reads throughout exposure
• linearize and least-square fit to obtain photon
  rate
• Noise Improvement ?(n/12)

48
Double Correlated Sample

• Reset ? Read start Sr ? Expose ? Read end Er
• Signal Er - Sr
• Low noise 9e
• Removes kTC noise
• Removes bias structure gt good for quick look
  display

49
Fowler Sample

• Reset ? n/2 reads at start, Sri ? accumulate
  charge for expose time ? n/2 reads at end, Erj

• Noise Improvement ?(n/2)
• Simple and easy to implement

50
Linear Fitting

• Advantages
• bias is small gt in real time see useful data on
  quick look display
• long exposures are punctuated by frequent display
• pixels that saturated can be measured properly
• attenuate 1/f noise (DC drift) by frequent
  samples
• Disadvantages
• data must be linearized in real time.
• complex and requires computing power gt PowerPC
  required
• more samples to get same improvement as fowler
• ?(N/12) versus?(N/2)
• every read means noise e.g. Mux. and amplifier
  glow
• other institutions have elected to use simpler
  fowler sampling

51
DSP code

• Motorola 56000 DSP
• Two DSPs
• VME interface board
• interrupts needed to sequence data transfer
• use GMOS code as starting point
• Timing Board
• considerable amount of code around for HAWAII-1
• either modify HAWAII-1 or wait for HAWAII-2 code
• linear fitting routine needs to be added
• preliminary command interface document written
• RSAA has considerable experience

52
VME Interface Board DSP Code

• VME boards command and data communication path
  between fiber link on timing board and IOC.
• DMA to IOC memory.
• Motorola 56001 DSP, on board memory 32K
• No special data processing or data sorting
  required
• Standard code exists, but needs interrupt to
  report progress of data transfer during data
  taking
• e.g. linear fitting, one command ? many frames of
  data
• GMOS code provides this feature and is good
  starting point

53
Timing Board DSP Code

• Major code controls all aspects of read out
• sets clocks and biases levels
• turn on and off the clocks and biases
• perform sequencing of clocks and video processor
  channels to readout detector
• controls data transfer
• Many other institutions use SDSU-2 to drive
  HAWAII detectors
• Modify their code for our needs

54
Grounding and Shielding
55
Grounding and Shielding

• Adhere to good grounding and shielding practices
• Design ground system
• Minimize earth loops
• star point
• make sure cryostat earthed to telescope only. no
  earth through components controller, cryocooler
  lines, cables etc
• Separate noisy grounds from low level signal
  grounds
• All cabling use shields
• Keep noisy circuitry and cables away from low
  level signals.
• Only earth shields at one end, preferably at
  source
• Orient cryocooler so that its radiated large
  magnetic field has least affect on the detector
  controller.
• Electrically isolate shields from each other
  especially noisy shields from low level shields
• eg. clocks from biases from output signal

56
Detector Ground

• Star point ground
• Use isolated internal shield around detector and
  detector wiring
• Cryostat grounded once to telescope
• Shield all cables.
• Connect shields at one end only
• separate noisy signals and cables from low level
  ones

57
Component Controller Ground

• Will the NIRI duplicate component controller
  compromise our grounding scheme by having
  multiple ground connections?

58
Detector Controller Issues

• Will the drift of the detector controller be low
  enough?
• Will the dummy pixel on the multiplexer be
  useful?
• Will we need to regulate the temperature of the
  detector controller?
• Will the on video board pre-amplifier noise be
  low enough?
• Are the three readout methods necessary?
• Will resistor load for detector output amplifier
  produce sufficient linearity? Should we consider
  current sources?
• Will NIRI duplicate component controller
  electronics compromise our grounding and
  shielding scheme?

59
Electrostatic Protection

• Protect detector from damage during assembly,
  disassembly, transport, storage, normal use.
• No static protection devices inside cryostat
• Provide static shorting mating connectors with
  ground lead.
• Write procedures for safely
• assembling and disassembling various detector
  components
• connecting and disconnecting the detector
  controller from cryostat.
• When controller connected protection is by
• video and bias boards provide resistance to
  ground
• clock board provides resistance to ground, zener
  diode protection and analog switch isolation.

60
Detector Characterization
61
Test Facilities

• Test Software
• characterize and optimize detector
• CICADA Instrument Control and Data Acquisition
• support SDSU-2 for CCDs
• modify to support IR detectors
• Data Analysis Software
• use IRAF
• suite of custom IRAF routines for data analysis
  under development
• NIFS specific routine will be added as needed
• Test Cryostat
• a suitable test cryostat will be built for
  characterizing the detector

62
Test Cryostat Measurement

• What measurements do we want to do
• read noise, dark current - completely blank off
  with cold stop
• photon conversion gain, full well capacity and
  linearity - illuminate evenly with stable source
  intensity (e.g.IR LEDs inside cryostat)
• quantum efficiency - illuminate with repeatable
  stable source of known intensity (e.g. black
  body) - absolute vs relative
• cross-talk and persistence - test image generator
  such as hot spot and lines
• fringing???? - should we wait until in final
  cryostat

63
Test Cryostat Proposed

• Modify existing Irlabs 8 double can cryostat
• Filter wheel J, H, K, blank for darks
• Cold optics for imaging - does not look out
  through wide angle
• IR LEDs for flat fields
• Test pattern and field mask wheel - measure
  cross-talk and persistence
• Mount in 74inch Coudee and do spectrograph tests
• Custom wire bring out all detector signals
• test each quadrant independently
• engineering array may have problems
• determine which signals can be paralleled
• test all wiring materials, components and
  construction techniques
• Accommodate final mount and wiring

64
Methodology
65
Operating Point

• Vary temperature and detector voltage, measure
  and plot
• read noise
• photon conversion gain
• dark current
• full well capacity
• output amplifier settling time
• cosmetics
• quantum efficiency
• Select suitable operating point
• good QE
• sufficient well depth, (50ke)
• minimizes read noise, dark current and cosmetic
  problems

66
Characterize

• Measure persistence and work out how to live with
  it
• Measure linearity and if need workout how to
  correct it
• Stability
• check whether mK control of detector temperature
  is adequate
• measure how detector controller varies with
  temperature and time and how these variations
  affect the detector
• biases and clock
• video processor chain
• if need regulate detector controller temperature
• Multiplexer and amplifier glow
• best technique to manage glow
• verify turning output amplifier off during
  exposure doesnt cause drift problems
• Measure fringing, determine its stability and how
  to flat field

67
Detector Characterization Issues

• Can we afford to build a test cryostat to do
  everything we want to do in the time allocated?
• Should we build test wiring or should we use the
  final wiring in the test cryostat?
• Can we reduce drift to acceptable level?
• Can we learn to live with persistence?
• Can we successfully manage amplifier glow?
• Will fringing be stable and be able to be flat
  fielded?
• Will we achieve low enough read noise, dark
  current at a sufficient well depth and quantum
  efficiency to do good science?

68
END