Spectrograph Detector - PowerPoint PPT Presentation

1 / 50
About This Presentation
Title:

Spectrograph Detector

Description:

... s/pixel with long tail beyond 0.1e/s/pixel at ... first HAWAII-2 device delivered to Klaus Hodapp Feb 00 ... Don Hall offer 5 m CdZnTe MBE array from NGST. ... – PowerPoint PPT presentation

Number of Views:94
Avg rating:3.0/5.0
Slides: 51
Provided by: PeterMc150
Category:

less

Transcript and Presenter's Notes

Title: 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
Write a Comment
User Comments (0)
About PowerShow.com