JWST's Near-Infrared Detectors: Ultra-Low Background Operation and Testing

1
JWST's Near-Infrared DetectorsUltra-Low
Background Operation and Testing

• Bernard J. RauscherSpace Telescope Science
  Institute

2
Outline

• What is a Near-Infrared Array Detector?
• JWST Science Drivers
• Detector Requirements
• Detector testing at STScI/JHU
• Optimal Use
• Summary

3
JWSTs IR Arrays are Hybrid Sensors

• PN junctions are bump bonded to a silicon
  readout multiplexer (MUX).
• Silicon technology is more advanced than other
  semiconductor electronics technology.
• The bump bonds are made of indium.

4
JWST Needs Very Good Near Infrared Detectors!

• Completing the JWST Design Reference Mission on
  time requires background limited near-infrared
  (NIR) broadband imaging
• Zodiacal light is the dominant background
  component in the NIR
• The total NIR detector noise requirement is
  therefore 10 e- rms in a t1000 seconds
  exposure.
• NIRSpec will probably be detector noise limited.
  The total noise goal is 3 e- rms per 1000
  seconds exposure

5
JWST Near Infrared (NIR) Detector Requirements
6
Detector Testing at STScI/JHUIndependent
Detector Testing Laboratory
7

• Past and present personnel

Eddie Bergeron Data Analyst
Tom Reeves Lab Technician
Robert Barkhouser Optical Engineer
Bernie Rauscher Project Scientist
Utkarsh Sharma Graduate Student
Mike Telewicz Intern
Gretchen Greene Mechanical Engineer
Steve McCandliss JHU Lead
Ernie Morse Data Analyst
Monica Rivera Intern
Scott Fels Intern
Don Figer Director
Sito Balleza Systems Engineer
Mike Regan System Scientist
Russ Pelton Technician
8
NIR Detector Characteristics

• Dark current
• Read noise
• Linearity
• Latent charge (persistence)
• Relative and Absolute Quantum efficiency (QE)
• Intra-pixel sensitivity
• Thermal stability
• Radiation immunity

9
Dark Current

• Lowest measured dark current is 0.006
  e-/s/pixel.

10
IDTL Measurements Read Noise

• Read noise is 10 e- for Fowler-8. (system read
  noise is 2.5 e-)

11
IDTL Measurements Conversion Gain
Per correlateddouble sample
12
Hawaii 1R with 5 um Cutoff
13
IDTL MeasurementsRelative and Absolute Quantum
Efficiency
We are currently working on better calibration to
enable measurements of absolute QE vs. wavelength.
14
Relative QE Maps

• Relative QE maps show significant structure

15
Hawaii Shirt
IDTL Test System
Hawaii Detector
16
Then Now
November 2000
November 2002
17
IDTL First Light Images
Rockwell HAWAII-1RG
Jun. 02 (MUX)
Jul. 02 (SCA)
18
IDTL Test System
Leach II Controller Electronics
Dewar
Entrance Window
Vacuum Hose
He Lines
19
Detector Readout System
T30-50 K
Unix Instrument Control Computer
Warm Harness
COTS Leach II IR Array Controller
T293 K
Cryogenic Harness
Detector Customization Circuit
JWST SCA
20
An Adaptable Readout System

• The only hardware change required to run a
  different detector is swap-in a DCC.
• We have DCCs for the following detectors.
• Raytheon
• SB-290
• SB-304
• Rockwell
• HAWAII-1R
• HAWAII-1RG
• HAWAII-2RG
• Each DCC is a multi-layer PCB. Extensive use of
  surface mount technology. Includes flexible
  neck to simplify interfacing.

21
Close-up ofDetector Customization Circuits (DCCs)
22
Optimal Use

• JWST Detector Readout Strategies
• Anomalies seen in other instruments
• Other effects
• Use of Reference Pixels

23
Detector Readout

• JWST science requires MULTIACCUM and SUBARRAY
  readout.
• Other readout modes can be implemented using
  parameters.
• For example, Fowler-8 can be implemented as
  MULTIACCUM-2x8.
• Cosmic rays may be rejected either on the ground
  or on-orbit.

24
NICMOS Anomalies ( how JWST will avoid them)

• Dark current
• JWST detectors already designed to minimize glow
• Careful detector characterization selection
• Do not exceed max temp. requirement!
• Bias drifts
• Good electronic design
• Avoid power supply coupling
• Avoid ground coupling
• Reference pixels will help
• Synchronous readout can help
• QE variations
• Careful detector characterization selection
• Amplifier glow
• JWST detectors should be much better than NICMOS

25
NICMOS Anomalies 2

• Persistence
• There will be persistence on JWST
• Strongly dependent on detector fabrication
  process
• Careful detector characterization selection
  needed to choose best detectors
• In IDTL, we are exploring mitigation measures

26
NICMOS Anomalies 3

• DC bias level drift
• Good electronic design is first line of defense
• Reference pixels should eliminated Pedestal
  drifts.
• Depending on reference pixel layout, reference
  pixels may help reject bands.
• Ghosts
• In NICMOS, may result from ground plane coupling
  within the MUX.
• Also seen in SIRTF InSb radiation testing.
• Good cable harness and electronic design help

27
NICMOS Detector Effects

• Linearity
• In NICMOS, 10 intrinsic non-linearity can be
  calibrated out to within 0.2.
• Well depth
• Well-depth is a function of reverse bias in
  photo-voltaic detectors.
• Well-depth can also depend on temperature.
• In the IDTL, we will study well depth as a
  function of reverse bias and temperature.

28
NICMOS Detector Effects 2

• QE
• Can depend on wavelength and temperature.
• Dark current bump
• This is a curious effect seen in NICMOS.

29
Reference Pixels

• All candidate JWST detectors have reference
  pixels
• Reference pixels are insensitive to light
• In all other ways, designed to mimic a regular
  light-sensitive pixel
• NIR detector testing at University of Rochester,
  University of Hawaii, and in the IDTL at STScI -gt
  reference pixels work!
• Reference pixel subtraction is a standard part of
  IDTL data reduction pipeline

30
Use of Reference Pixels

• JWSTs NIR reference pixels will be grouped in
  columns and possibly rows
• Most fundamentally
• reference pixels should be read out in exactly
  the same manner as any normal pixel
• Data from many reference pixels should be
  averaged to avoid adding noise to data
• We have begun to explore how reference pixels
  should be used. Approaches considered include the
  following.
• Maximal averaging (average all reference pixels
  together and subtract the mean)
• Spatial averaging
• Temporal averaging
• Spatial averaging is now a standard part of IDTL
  calibration pipeline

31
A Picture of IDTL System Noise

• Shorting resistor mounted at SCA location
• 1/f tail causes horizontal banding.
• Total noise is 7 e- rms per correlated double
  sample.

32
Averaging small numbersof reference pixels adds
noise

• Averaged the last 4 columns in each row and
  performed row-by-row subtraction

33
Spatial Averaging
This is a standardpart of the IDTL
datacalibration pipeline

• In spatial averaging, data from many (64 rows)
  of reference pixels are used to calibrate each
  row in the image
• A Savitzky-Golay smoothing filter is used to fit
  a smooth and continuous reference column
• This reference column is subtracted from each
  column in the image
• Using this technique, we can remove some 1/f
  noise power within individual frames
• In practice, this technique works very well

34
Spatial Averaging Before After
Before
After
35
Temporal Averaging

• Dwell on the reference pixel and sample many
  times before clocking next pixel
• Potentially removes most 1/f
• Not tried this in IDTL yet. U. Hawaii has
  reported some problems with reference pixels
  heating up

36
Temporal Averaging Before After
Before
After
37
Summary of Reference Pixel Calibration Methods

• Spatial averaging works well using a Rockwell
  HAWAII-1RG detector
• Based on conversations with U. Rochester, we
  foresee no problems with SB-304
• Temporal Averaging is promising. More work needed
  using real detectors.

38
Summary

• The Independent Detector Testing Laboratory
  (IDTL) at STScI/JHU is up and running
• Test results including dark current, read noise,
  conversion gain, relative quantum efficiency, and
  persistence are in good agreement with other JWST
  test labs
• Reference pixels work and are an invaluable part
  of the data calibration pipeline
• We have explored three techniques for using
  reference pixels
• Maximal averaging,
• Spatial averaging,
• Temporal averaging
• Spatial averaging works well and is robust
• Early reports from U. Hawaii using temporal
  averaging are not encouraging due to reference
  pixel self-heating. More work is planned in the
  IDTL