Transcript Slide 1

James Webb Space Telescope:
University of Rochester Detector Testing
on Raytheon SB-304 InSb SCAs
2 Sep 2003
Craig McMurtry, William Forrest,
Judith Pipher, Andrew Moore
Overview
• Introduction
• SB-304 operation
– Number of clocks, biases, output and other
• Calibration of InSb SB-304 SCAs
– Source Follower Gain
– Capacitance
– Well Depth
– Linearity
• Dark current
– Methods of measurement
– Results
• SCA 006
– Toptimum and Tmax
– Dark current versus inverse temperature (Arrhenius plot)
• SCA 008
2
Overview (continued)
• Noise
– Methods of measurement
– Results
• System Noise
• Read noise in 100 seconds integration
– SCA 006
– SCA 008
• Total noise in 1000 seconds integration
– SCA 006
– SCA 008
– Summary of noise
3
Overview (continued)
• Quantum Efficiency
– Methods of measurement
• UR dewar optics and calibration equipment
• Responsive Quantum Efficiency (RQE)
• Detective Quantum Efficiency (DQE)
– Results
• SCA 006
• SCA 008
– Comparison to AR coating
• Latent or Persistent Image Performance
– Methods of measurement
– Results
– Possible amelioration techniques
4
Overview (continued)
• Operability
– Definitions
– Results
• Basic operability
• Radiometric Stability
– Method of measurement
– Results
5
Overview (continued)
• MTF and Electrical Cross-talk
– Methods of measurement
• Cosmic Ray Pixel Upsets
• MTF using knife edge and circular apertures
– Results
• Summary of SB-304 InSb SCA performance
6
Introduction
• Raytheon Detectors Proposed for JWST NIRCam and NIRSpec
– InSb detector technology
• 0.5 – 5.3 mm photo-response
– Based on SB-304 Read Out Integrated Circuit (ROIC) or multiplexer
• 2048 x 2048 active pixels
• 2 columns of 2048 reference pixels multiplexed to four outputs
• Total readout format is 2056 x 2048
– University of Rochester provided detector array testing facilities for
JWST level requirements
• Competition was/is with Rockwell Scientific and University of Hawaii
– HgCdTe detector technology
• 5 mm cutoff
7
JWST Requirements
Parameter
Requirement (Goal)
SCA Format
2048 x 2048 pixels
Fill Factor
/95% (100%)
Bad Columns/Rows
<5 containing >1000
Bad Pixel Clustering
< 20 cluster up to 20 pixels
Pixel Operability
>98%
Total Noise 1000 s
[9 e- (2.5 e-)
Read Noise for single read
[15 e- (7 e-)
Dark current
< 0.01 e-/s
8
JWST Requirements
Parameter
Requirement (Goal)
DQE
70% 0.6[ l [ 1.0 mm
80% 1.0[ l [ 5.0 mm
(90%; 95%)
Well Capacity
> 6x104e- (2x105e-)
Electrical Cross-talk
<5% (<2%)
Radiometric Stability
1% over 1000 s
Latent Image
< 0.1% after 2nd read following >80%
full well exposure
Frame Read Time
12 sec (<12 sec)
Pixel read rate
100KHz; 10 ms/pix
Sub-array read
0.2 s for 1282 pixels
9
SB-304 Operation
• Number of required connections
– 7 Clocks
• pC1, pC2, pR1, pR2, vRstG, pRstR, VrowOn
– 9 Biases
• Vp, VnRow, VnCol, VddOut, Islew, Vssuc, Vdduc, Vdetcom, VrowOff
– 4 Output
– 1 ground
• Additional connections
– 2(4) wires for temperature sensor
– 1 for diagnostic Tend
– 1 clock pScanCol for bi-directional control, fast guiding
– 1 external load current source for output (warm connection only)
10
Calibration
• Source Follower Gain
– SCA 006 SFGain=0.777
– SCA 008 SFGain=0.785
• Capacitance
– Noise2 vs Signal method
– SCA 006
• 66 fF
• 3.22 e-/ADU
– SCA 008
• 68 fF
• 3.32 e-/ADU
11
Calibration
• Linearity
– Plotted Signal Rate vs
Signal (C0/C)
– Small flux over long
integration times
• Well Depth (Capacity)
– @ 300 mV applied
detector bias
– SCA 006 well depth =
1.4 x 105 e– SCA 008 well depth =
1.3 x 105 e– Larger well depths
possible with little or
no increase in dark
current
12
Dark Current Test Methods
• Dark dewars are difficult to
make and keep dark
– Using an opaque mask
placed in contact with
InSb surface, UR dewar
light leak < 0.006 e-/s
• 3 Methods of measurement
– Usually yield same
values, although some
discrepancies possible
• Dark Charge versus
integration time
– With reference pixel
correction, accurate for
moderate dark currents
– Lengthy measurement
13
Dark Current Test Methods
• Noise2 versus
integration time
– With reference pixel
correction, accurate
for small dark
currents
– Also, lengthy
measurement
14
Dark Current Test Methods
• SUTR Dark Charge vs. time
– With reference pixel
correction, accurate for
small dark currents
– Relatively short
measurement (single
2200 sec integration)
– Addition of possible
charge per read due to
higher read rate
• Confuses dark current
measurement
• No detectable added
noise
15
Dark Current Results
• SCA 006
– Idark = 0.012 e-/s @
T=30.0K (Toptimum)
– Idark = 0.024 e-/s @
T=32.3K (Tmax)
– Charge per read of
0.09 e-/read
• Again, no detectable
added noise
– No measurable amp
glow or digital circuit
glow
16
Dark Current Results
• SCA 008
– Idark = 0.025 e-/s @ T=30.0K
– Charge per read of
0.07 e-/read
– No digital circuit glow
– Slight glow from output
amplifier
• 0.05 e-/s including dark
current
• Covers small region (see
operability section)
• Known multiplexer defects
(shorts)
– Amp glow not seen on
other multiplexers
17
System Noise
• System Noise
– Shorting resistor placed
between signal (video)
and signal reference
lines (analog ground)
– T=295K
– Connected and
functioning detector in
dewar to allow typical
voltage/current paths
which may cause cross
talk (worst case)
18
Read Noise Results
• Read noise
– Measured at T=30.0K
– All integration times are
100 s
• SCA 006 results
– Follows 1/sqrt(N) where
N is the number of
Fowler sample pairs
19
Read Noise Results
• SCA 008 results
– Follows 1/sqrt(N)
20
Total Noise Measurement Methods
• Methods of measurement
– Box average (often called
“spatial” noise method) uses
the {standard deviation of
mean}/sqrt(2) of difference of
two 1000 sec Fowler-8 images
– Full frame average (“spatial”)
noise computed using
difference of two 1000 sec
Fowler-8 images, and plotting
histogram of pixel values
• The width of the distribution
corresponds to the average
noise; mean is DC offset
• Gaussian fit rejects cosmic ray
• SCA 006 at right
21
Noise Measurement Methods
• Methods of measurement (cont)
– Temporal noise measurement
is computed by taking the
standard deviation of the
mean per pixel for a large
number of 1000 sec Fowler-8
images (time series)
• Distribution is typically a
Gaussian whose width
depends on the number of
images taken.
• Cosmic Ray hits removed
from single images (4 s
clipping).
22
Noise Results
• Total Noise Requirement: < 9 e- in 1000 sec using Fowler-8 sampling
– SCA 006
• 6.2 e- (Temporal method), 6.7 e- (Full frame spatial method) @ T=30.0K
– Note on charge per read: temporal noise data are Fowler-8 images that
were re-constructed from 98 samples of a SUTR series. From the dark
current results, 0.09 e-/read was inferred. One would expect to see
(98-16 read)*(0.09e-/read) worth of noise power. However, the noise
for the reconstructed Fowler-8 images of temporal method was LESS
than the noise for standard Fowler-8 spatial method, i.e. no detectable
noise contribution.
• 6.4 e- (Full frame spatial method) @ Tmax = 32.3K
• For 1000 sec Fowler-1, total noise is 12.0 e- (temporal method) @T=30.0K
– SCA 008
• 7.9 e- (temporal method) @ T=30.0K
23
Quantum Efficiency
• UR dewar cross section optical
path
• AW is as simple as possible.
• All IR filters from OCLI or Barr
– Transmission traces taken
at room temperature and
77K
• Visible filter, KG-5
– Transmission trace at room
temperature and 4.2K
• Still have some optical
problems (large angles!), likely
interference patterns and
vignetting
– Central portion illuminated
well
24
Quantum Efficiency
• Reconstructed psuedo-flat fields for SCA 008, cos4q corrected
– Most effects are caused by dewar optics, not detector; corners are vignetted
– J band on left, L’’ (3.81mm) on right
25
Quantum Efficiency
• Photon sources and calibration equipment
– For l > 3.0 mm, photon source is room temperature black body
surface monitored with a calibrated temperature sensor
• Subtract “extra signal” from image taken of liquid nitrogen cup
– For 1.0 mm < l < 3.0 mm, photon source is NIST calibrated black body
(Omega BB-4A, 100 – 1000 C, e =0.99)
– For l<1.0 mm, photon source is stabilized visible light source feeding
an integrating sphere with a NIST calibrated Si diode detector
• Responsive Quantum Efficiency -- can be > 100% due to gain
– RQE = signal/(expected #photons)
• Signal is averaged signal measurement, corrected for non-linearity
• Expected # photons from NIST calibrated detector or spectral black body
calculations
• Detective Quantum Efficiency -- is < 100%
– DQE = (Signal/Noise)2/(expected #photons)
• Noise obtained via standard deviation of difference of two measurements
26
Quantum Efficiency Results
RQE;
DQE
0.65mm
RQE;
DQE
0.70mm
RQE;
DQE
1.25mm
RQE;
DQE
1.65mm
RQE;
DQE
2.19mm
RQE;
DQE
3.81mm
RQE;
DQE
4.67mm
RQE;
DQE
4.89mm
SCA
006
88%;
82%
105%;
95%
107%;
97%
96.2%;
96.7%
84.6%;
85.3%
97.1%;
98.5%
84.7%;
85.0%
80.1%;
-
SCA
008
-
-
114%;
97.1%
-
-
-
86.8%;
-
-
DQE closely matches expected value from AR coating transmission (see
Raytheon data on AR curve). From this, we infer that the optical fill factor is
> 98%.
27
Latent Image Measurement Method
Our test procedures are described here (since the result one gets depends critically on the
exact procedure):
•
•
•
•
•
•
•
•
•
•
•
•
•
•
a. Very dark control region on array provided by an opaque mask of black paper.
b. Use nominal bias. The number of latent traps populated depends upon the applied bias and depletion width.
c. Wait at least 15 minutes on cold dark slide (assures no prior latents).
d. Take multiple dark exposures for use as background level.
e. Move directly from cold dark slide to filter's edge (this is the source). No other filter is allowed to pass in front of optical path in
this transition. Use of filter edge to illuminate array provides a gradient of fluxes across array to allow choice in flux/fluence levels
during analysis. Should do tests at several wavelengths.
f. Integrate for Source Exposure Time. The number of latent traps populated depends upon the applied bias and thus depletion
width. If the depletion width decreases (as it does during integration under illumination), then more traps near the implant will be
exposed and collect charge. See Benson et al. ("Spatial distributions hole traps and image latency in InSb focal plane arrays",
Proc SPIE Vol. 4131, p. 171-184, Infrared Spaceborne Remote Sensing VIII) specifically figures 6 and 7.
g. Move back to cold dark slide (again, no other filters pass array).
h. Delay time is time to move filter wheel plus reset time plus time to mid-point of pedestal (e.g. JWST minimum is 6s in Fowler-1).
Propose 30s delay =expected JWST dither time. Any amelioration techniques allowed during this interval (e.g. autoflush in the
STScI tests).
i. Take "darks" at Latent Integration Time in a loop such that a pair of tests {(1 and 2) or (4 and 5)} are completed for the same
single source exposure. UR usually takes twice as many darks as required. Multiple sampling and/or multiple pixel average
assumed.
j. Reduction: All statistics are done with 4 sigma clipping to eliminate dead/hot pixels and cosmic rays. Use 4 column by 25 row
box averages (# of columns chosen to keep fluence roughly constant over box - gradient from filter edge, while # of rows chosen
to reduce pixel to pixel variation).
A. Remove background level due to any light leak or dark current using prior
dark frames.
B. Remove any frame-to-frame instability (using reference pixels or masked off
region as reference level).
28
Latent Image Results
Test
#
Srce
Flux
(e-/s)
Source
Exposure
(s)
Source
Fluence (e-)
Delay
(s)*
Latent Integr’n
Time (s)
Max. Desired
Latent
Fluence
(e-: %)
Meas’d (%)
Latent
Fluence
SCA006 ; SCA008
1
300
100
30,000
30
100
9
0.3
2
300
100
30,000
1000
100
0.9 ; 0.003
0.017 ; ≤0.01
3
30
1000
30,000
30
1000
4.5 : 0.015
;
4
300
500
150,000
30
100
90
; 0.06
0.48 ; 0.22
5
300
500
150,000
1000
100
9
; 0.006
0.03 ; ≤0.01
6
3
10,000
30,000
200
8000
Noise level
7
15
10,000
150,000
200
8000
Noise level
; 0.03
; 0.12
29
Operability
• Operability is affected by two types of defects:
– Missing contact between InSb diode implant and multiplexer unit cell
• First InSb bump-bonding to mux had moderate outages.
• Significant strides made in very short time (see next slides).
– PEDs (Photo-emissive defects)
• Defect centers that glow (both IR and visible photons).
• Techniques in place which either eliminate or dramatically reduce glow
region such that ~20-40 pixel diameter region fail operability.
• Future multiplexers will have additional circuitry to fully eliminate all
PEDs.
• Foundry improvement to reduce/eliminate defects.
30
Operability
• SCA 006
– Basic Fail = 13.5%
– Large fraction failing are
unconnected pixels
31
Operability
• SCA 008
– Basic Fail = 1.94%
– Slight amp glow in lower
left
32
Radiometric Stability
• Method of measurement
– Using similar technique as RQE measurement at l= 3.50 mm, a room
temperature black body source was the source of “stable” flux.
– A calibrated temperature sensor was used to monitor/calibrate
variations in the temperature of the black body (radiation source).
– A series of integrations were then taken over a 9 hour period.
– Most of the errors or inaccuracies in this measurement are a result of
source calibration error or instabilities in our system electronics and
not due to the SCA itself.
• Result
– SCA 006 exhibited instabilities < 0.07% over 1000 s and < 0.19% over
the total 32000 s.
– Further improvement by factor of 10 - 100 may be gained by using
our NIST calibrated black body source.
33
MTF and Electrical Cross-Talk
• Methods of measurement
–MTF using knife edge and circular apertures placed in contact
with InSb surface
–Cosmic ray hit pixel upset for electrical cross-talk
34
MTF and Electrical Cross-Talk
• MTF results
– Edge spread functions
shown for two
wavelengths
– Edge spread modeled by
diffusion and rectangular
pixel function which is the
ratio of {pixel pitch/
distance between photon
absorption and the
depletion region}
35
MTF and Electrical Cross-Talk
• MTF results (cont.)
– From the best fit model parameter, z (frequency in cycles/thickness)
can be determined, which in turn leads to MTF:
MTF = 0.64 (2 e –2pz)/(1 + e-4pz)
– If Nyquist frequency is taken as ½ z, then MTF = 0.45
• Similar measurement on SB-226 InSb SCA produced MTF=0.52
– If Nyquist frequency is taken as ¼ z, as in Rauscher’s MTF
document, then MTF = 0.58
• Exceeds (existing) requirement of 0.53 in NASA JWST 641 document
36
MTF and Electrical Cross-Talk
• Cosmic ray hit pixel upsets
used to quantify electrical
cross-talk
– Used CRs which appear to
be normal incidence with
charge predominantly in
one pixel and equal
distribution to neighbors
– Histogram of 30K dark data
difference showing peaks
at 0.1% for next nearest
neighbors and 0.5-1.2% for
nearest neighbors
– Cross talk is < 2%
37
MTF and Electrical Cross-Talk
• 4th pixel over electrical cross-talk
– 4 interleaved outputs = next pixel on same output is 4 pixels away
– Deterministic, can be removed or corrected in software
– Below is a table of pixel values in percentage of a single cosmic ray
event; notice 4th pixel over is 2%
0
0.025
0.012
0.025
0.012
-0.025
0.037
-0.037
0.012
0.099
0
-0.025
0.074
0.546
0.099
0
0.037
-0.012
0.025
-0.062
0.012
-0.050
1.142
100
0.782
0.137
-0.248
2.062
-0.211
0.012
0.062
0.012
0.161
0.733
0.012
0.062
-0.074
0
-0.050
0
0.025
-0.062
-0.037
0.012
0.012
0
0.074
0.025
0.012
0.001
38
Power Dissipation
• Power Dissipation per 2K x 2K InSb detector array
– Original requirement was < 1 mW per 1K x 1K array.
– Measured by summing powers generated by voltages and currents
{see Wu, et al., Rev Sci Inst., 68, 3566 (1997)}.
– Total power dissipation on ROICs with moderate shorts < 0.37 mW
– Total power dissipation on ROICs with no shorts < 0.1 mW
39
Additional Tests
• NASA Ames conducted proton radiation testing at UC Davis
– Please see paper “Radiation environment performance of JWST
prototype FPAs” McCreight, et al., SPIE Vol. 5167 (in publication)
• STScI IDT Lab conducted independent tests on both InSb detector
arrays from Raytheon and HgCdTe detector arrays from Rockwell
Scientific.
– Please see paper “Independent testing of JWST detector prototypes”
Figer, et al., SPIE Vol. 5167 (in publication)
40
Summary of
SB-304 InSb SCA Performance
Parameter
Requirement (Goal)
SB-304-006 Result
SB-304-008 Result
SCA Format
2048 x 2048 pixels
2048 x 2048 active
+ 2 reference
columns
2048 x 2048 active
+ 2 reference
columns
Fill Factor
/95% (100%)
/98% (100%)
/98% (100%)
Bad
Columns/Rows
<5 containing >1000 No
Yes
Bad Pixel
Clustering
< 20 cluster up to
20 pixels
No
Yes
Pixel Operability
>98%
86.5% basic
98.1% basic
Total Noise 1000 s
[9 e- (2.5 e-)
6.2 e-
7.9 e-
Read Noise for
single read
[15 e- (7 e-)
12 e- (CDS)
14.5 e- (CDS)
Dark current
< 0.01 e-/s
0.012 e-/s
0.025 e-/s
41
Summary of
SB-304 InSb SCA Performance
Parameter
Requirement (Goal)
SB-304-006 Result
SB-304-008 Result
DQE
70% 0.6[ l [ 1.0 mm
80% 1.0[ l [ 5.0 mm
(90%; 95%)
82% @ 0.65 mm
97% @ J,H,L’’
97% @ J
Well Capacity
> 6x104e- (2x105e-)
1.4 x 105e-
1.3 x 105e-
Electrical Crosstalk
<5% (<2%)
<1.3% (nearest and
next nearest pixel)
<1.3% (nearest and
next nearest pixel)
Radiometric
Stability
1% over 1000 s
< 0.07% over 1000s < 0.07% over 1000s
Latent Image
< 0.1% after 2nd read
following >80% full
well exposure
0.3%
(no amelioration)
0.12%
Frame Read Time
12 sec (<12 sec)
< 11 sec
< 11 sec
Pixel read rate
100KHz; 10 ms/pix
100KHz; 10 ms/pix
100KHz; 10 ms/pix
Sub-array read
0.2 s for 1282 pixels
<0.05 s for 1282
<0.05 s for 1282
42
Conclusions
• Raytheon has produced a robust, mature InSb detector array
technology.
• Both the InSb detector arrays from Raytheon and the HgCdTe detector
arrays from Rockwell Scientific have demonstrated excellent
performance.
• The University of Arizona has selected Rockwell Scientific to produce
the NIRCam SCAs and FPAs.
– Congratulations to UH and RSC!
43