Recent Performance Improvements, Calibration Techniques and Mitigation
Strategies for Large-format HgCdTe Arrays
G. Finger, R. Dorn, S. Eschbaumer, D. Ives, L. Mehrgan, M. Meyer, J. Stegmeier
Hawaii-2RG close to prefect wrt basic parameters noise, QE, darkcurrent
comparison of 4 methods to determine conversion gain persistence of HgCdTe Hawaii-2RG arrays mitigation strategy to reduce persistence method to measure persistence in darkness
Introduction
Noise comparison H2RG #119 / H2RG #184
H2RG #119 (X-Shooter)25.3 erms on IR active pixels7.7 erms on reference pixels
H2RG #184 (KMOS)6,9 erms on IR active pixels5.8 erms on reference pixels
bond pad contact resistance improved noise reduced from 25.3 to 6.9 erms
Readout noise < 10 erms for DCS on 5 new science arrays (KMOS and SPHERE)
Noise of KMOS arrays with Fowler sampling
Reduce noise with multiple nodestructive sampling
Noise 2.2 erms for 32 Fowler pairs
Noise map of crossdispersed spectrum slit open / warm instrument shutter closed integration time = 600s ( 903 nondestructive readouts) limited by shot noise of photon background which is dominated
by scattered light of K-band (5E-2 e/s/pixel)
J order 26
K order 11
Dark current of c =2.5 m HgCdTe
Crossdispersed echelle spectrum with slit closed
Cut levels :
0 - 5E-3 e/s/pixelat T=81K , Vbias=1V
1
2
dark current outside optical field:4.2 E-4 e/s/pixel
dark current in J1.3 E-3 e/s/pixel
T=81K, VBIAS = 1V
Dark current of c =2.5 m HgCdTe
Dark current versus temperature
Quantum efficiency high over the entire sensitive range of the array
Measurement at optical wavelengths pending
IPC with single pixel reset
uniformly illuminate arraywith high flux integration time 1 s
Use guide mode of Hawaii-2RG muxguide window size 1x1
Reset single pixel before readoutintegration time < 500s
uniformly illuminate arraywith high flux integration time 1 s
Use guide mode of Hawaii-2RG muxguide window size 1x1
Reset single pixel before readoutintegration time < 500s
Observe capacitive coupling on next neighbors
IPC with single pixel reset
recent improvements of IPC
•improvements in multiplexer layout resulted in reduction of coupling coefficient :
#184=1.7% #226=1.4%
H2RG #184
H2RG #226
Conversion gain and single pixel reset
Assuming a
• Ori Fox method of classical propagation of errors: used also by Teledyne
Cij is covariancebetween pixels i and j
true variance
Conversion gain and single pixel reset
Assuming a
• Ori Fox method of classical propagation of errors: used also by Teledyne
Cij is covariancebetween pixels i and j
• =0.17 : correction factor 1+8 =1.13
true variance
Conversion gain and single pixel reset
variance versus signal:
2.26 e/ADU single pixel reset IPC
correction1.96 e/ADU
Conversion gain from integrated autocorreolation
variance versus signal:
2.26 e/ADU single pixel reset IPC
correction1.96 e/ADU
integrated autocorrelation versus signal: 2.04 e/ADU
Conversion gain by capacitance comparison method
Vreset
Dsub
CextC0
Reset SFD
V
Relais
Detector
01
,2,1, )( CVCVVnpixel
iinextextext
dc level drifton external capacitor
sum of signal of all pixels
slope = C0/Cext
Conversion gain by capacitance comparison method
Vreset
Dsub
CextC0
Reset SFD
V
Relais
Detector
Ccryo
Ccryo difficult to estimate without risk for detector:Ccryo includes capacitances of cable, preamplifier board and wirebond ceramics
Ceramic capacitors on HAWAII-2RG wirebond ceramics show strong temperature dependence:T=296 K C=1FT=77 K C=276 nF
Conversion gain by capacitance comparison method
0
0
0
1extcryoext CCCC
C0=32.8 fF Ccryo=394 nF
C0/e=205e/mV (in our setup: 1.89e/ADU)
Vreset
Dsub
CextC0
Reset SFD
V
Relais
Detector
Ccryo
0 measured with Cext removed
comparison of methods to obtain conversion gain
method conversion gain C
[e/ADU] [e/mV] [fF]
variance versus signal 2.26 245 39.2
integrated autocorrelation (Moore et al) 2.04 221 35.4
single pixel reset (Fox et al) 1.96 212 34.0
capacitance comparison 1.89 205 32.8
Remarks:
•first three methods are stochastic (rely on noise measurement)•single pixel reset measures coupling coefficient but assumes only coupling to next neighbors•capacitance comparison is direct and robust method
taking into account coupling to all pixels taking into account cable capacitance and cold ceramic capacitors at detector
Persistence: X-Shooter as test bench
•with slit closed: instrumental background: 4.2E-4 e/s/pixel
•ideal for persistence tests
Persistence: lamp on
DIT=1.65s slit open ThAr lamp
Persistence: slit closed
first dark exposure with DIT=128s after 2048s exposure with open slit
Persistence versus stimulus
Persistence of first 2 min. dark exposure is ~6.3e-4 of stimulus
Persistence at different wavelengths
Persistence almost the same at
=1.07 m
and =2.2 m
Persistence model of Roger Smith
p npn -junction
charge trapped when location of trapbecomes undepleted and is released in next dark exposure
traps populated when exposed to mobileelectrons and holes
p npn -junction
charge trapped when location of trapbecomes undepleted and is released in next dark exposure
traps populated when exposed to mobileelectrons and holes
Persistence model of Roger Smith
Mitigation of persistence: global reset detrapping
p npn -junction
charge trapped when location of trapbecomes undepleted and is released in next dark exposure
traps populated when exposed to mobileelectrons and holes
keep global reset switch closed after science exposure allow de-trapping of charge
28
•Slit open
Mitigation of persistence: global reset detrapping
29
•First 2 minute dark exposure without global reset de-trapping
Mitigation of persistence: global reset detrapping
30
•Keep reset switch of all pixels permanently closed with global reset for 128 s at the end of bright exposure
to force depletion width to stay wide avoiding population of traps• de-trapping time is 128 s• close slit and return to normal operating mode taking dark exposures•Persistence in first dark exposure reduced by factor of 9
•First 2 minute dark exposure with global reset de-trapping
Mitigation of persistence: global reset detrapping
31
• if reset closed before switching on bright source and kept closed until slit closed again persistence is zero•global reset is an electronic shutter which protects detector from persistence while exposed to bright illumination
•First 2 minute dark exposure with global reset always closed during bright exposure
Mitigation of persistence: global reset detrapping
32
• In first 2 minute dark exposure intensity of persistence is reduced by a factor of 9 with global reset de-trapping•Duration of detrapping 128 s
without global reset de-trapping
with global reset de-trapping
reset always closed during bright exposure
Mitigation of persistence: global reset detrapping
Method to measure persistence in darkness
hypothesis:
persistence is generated by the change of the voltage across pn junction of pixel diode
instead of using light to shrink depletion region reduce bias voltage in selected area of array using the window mode of the Hawaii-2RG multiplexer and the global rest
outside window normal operation of the array
Persistence electrical /optical
•Generated with bias change in selected area using global reset
•Generated with light source
Persistence electrical /optical
• red diamonds: persistence generated with light source on /off•black triangles: in selected area using global reset persistence generated with bias low / high• decay with similar time constants
Persistence measured in darkness
•Measure persistence of all 3 KMOS detectors in one gouniformity, cosmetics, dark current. readout noise, persistence
•GL scientific mosaic mount •128 channel cryo-preamps , flex boardsand vacuum connectors
Persistence measured in darkness
•Mosaic test facility:no windowno opticsdetector covered by black plateflux < 1E-3 e/s/pixel
Persistence measured in darkness
•integration time 120 sec•operating temperature 66K•generated with bias change in darkness on selected area using global reset
Persistence versus time
•persistence lasts for > 1500s•persistence is device dependent•array # 184 is better
Persistence versus temperature
•persistence is device dependent•persistence of devices #211 and #212 has a maximum at T=66K•persistence of device #184 does not have this temperature maximum
Persistence versus detrapping time
•peristence is decreases with increasing detrapping time
Persistence versus duration of illumination
•persistence increases when detector is exposed to the bright source for a longer time
Persistence versus signal intensity
•persistence increases with increasing stimulus•bias = DSUB – VRESET VRESET=0.5V
increasing signal ( brighter light source)
Global reset de-trapping: on sky test
• after global reset detrapping vertical stripes in first difference images of two 1200s exposures • intensity of stripes in first difference ~ 0.03 e/s/pixel
Global reset de-trapping: on sky test
• profile of vertical stripes in difference images of two 1200s exposures• intensity of stripes in first difference ~ 0.03 e/s/pixel
Global reset de-trapping: on sky test
• vertical stripes in first difference images of two 1200s exposures• intensity of stripes in first difference ~ 0.03 e/s/pixel• stripes located at start of fast shift register
Global reset de-trapping: on sky test
• keep clocks running during global reset detrapping• no vertical stripes in first difference images of two 600s exposures
Global reset de-trapping: on sky test
• profile of vertical stripes in difference images of two 1200s exposures• intensity of stripes in first difference ~ 0.03 e/s/pixel•intensity of stripes in second difference ~ negligible•to be further investigated
Global reset de-trapping: on sky test
• automatic flexure compensation: line intensity 45000 e/s/pixel
Global reset de-trapping: on sky test
• automatic flexure compensation: line intensity 45000 e/s/pixel•First 1500 s dark exposure : no persistence !
readout noise improved by a factor of 2 on new H2RG (6.9 erms single DCS)
shot noise limited operation achieved in cross dispersed spectrometer in J PTF corrected by single pixel reset factor yields conversion gain
which agrees with capacitance comparison method within 3% persistence model confirmed by experiment:
traps at edge of valence and conduction bands persistence is a consequence of changing bias voltage no persistence is expected with CTIA since bias voltage constant
new method to measure persistence in darkness global reset detrapping successfully tested on sky without residuals
Conclusion
HAWK-I first light
Tarantula Nebula
THE END
TLI: Threshold Limited Integration
Multiple nondestructive readouts scheme
Set saturation level If signal exceeds
saturation level, readout not used to calculate slope of integration ramp
Extrapolate signal to DIT In effect different
integration times for pixels exceeding saturation level
TLI: Threshold Limited Integration
Multiple nondestructive readouts scheme
Set saturation level If signal exceeds
saturation level, readout not used to calculate slope of integration ramp
Extrapolate signal to DIT In effect different
integration times for pixels exceeding saturation level
TLI: Threshold Limited Integration
Multiple nondestructive readouts scheme
Set saturation level If signal exceeds saturation
level, readout not used to calculate slope of integration ramp
Extrapolate signal to DIT In effect different integration
times for pixels exceeding saturation level
Signal can be >> 107 e Gain of
> 2 orders of magnitude in dynamic range
TLI: Threshold Limited Integration
Multiple nondestructive readouts scheme
Set saturation level If signal exceeds
saturation level, readout not used to calculate slope of integration ramp
Extrapolate signal to DIT In effect different
integration times for pixels exceeding saturation level
Full well 1E5 electrons With TLI it is possible to
integrate > 1E7 electrons
with TLI
without TLI
Saturation level
Trapping model
P
N
-- - -
-
++++
+
next dark exp.
(small bias reduction)
The released charge reduces the bias voltage. persistence
- - -
+ + +
-- - -
-
++++
+
high flux signal
(low bias)
As signal accumulates the depletion width is reduced. Traps newly exposed to charge can capture some mobile carriers.
Trapped holes
Trapped electrons
- - -
+ + +
++++
+
-- - -
-
reset
(large reverse bias)
At “reset” the wider depletion region is restored, but trapped charge stays behind.
Depleted
Mobile electrons
Mobile holes
dark idle
(large reverse bias)
All traps have released their charge in depletion region
++++
+
-- - -
-
-+
R.Smith, SPIE 7021-22, Marseille 2008-06-24
Dark current versus temperature
generation-recombination limited above 110K
Idark exp(-Teff/T)
Idark exp(-Egap/(1.7 KT)) surface leakage and
tunneling below 100K
1.7E-3 e/s/pixel