1. European Southern Observatory
Performance of large format HgCdTe and InSb arrays for low background applications
European Southern Observatory
G. Finger, H.Mehrgan, M.Meyer, A.F.M. Moorwood,G. Nicolini and J. Stegmeier

2. Outline Limitations of HgCdTe 1Kx1K (Hawaii I)
photon shotnoise of multiplexer glow
image degradation by reset anomaly
Limitations of InSb 1Kx1K (Aladdin II)
thermal drift for long integrations
readout noise
readout speed and crosstalk
Future projects
HgCdTe 2Kx2K (Hawaii II) with lc=1.9 mm for NIRMOS
InSb 1Kx1K mosaic for CRIRES
128 channel IRACE data acquisition system

3. Noise reduction by multiple nondestructive readouts
Follow-up-the-ramp sampling (FUR): at equidistant time intervals nondestructive readouts least squares fit: slope of integration ramp
FUR
Fowler: nondestructive readouts at start and at end of ramp least squares fit: slope of integration ramp for n>>1:
Fowler

4. Comparison of Fowler and Follow-up-the-ramp sampling
white, bandlimited noise number of nondestructive readouts >> 1
Follow-up-the-ramp sampling is best for continuous nondestructive sampling no glow:
Fowler is best if number of nondestructive readouts has to be severely limited because of glow:

5. Hawaii 1Kx1K MCT array LPE HgCdTe on Al2O3 substrate (PACE1)
cutoff wavelength lc = 2.5 mm
pixel size 18 mm
installed in SOFI at NTT (3.5 m) and in ISAAC at VLT (8m)

6. Cryogenic operational amplifier
On-chip source follower not used
internal bus of detector directly connected to +5V with 200 KW load resistor
linear CMOS opamp (TI LinCMOS TCL2274) 15nV/sqrt(Hz)
contribution to read noise
2 electrons rms
for double correlated sampling at frame rate of 1 Hz
reduced glow
Gain=1+2RF/RG

7. Readout Noise versus number of nondestructive readouts
Follow-up-the-ramp sampling:
number of readouts n
proportional to integration time
s2mult/s2dc=6(n-1)/(n(n+1))
frame rate 1Hz, 64 nd samples
integration time 67 sec
rms noise 6.3 electrons
darkcurrent too small for shot noise
Full frame, 24 mV/ADU
Shot noise

8. Darkcurrent of 1Kx1K MCT
sample integration ramp for 3 hours every 15 minutes
Darkcurrent 49 e / hour at T = 65K

9. Shot noise of shift register glow1 2
shift register glow localized at four points on edges (1,4) and in corners (2,3)
Remark: arcs are due to reflection of light emitted by
glow centers 4 3

10. Shot noise of shift register glow1 2
Cover glow centers and paint detector environment black
glow measured close to glow center in center of the array 4 3

11. Shift register glow induced detector signal
Close to glow center: 49 electrons/frame
Center of array: 0.2 electrons/frame

12. Readout Noise versus number of nondestructive readouts
Follow-up-the-ramp sampling:
number of readouts n
proportional to integration time
s2mult/s2dc=6(n-1)/(n(n+1))
frame rate 1Hz, 64 nd samples
integration time 67 sec
rms noise 6.3 electrons
darkcurrent too small for shot noise
Full frame, 24 mV/ADU
Shot noise of
shift register glow

13. Electroluminescence of Hawaii I shift registers
Optical photons emitted by mux
324 photons/frame in 2p ster

14. Coping with shift register glow
On-chip amplifiers not used: no glow
only source of glow is addressing of pixels
reduce addressing (number of readouts)
increase number of conversions per pixel
sufficient samples to beat down noise without glow penalty
Conversion
strobes
Video signal

15. Shift register glow
With off chip cryogenic operational amplifiers photon shot noise of shift registers glow limits multiple sampling techniques for noise reduction
For continuous readout of full frame Iglow=720 electrons/hour Iglow=0.2 electrons/full frame
Solutions : 1) interleaved clocking of shift registers 2) digital filter of video signal by multiple sampling of single pixel less readouts & clocking of shift registers still sufficient samples to beat down noise

16. Reset Anomaly Dummy line without start pulse
Signal nonlinear function of time
Difference of 2 flux levels linear in time
Introduce wait state after reset
Integrate using multiple nondestructive readouts
Do not reset array after moving telescope to reduce overhead
Reset array
Wait
Integrate
Move telescope

17. Double Correlated sampling
Pixel to pixel noise
raw erms
difference 12erms
strong gradient at start of readout
Stare time 300 seconds
cut levels: 0 / 90 el.

18. Follow-up-the-ramp sampling
Pixel to pixel noise
raw erms
difference 6.9 erms
number of readouts 32
subtract row crosstalk:

19. Fowler sampling & Digital Filter
Pixel to pixel noise
raw erms
difference 3.7 erms
no gradient
row crosstalk
number of readouts 32
8 samples / pixel Stare time 300 seconds cut levels: 0 / 90 el.

20. Fowler sampling & Digital Filter & no reset
Pixel to pixel noise
raw erms
difference 3.5 erms
noiseraw=noisediffernce
no gradient
no row crosstalk number of readouts 32
8 samples / pixel Stare time 300 seconds cut levels: 0 / 90 el.

21. Uniformity of Hawaii 1Kx1K raw image on stabilized integration ramp
Introduce delay > 30 s after reset
with 32 fowler pairs uniformity of raw image 3.66 e rms pixel to pixel
no difference required : equivalent noise of 2.6 e rms
discontinuity at quadrant borders: < 4 electrons
3.66 erms
Dit = 60 s , Cuts -5 / 40 electrons

22. Aladdin 1Kx1K InSb array
backside illuminated InSb thinned by diamond turning
cutoff wavelength lc = 5.2 mm
pixel size 27 mm
foundry run on best effort basis 6 SCA’s delivered and tested 1SCA has PED, 2 SCA’s have cracks 3 SCA’s useful for instruments

23. Aladdin InSb 1Kx1K detector mount
glassfiber epoxy structure supports detector
manganin flexible board between detector at 30 K and fan-out board at 77K
32 video channels 32 reference channels

24. Aladdin #3 & #4 #5: QE H = 0.82 (ISAAC) darkcurrent = 0.017 e/s

25. Darkcurrent Measurement using open In bump bonds
Triangles: measured integration ramp
Diamonds: dead pixels
Open In bump bonds are used to monitor drifts
Squares: drift corrected integration ramp
darkcurrent at 28.5 K: e/s/pixel

26. Darkcurrent Measurement using open In bump bonds
Triangles: measured integration ramp
Open In bump bonds are used to monitor drifts
Squares: drift corrected integration ramp
darkcurrent at 25 K: e/s/pixel

27. Temperature drift of dead pixels in Aladdin array
Temperature drift: 1700 electrons / K
required temperature stability of array: 6 mK

28. Readout Noise and Gain of 1Kx1K InSb
65 e rms for DC
Readout noise 9.7 erms 128 multiple samples readout time 42.6 ms integration time 7 s
Gain 11 e / ADU

29. Analog bandwidth of Aladdin II Video Signal
Vload = 5V
Risetime 556 ns Falltime 758 ns Pixeltime 1.1 ms Frametime 52 ms
Aladdin II limited by analog bandwidth of on chip amplifier
Aladdin III faster by a factor of 2

30. Crosstalk limits readout speed
Vload = 4V
Crosstalk for frame time of 52 ms for frame time of 104 ms
Aladdin 3 improves speed by factor of 2

31. Aladdin 1024x1024 InSb Results Quantum efficiency: J 0.89 H 0.73K
L 0.68
M 0.74
read noise: 65 e rms for DC e rms with multiple
readout speed :10 frames/s crosstalk 0.5% frames/s crosstalk 7%
darkcurrent in selected region: e/s at well of 1.0E4 , T=25 K

32. True Colour Image M17 Aladdin 1024x1024 InSb
Installed in ISAAC February 2000
integration time 5 minutes
Red: PAH 3.28 mm Green: Brg mm Blue: H mm
Field: 150x150 arcsec single shot
it works

33. InSb 1Kx1K array mosaic for CIRES
Mosaic of three 1Kx1K InSb Aladdin arrays at 30K
package:
three single sockets
minimum noise : cryo-opamps at 77K

34. Hawaii x2048 HgCdTe arrays
one LPE 2048x2048 Hawaii II array with lc=2.5mm on sapphire substrate delivered in a few months
four MBE 2048x2048 Hawaii II arrays with lc=1.9mm on CdZnTe delivered in 2 years
need 128 channel IRACE Data Acquisition for NIRMOS

35. Differences between Hawaii 1Kx1K/Hawaii 2Kx2K
32 video channels & shuffled output
interleaved clocking to reduce shift register glow
each quadrant has reference unit cell to monitor drifts
package Zero Insertion Force Socket Pin Grid Array (ZIF PGA) cooled by central pins
pixel size 18 mm

36. IRACE Acquisition boards
ADC 4322: 16 bit 2MHz
4x8 = 32 channel system
1Kx1K array: frame rate 20 Hz
Aladdin
LTC1604CG: 16 bit 333KHz
16x8 = 128 channel system
4Kx4K array: frame rate 2Hz
NIRMOS

37. IRACE 128 channel system for NIRMOS
Nirmos has four 2Kx2K arrays each having 32 video channels
new ADC board developed 16 channels / board LTC1604CG: 16 bit 333KH
16x8 = 128 channel system
readout of 4Kx4K array at frame rate 2Hz Status:
16 channel ADC board tested
128 channel front-end prototype tested
Interface
to Ultrasparc
Detector front end
mounted on Instrument

38. Conclusions Limitation by photon shot noise
use Fowler sampling with limited number of readouts
apply digital filtering of video signal
Limitation by reset anomaly
switch off reset clock
pixel to pixel noise of raw frame is 3.6 erms (equivalent to 2.6 erms)
Limitation by thermal drift for long integrations
use dead pixels to monitor drift
darkcurrent of e/s measured with InSb
design reference row into mux for differential signal chain (Hawaii1R)
Limitation of readout speed
use all video channels for faster readout without increasing noise
IRACE can provide 128 channels for four Hawaii 2

39. IRATEC test camera & IRACE readout electronics
IRACE 32 channel DFE: 8x4 ADC’s for 32 video channels frame rates up to 20 Hz bandwidth of detector limits frame rate to 10 Hz
IRATEC test camera for 1Kx1K arrays Offner design 2 filter wheels bandpass / neutral density
IRACE
IRATEC

40. Cosmetics of long exposures
Warm pixels
continuous ramp
Cosmic rays
charge jump

41. Histogram of charge generated by cosmic rays
At Paranal (altitude 2600 m) pixels/hour hit
maximum charge jumps <2E4 e less than full well
correction possible by multiple sampling
exponential increase of events towards less charge
more warm pixels, but arrays improve (MBE)

42. Persistence effect
Persistence reduced by multiple reset frames prior to readout
After exposure to bright point source of 5E7 phot/sec signal proportional exp(-t/ t)
t<30 sec : t=11 sec
t> 30 sec: t=48 sec
no temperature dependence

43. Sensitivity profile of single pixel
IR fiber scanned in steps of
2.16 mm across pixel
intensity of next neighbors:
15% (along fast shift register)
6% ( along slow shift register)
for centered spot 58% of total intensity on 1 pixel