Infrared Detector & ASIC Technology Markus Loose STScI, May 8, 2014.

Infrared Detector & ASIC TechnologyMarkus Loose

STScI, May 8, 2014

STScI Lecture 2

Outline

May 08, 2014

• CMOS-based Detectors (infrared)– General Properties of Solid State Detectors– CMOS-based Multiplexers– Examples

• Control ASICs– General Description– Example: SIDECAR ASIC

– Preamp, Biases, ADCs– Noise performance and issues

• Conclusion

STScI Lecture 3

CMOS-based Infrared Sensors

May 08, 2014

( CMOS: Complimentary Metal Oxide Silicon )

4

The Ideal Detector

• Detect 100% of photons

• Each photon detected as a delta function

• Large number of pixels

• Time tag for each photon

• Measure photon wavelength

• Measure photon polarization

Oct 15, 2009 Scientific Detector Workshop, Garching, Germany

Up to 98% quantum efficiency

One electron for each photon gfdg

~1,400 million pixels (>109)

No - framing detectors

No – defined by filter

No – defined by filter

Plus Read Noise and other “Features”

STScI Lecture 5May 08, 2014

Hybrid Imager Architecture

Image of indium bump array in comparison to

human hair(credit: Laser Focus World)

STScI Lecture 6

Energy of a Photon

May 08, 2014

Wavelength (m) Energy (eV) Band

0.3 4.13 UV

0.5 2.48 Vis

0.7 1.77 Vis

1.0 1.24 NIR

2.5 0.50 SWIR

5.0 0.25 MWIR

10.0 0.12 LWIR

20.0 0.06 VLWIR

• Energy of photons is measured in electron-volts (eV)• eV = energy that an electron gets when it “falls” through a 1 Volt field.

STScI Lecture 7

An Electron-Volt (eV) is extremely small

May 08, 2014

1 eV = 1.6 • 10-19 J (J = joule)

1 J = N • m = kg • m • sec-2 • m

1 kg raised 1 meter = 9.8 J = 6.1 • 1019 eV

• The energy of a photon is VERY small– The energy of a SWIR (2.5 m) photon is 0.5 eV

• Drop a peanut M&M® candy from a height of 2 inches– Energy is equal to 6 x 1015 eV (a peanut M&M® is ~2 g)– This is equal to 1.2 x 1016 SWIR photons

• 1 million x 1 million x 12,000• The number of photons that will be detected in ~1 million images from

the James Webb Space Telescope (JWST)• A 2-inch peanut M&M® drop is more energy than will be detected during

the entire 5-10 year lifetime of the JWST !

STScI Lecture 8

Photon Detection

May 08, 2014

Conduction Band

Valence Band

Eg

For an electron to be excited from theconduction band to the valence band

h > Eg

h = Planck constant (6.6310-34 Joule•sec)n = frequency of light (cycles/sec) = /c Eg = energy gap of material (electron-volts)

Material Name Symbol Eg (eV) c (m)

Silicon Si 1.12 1.1

Indium-Gallium-Arsenide InGaAs 0.73 – 0.48 1.68 – 2.6

Mercury-Cadmium-Teluride HgCdTe 1.00 – 0.07 1.24 – 18

Indium Antimonide InSb 0.23 5.5

Arsenic doped Silicon Si:As 0.05 25

c = 1.238 / Eg (eV)


CCD Approach CMOS Approach

PixelCharge generation &charge integration

Charge generation, charge integration &

charge-to-voltage conversion

+

PhotodiodePhotodiode Amplifier

Array ReadoutCharge transfer

from pixel to pixel

Multiplexing of pixel voltages: Successively

connect amplifiers to common bus

Sensor Output Output amplifier performs

charge-to-voltage conversion

Various options possible:- no further circuitry (analog out)- add. amplifiers (analog output)- A/D conversion (digital output)

General CMOS Detector Concept

STScI Lecture

General Architecture of CMOS-Based Image Sensors

May 08, 201410

Pixel Array

Horizontal Scanner / Column

Buffers

Ver

tica

l S

can

ner

fo

r R

ow

Sel

ecti

on

Control &

TimingLogic

(optional)

Bias Generation

& DACs(optional)

Analog Amplification

A/D conversion (optional)

Analog Output

Digital Output

STScI Lecture 11

IR Multiplexer Pixel Architecture

May 08, 2014

Vdd

amp drain voltage

OutputDetectorSubstrate

PhotvoltaicDetector

STScI Lecture 12


May 08, 2014

Vdd

amp drain voltage


PhotvoltaicDetector

Vreset

reset voltage

Reset

STScI Lecture 13


May 08, 2014


PhotvoltaicDetector

Vreset

reset voltage

Reset

Vdd

amp drain voltage

Enable “Clock” (red)

“Bias voltage” (purple)


Special Scanning Techniques in CMOS

• Different scanning methods are available to reduce the number of pixels being read:– Allows for higher frame rate or lower pixel rate (reduction in noise)– Can reduce power consumption due to reduced data

Random Read• Random access

(read or reset) of certain pixels

• Selective reset of saturated pixels

• Fast reads of selected pixels

Subsampling• Skipping of certain

pixels/rows when reading the array

• Used to obtain higher frame rates on full-field images

Windowing• Reading of one or

multiple rectangular subwindows

• Used to achieve higher frame rates (e.g. AO, guiding)

Binning• Combining several

pixels into larger super pixels

• Used to achieve lower noise and higher frame rates

* Binning is typically less efficient in CMOS than in CCDs.

STScI Lecture 15

Possible Reset Schemes for HxRG

May 08, 2014

Reticle

Stitched CMOS Sensor

Pixel by pixel reset Line by line reset Global reset

Full field

Window

STScI Lecture 16

Astronomy Application: Guiding

May 08, 201416

• Special windowing can be used to perform full-field science integration in parallel with fast window reads.Þ Simultaneous guide operation and science

data capture within the same detector.

Full field row Window Full field row

Full field row

Window Window

Full field row Full field row

• Two methods possible:– Interleaved reading of full-field and window

• No scanning restrictions or crosstalk issues• Overhead reduces full-field frame rate

– Parallel reading of full-field and window• Requires additional output channel• Parallel read may cause crosstalk or conflict• No overhead maintains maximum full-field frame

rate

STScI Lecture 17

Guide Mode Demonstration

May 08, 2014

• Video shows a small window in the center, read frequently for guiding, while the full field is read slowly in the background


H A W A I I - 2 R G

H A W A I I - 2 R GHgCdTe Astronomy Wide Area Infrared Imager

with 2k2 Resolution, Reference pixels and Guide Mode

2k x 2k HAWAII-2RG with HyViSI detector 2 x 2 Mosaic of HAWAII-2RG detectors

Teledyne HAWAII-2RG Hybrid Detector Array


5 MHz column buffersfast normal shift register + logic

glow

and

cro

ssta

lk s

hiel

d

g low and crosstalk shield

Additional row of reference pixels for diagnostic purposes

2048 x 2048 pixel array(2040 x 2040 sensitive pixels)

4 rows and columns containing reference pixels

4 rows and columns containing reference pixels

serial interface

clock buffers

fast guide shift register + logic

Slow

gui

de s

hift

regi

ster

+ lo

gic

Slow

nor

mal

shi

ft re

gist

er +

logi

c

decoders for horizontal start and stop address

I/O Pads & output buffers

deco

ders

for v

ertic

al s

tart

and

sto

p ad

dres

s

• All pads located on one side (top)

• Approx. 110 doubled I/O pads (probing and bonding)

• Three-side close buttable

• 18 µm pixels

• Total dimensions: 39 x 40.5 mm²

H2RG Block Diagram

20May 08, 2014 STScI Lecture

JWST - James Webb Space Telescope 15 Teledyne 2K×2K infrared arrays on board (~63 million pixels)

• International collaboration• 6.5 meter primary mirror and tennis court size sunshield• 2018 launch on Ariane 5 rocket• L2 orbit (2.4 million km from Earth)

Two 2x2 mosaicsof SWIR 2Kx2K

Two individual MWIR 2Kx2K

NIRCam(Near Infrared Camera)

• Wide field imager• Studies morphology of objects

and structure of the universe• U. Arizona / Lockheed Martin

• Spectrograph• Measures chemical composition,

temperature and velocity• European Space Agency / NASA

NIRSpec(Near Infrared Spectrograph)

1x2 mosaic of MWIR 2Kx2K

FGS / NIRISS (Fine Guidance Sensors /Near-Infrared Imager and

Slitless Spectrograph)

• Acquisition and guiding• Images guide stars for telescope

stabilization• Canadian Space Agency

3 individual MWIR 2Kx2K


Control ASICs

22

How to Operate an Image Sensor?

• Sensor/Detector requires:– DC bias and reference voltages

• Set properties like offset, bandwidth, reverse detector bias• Voltages need to be programmable to allow optimal performance• Very low noise to not contribute to the read noise (< 10µV noise)

– Clocks/Digital control signals• Responsible for controlling the readout timing and sensor configuration• Configurable timing required

– Video Signal Readout• If digital output sensor, job is mostly done. Simply route to FPGA for

data acquisition and storage• If analog output sensor:

– Amplify/buffer analog signal– Perform analog-to-digital conversion, then route digital data to FPGA

May 08, 2014 STScI Lecture

STScI Lecture 23

ASIC as Control Electronics

May 08, 2014

Replace this

with this!

1% volume1% power1% hassle


Digital ControlMicrocontroller for Clock Generation

and Signal ProcessingBias

Generator

Amplification and A/D

Conversion

Data Memory

Program Memory

Data Memory

Digital I/O

Interface

SIDECAR

Exte

rnal

El

ectr

onic

s

Mul

tiple

xer,

e.g.

HAW

AII-

2RG

analog mux out

bias voltages

clocksmain clock

data in

data out

synchron.

Digital Generic I/O

System for Image Digitization, Enhancement, Control And Retrieval

SIDECAR ASIC Architecture


SIDECAR ASIC Features• 36 analog input channels, each channel provides:

– 500 kHz A/D conversion with 16 bit resolution – 10 MHz A/D conversion with 12 bit resolution – gain = 0 dB …. 27 dB in steps of 3 dB– optional low-pass filter with programmable cutoff– optional internal current source (as source follower load)

• 20 analog output channels, each channel provides:– programmable output voltage and driver strength– programmable current source or current sink– internal reference generation (bandgap or vdd)

• 32 digital I/O channels to generate clock patterns, each channel provides:– input / output / high-ohmic– selectable output driver strength and polarity– pattern generator (16 bit pattern) independent of microcontroller– programmable delay (1ns - 250µs)

• 16 bit low-power microprocessor core (single event upset proof)– responsible for timing generation and data processing– 16 kwords program memory (32 kByte) and 8 kwords data memory (16 kByte)– 36 kwords ADC data memory, 24 bit per word (108 kByte)– additional array processor for adding, shifting and multiplying on all 36 data channels in parallel

(e.g. on-chip CDS, leaky memory or other data processing tasks)


SIDECAR Operating a HAWAII-2RG / 1RG

PCPCSoftware for

SIDECAR Control and Data Capture

PCI Cardor USB

interface

Vreset

Dsub

Clock

000 w

384 w

0cd w

HAWAII-2RG HAWAII-2RG SIDECAR ASICSIDECAR ASIC

Analog Supply

Data In

Data Out

Master Clock

Digital Supply

3.3V 3.3V

Analog Video

Biases

Power Supply

Serial Interface

Clocks

Only 7 lines needed to operate the SIDECAR ASIC

in base configuration (3 signal & 4 power lines)

The SIDECAR ASIC provides all 27 signals required to operate the

HAWAII-2RG

The microcontroller driven SIDECAR ASIC generates

all biases & clocks and digitizes the analog video

outputs

Sensor Chip Assembly

Sensor Chip Assembly

Inside the dewar at cryogenic temperatures

May 08, 2014 STScI Lecture 27

SIDECAR ASIC Flight Package for JWST

• Ceramic board with ASIC die and decoupling caps

• Invar box with top and bottom lid• Two 37-pin MDM connectors

– FPE-to-ASIC connection– ASIC-to-SCA connection

• Qualified to NASA Technology Readiness Level 6 (TRL-6)

• 15 mW power when reading out of four ports in parallel, with 16 bit digitization at 100 kHz per port.

FPE side

SCA side

SIDECAR

Comparison:

LGA Package

ACS (HST)


Missions Employing SIDECAR ASICs• James Webb Space Telescope

– NIRCam, NIRSpec, FGS/NIRISS instruments– H2RG IR detectors, T = 38K (ASIC), planned launch in 2018

• Hubble Space Telescope– ACS (Advanced Camera for Surveys)– CCD detector, T = 300K (ASIC), launched in 2009

• Landsat Data Continuity Mission– TIRS (Thermal InfraRed Sensor) instrument– QWIP detector, T = 300K (ASIC), launched in 2013

• OSIRIS-REx Asteroid Mission– OVIRS (OSIRIS-REx Visible and IR Spectrometer) instrument– H1RG IR detector, T = 300K (ASIC), planned launch in 2016

• Euclid Mission– NISP (Near IR Spectrometer Photometer) instrument– H2RG IR detector, T = ~140K (ASIC), planned launch in 2020

• MOSFIRE (Multi-Object Spectrometer For Infra-Red Exploration)– H2RG IR detector, T < 120K (ASIC), deployed at the Keck Telescope

• FourStar Wide Field Infrared Camera– H2RG IR detector, T < 120K (ASIC), deployed at the Magellan Baade 6.5m Telescope

JWST

HSTLDCM

OSIRIS-RExEuclid


Pre-Amplifier Block Diagram

• Capacitor Feedback Design• Gain programmable by

setting Cin and Cfb• Gain = Cin/Cfb

• Low pass filter with programmable cutoff

S3

S1

S1

V1

V2

V3

V4

S3

S4

S4

V1

V2

Vref mid

bypass

S5 S5

SAR_En SAR ADC output

LPF

Pipe_En Pipeline ADC output

S2

S2

Cin

Cin

Cfb

S6

VpremidrefS6

VpremidrefCfb S6

S6


Preamp Drift and MitigationData taken as 512 x 64 frames for efficiency, Gain = 4

Drift

kTC row noise

kTC removed (CCD mode)

σ= 52 ADU

σ= 2.6 ADU

σ= 13.9 ADU


Noise Reduction by Using Multiple ADC Channels

1 ADC

2 ADCs

4 ADCs

6 ADCs

8 ADCs

• PreAmp inputs shorted to ground (lowest noise signal in order to be dominated by ADC noise)

• PreAmp gain set to 4 (12 dB)• Noise measured by using multiple

preamp and ADC channels in parallel (1, 2, 4, 6, and 8)

• Noise reduces almost as the square root of the number of channels used

0

0.5

1

1.5

2

2.5

3

0 1 2 3 4 5 6 7 8 9

Number of averaged ADC channels

Ou

tpu

t N

ois

e [

AD

U]

0

5

10

15

20

25

30

35

40

45

Inp

ut

Re

ferr

ed

No

ise

[µ

V]

STScI Lecture 32

Bias Generator Block Diagram

• SIDECAR has 20 Channels• Each Channel provides

programmable voltage and current sources

• Noise is caused mostly by buffer (1/f noise of MOS transistors)

• Drive strength of buffer can be adjusted to modulate the bandwidth

• Feedback compensation can be adjusted for stability

• Buffer can be configured for single or dual stage operation

• => Tuning required for optimal noise performance

May 08, 2014


Bias Generator Noise• Bias output 1 routed back into PreAmp• PreAmp gain set to 22 (27 dB)• Use 4 ADCs in parallel to reduce PreAmp & ADC noise• Noise on bias without filtering is about 35µV (11.6

ADU)• Noise can be reduced by RC filtering to less than 5µV

0

2

4

6

8

10

12

14

0.001 0.01 0.1 1 10 100 1000

RC filter time contant [ms]

Ou

tpu

t N

ois

e [A

DU

]

0

5

10

15

20

25

30

35

40

45

Inp

ut

Ref

erre

d N

ois

e [µ

V]

total noise

bias noise

Bias noise as a function of RC filter time constant

PreAmp & ADC noise floor

Unfiltered Noise of Bias Output 1

Filtered Noise of Bias Output 1 (tRC = 360 ms)


Noise Power Spectrum of the Bias Outputs

-60

-50

-40

-30

-20

-10

0

10

20

0.1 1 10 100 1000 10000 100000

Frequency [Hz]

No

ise

[d

Bµ

V/r

oo

tHz]

-60

-50

-40

-30

-20

-10

0

10

20

0.1 1 10 100 1000 10000 100000

Frequency [Hz]

No

ise

[d

Bµ

V/r

oo

tHz]

FFT of temporal noise measurement with RC filter set to tRC= 3 µs

FFT of temporal noise measurement with RC filter set to tRC= 3 ms


-60

-50

-40

-30

-20

-10

0

10

20

0.1 1 10 100 1000 10000 100000

Frequency [Hz]

No

ise

[d

Bµ

V/r

oo

tHz]

-60

-50

-40

-30

-20

-10

0

10

20

0.1 1 10 100 1000 10000 100000

Frequency [Hz]

No

ise

[d

Bµ

V/r

oo

tHz]

FFT of temporal noise measurement with RC filter set to tRC= 360ms

FFT of temporal noise measurement with grounded PreAmp inputs (i.e. noise floor)

Noise Power Spectrum of the Bias Outputs, Part 2


Analog-to-Digital Conversion

• Quantization noise of an ADC is (1/√12) Least Significant Bit = 0.289 LSB

• Typically set gain of amplifier chain so that quantization noise is much less than readout noise. If readout noise is 4 electrons, set gain so that LSB equals ~2 electrons

• 16 bit ADC is most commonly used in astronomy. At ~2 electrons per ADU (analog to digital unit), or LSB, full well of a 16 bit ADC will be ~130,000 electrons; good match to the typical full well of a CCD or Short-Wave IR detector of 100,000 electrons.

Highly exaggerated quantization noise“Don’t do this at home”


Differential Non-Linearity (DNL)

DNL = (VD+1 – VD) / VLSB-Ideal – 1

Code 10 is missingDNL = -1

Code 10 is reducedDNL = -0.5

Code 100 is increasedDNL = +1

• DNL describes the distance of an ADC code from its adjacent code. • It is measured as a change in input voltage magnitude, and then converted to

number of Least Significant Bits (LSBs).


Integral Non-Linearity (INL)

• INL describes the deviation of the ADC transfer function from a straight line • It can be computed as the integral of the DNL, and is expressed in LSB

INL = (VD – VZero) / VLSB-Ideal – D


16-bit ADC Linearity

Output Code

DN

L [ L

SB ]

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

0 10000 20000 30000 40000 50000 60000

DNL

Output Code

INL

[ LSB

]

-4

-3

-2

-1

0

1

2

3

4

0 10000 20000 30000 40000 50000 60000

INL

• Differential Non-Linearity: < ± 0.3 LSB• Integral Non-Linearity: < ± 0.2 LSB• Temporal Noise: 2.7 LSB


ADC Linearity Pitfalls

Optimal Vcm

Vcm off by 80 mV

Vcm off by 160 mV

• Differential ADC is composed of 2 separate single-ended ADCs

– If one of the two ADCs saturates before the second one does, the transfer slope changes by 2

Slope change

Slope change

• Requires careful adjustment of the ADC reference and common mode voltages

• Simultaneous optimal tuning for all channels does not exist due to component mismatch

• Avoid lower and upper end of ADC for science


1/F Noise in NIRSpec/JWST

• Traditional CDS • Optimal CDS

σCDS ~ 18 e- rms σCDS ~ 8 e- rms


IRS^2 Noise Reduction ModeExample: NIRSpec 1000s Dark Up-the-Ramp Signal

• Improved Reference Sampling & Subtraction (IRS^2) is a method to utilized reference pixels in a more efficient way

– In every output channel, read reference pixels from the top or bottom or rows in-between the regular science pixels (e.g. read 4 reference pixels every 16 science pixels)

– Use Fourier analysis to determine the frequency-dependent correlation between signal and reference pixels, and subtract the reference pixel signal accordingly

Nominal (no IRS^2) IRS^2 IRS^2 (using real pixels as reference)


ACS 1/f NoiseBias Frame without correction(superbias subtracted)

Bias Frame with correction(superbias subtracted)

STScI Lecture 44

Conclusion

• Infrared Detectors (Image Sensors)– Hybrid design: Detector material bump-bonded to readout chip– Different detector materials possible

• HgCdTe for astronomy: lowest dark current and adjustable bandgap– Flexible readout options like guide mode or single-pixel reset

• Control ASICs– Provide all functions to operate the detector

• Clocking, Biasing, A/D conversion– Single-chip solution in contrast to discrete electronics

• Lower power, space, weight– Can run cryogenically (next to the cooled detector)– Performance Improvements desired (lower noise, no artifacts)

May 08, 2014

Infrared Detector & ASIC Technology Markus Loose STScI, May 8, 2014.

Documents