Scientific Detectors in the

Visible and Infra-Red

April 2021

A three day intermediate-level course covering the principles

of photon detectors as well as the construction, operation and
characterisation of scientific imaging systems.

Simon Tulloch
[email protected]
Course Presenter:
Dr. Simon Tulloch has worked for 20 years in observatory operations and
scientific instrument development. He has built cryogenic cameras at the
Isaac Newton Group, the European Southern Observatory and as a
private contractor at the University Complutense. He has a Ph.D. from
the University of Sheffield titled “Astronomical Spectroscopy with
Electron Multiplying CCDS” and has extensive publications in the fields
of low noise detectors, signal processing, detector characterisation and
image-persistence correction.

At whom the course is oriented:

The course is oriented at physicists, engineers and astronomers who are
working on instrumental projects that involve the use of CCDs, CMOS
sensors or IR imaging arrays. The course would also be a valuable
introduction for those wishing to become scientific detector specialists.

Previous grounding:
A basic grounding in physics is required. Some knowledge of electronics
would be desirable but is not essential. The course contains very little
mathematics.

Course Summary:
The course begins with an introduction to the solid-state physics
underlying the operation of modern scientific visible and near infra-red
detectors. The basic principles of photo-diodes, CMOS and CCD
detectors are then outlined. Visible wavelength detectors are then
described in more detail. It is explained how CCD detectors have been
progressively improved over the last 50 years and how their
performance is now closely approaching that of an ideal detector. The
current state of alternative CMOS design detectors and how they
compare to CCDs is also described. Photon counting detectors such as
the Electron Multiplying CCD and the Leonardo Saphira are also covered.
The course then moves to longer wavelengths, covering the use of
Teledyne Hawaii HxRG Near-Infra-Red (NIR) detectors and the Raytheon
Aquarius Mid-Infra-Red (MIR) detectors. The various noise sources that
degrade scientific observations with any detector and the way they
combine statistically are examined in depth. The course finishes with a
study of essential detector related topics such as characterisation, cryo-
vacuum systems, thermal control and controller electronics.
 SPECIFIC TOPICS COVERED
Underlying Physics:
Semiconductors and the Periodic Table
The PN junction and its sensitivity to light.
The electromagnetic spectrum and atmospheric transmission.
Understanding detector performance data.
Statistics of photons.

CCDs:
Physical structure.
CCD clock sequence design.
Output amplifier variants.
Noise sources: read noise, dark current, reset noise and photon noise.
Boosting performance with backside illumination and thinning.
Deep-depletion CCDs and Fringing.
Anti-reflective coatings and Quantum Efficiency.
Hi-Rho and Bulk-silicon CCDs.
Electron-multiplying CCDs and their advantages/disadvantages.
Traps.
High-speed operation.
Low-noise preamplifiers.
Analogue and Digital video processor design.
CCD mosaics.
CCDs in Space.
ESD precautions.

CMOS sensors:
Physical structure.
Global and Line Reset.
Unit cell circuitry: 3T,4T and 5T designs.
Next generation CMOS
Hybrid CMOS.

Infra-red sensors:
Mercury Cadmium Tellurium as a sensor material.
The Teledyne HxRG detectors.
The Saphira detector.
Blocked Impurity Band detectors.
Fowler Sampling.
Persistence.
Subtle image defects.

Cryo-vacuum systems:
The Planck curve and the need for cooling.
Cooling methods.
Thermal sensing and control.
Vacuum materials.

Detector Characterisation:
Measuring Quantum Efficiency.
Measuring gain and noise.
Flat-fielding.
Use of Iron-55 X-rays.
Mosaic Coplanarity.
Sample Material
 Semiconductor Junctions
Creates a built-in electric field within the device. Strong
field is present in a “depletion region” which is free of
charge carriers. The resultant structure is known as a
“Diode”.

Electrical
Symbol
Depletion
region
p
Electrical potential

The band
structure is
distorted
across the
junction.
n

Photoelectric effect
Photon absorption lifts an electron from the Valence Band to the
Si Si Si
Conduction Band. It can then move freely through the Silicon
lattice. A Hole is left behind in the Valence band which is also
free to jump from atom to atom. Both Electrons and Holes allow
current to flow.
Si Si Si
Covalent
bonds due to
pairing of valence
Si
Photon Si Si
electrons.
absorption
Electron

Hole

Creates charge 3D Structure

carrier pair.
 The PhotoDiode
Only ever operated with zero or reverse bias.
Vout

Simple resistive Op-amp solution.

load: non-linear Maintains zero bias
performance. across photodiode.
Preferred method.
Electrical Illumination
Symbol

Traps
Of fundamental importance in
detector manufacture.
Surface Traps:
Dangling covalent bonds at
interface between the semi-
conductor and the surface
oxide. Passivated in Silicon
using Hydrogen (not 100%
effective).
Bulk Traps:
Due to impurities, crystal
boundaries and lattice
dislocations.

Surface trap density usually much higher than bulk.

Signal charge should be kept away from surface if
possible.
Johnson Noise
Thermal motion of electrons in a resistor generates a noise voltage.

Noise White noise

Bandwidth Hz
temperature Kelvin
Boltzmann´s constant
1.38 x 10-23 J K-1

A 1kOhm resistor
at a temperature
of 300K
measured across
a 10kHz
bandwidth
generates
Johnson noise of
First explained at Bell labs
400nV RMS.

Flicker Noise
Typical MOSFET amplifier noise spectrum (Log:Log scale)
 CCD Analogy

CCDs : Backside Thinning

Necessary step for scientific-performance sensors
Basic “webcam” quality CCDs
Scientific quality “thinned,
are frontside illuminated Remove wafer using acid etch, flip the backside illuminated” CCD
CCD and illuminate from backside

Thick CCD Thinned CCD

D

Extremely expensive but

detects almost all photons.
Cheap to manufacture but most Particularly good in the
photons are either reflected or blue.
absorbed in the electrodes.

(refractive index of Silicon =4)

CCD : Silicon implementation

Bulk or Hi-Rho CCDs

For CCDs greater than 50Pm thick it is not possible to use epitaxial
Silicon. Instead the entire CCD is manufactured using the bulk wafer.
Excellent red-end response. Requires “Back Bias”.

Teledyne E2V CCD250

Hi-Rho CCD used in LSST.

temp=178K

(Needed to fully
deplete device.) Paul Jorden Teledyne E2V
CCD: Charge accumulation

Photocharge stored well

away from the surface of the
Silicon thus avoiding surface
traps and keeping dark
current low.

CCD: Charge Transfer

Achieved by modulation of electrode voltages
Photodiode Arrays
At least one extra transistor is actually required (for row select).
The 3T Unit Cell. row-reset bus
transistor bias voltages

reset transistor output transistor

(switch) (source follower buffer)

row-select bus

column bus
row select transistor
(switch)

Line reset

Line
Address
Line Reset
decoder
Read-Reset-Read mode (Rolling Shutter)
Gives fastest possible readout whilst retaining high observing efficiency.
Good for high-background astronomy but for dynamic scenes causes
distortion as the maximum frame rate is approached.
EXPOSURE Line 1

EXPOSURE Line 2

Each line has same exposure

EXPOSURE Line 3
time but they are not
simultaneous.
EXPOSURE Line 4

EXPOSURE Line 5

EXPOSURE Line 6

Read Up the Ramp

Taking multiple measurements (samples) during exposure is the favored
method for low-background astronomy. Has the advantage of much lower
read noise.
row 1
row 2
A least-square fit
row 3 must be done in
row 4 real-time on the
data points.
row 5
row 6

Exposure time
Exposure time is
time between first
and last samples.

Samples
 Main problem with current
CMOS: lack of red response
Consequence of thin epitaxial Silicon, at most 18um, typically much less.

4T design Absorption length in Silicon

STI STI

PPD=Pinned photodiode TG=Transfer Gate OD=Output Drain

STI=Shallow trench isolation RG=Reset Gate RD=Reset Drain

The Stefanov solution

Collaboration between Open University (UK) and Teledyne-E2V opens way to
CCD-like QE in CMOS sensors. Uses “DDE” implants to isolate p-wells.

“Design and performance of a Pinned Photodiode CMOS Image Sensor using Reverse Substrate Bias”
Open Research Online, Konstantin D. Stefanov, 2018
 Cutting edge CMOS
Teledyne E2V Hi-Rho Capella
2048 x 2048 x 10Pm pixels
12e- read noise
90%
90ke- FWC
80%
20 fps 2.5x QE
70%
60% 2x QE
50%
Epi thickness
40%
30% 8um 20um 35um
20%
Uses up to 35mm thick 10%
epitaxial fully depleted 0%
400 500 600 700 800 900 1000 1100
by application of back bias
technique. Gives the red
response required by modern
astronomy.

Teledyne Princeton Instruments

LACera
Wafer-scale backside illuminated CMOS.

up to 8k x 8k x 10Pm pixels
sub-electron read noise
>80ke- FWC
10 fps for 8k variant
90% peak QE
 Hybridisation
Solution to low fill factor experienced by monolithic CMOS.
Photodiode material doesn’t have to be Silicon.

Two halves of detector sandwich are mechanically bonded

in a hydraulic press.

Blocked Impurity Band (BIB) Detectors

Also known as Impurity Band Conductors (IBC). Used extensively to detect
IR radiation between 5Pm and 28Pm.

Key Points:

Unipolar device i.e. no junction.

n.

Photoconductive operation
i.e. electrical field that sweeps
apart photo-generated charge e
pairs is externally applied.

Photons are absorbed by the

dopant atoms not the Silicon.

To be used on ELT-METIS, JWST MIRI. Currently used on VLT-VISIR and Spitzer.

 Photon Statistics
Pixel

8 Photons 6 Photons 6 Photons 11 Photons 10 Photons

Source of
constant Exposure 5 Exposure 4 Exposure 3 Exposure 2 Exposure 1
intensity

Photon “bunching” gives a fundamental uncertainty in how accurately we can measur

intensity of an optical signal. Variously described as :

“Photon Noise” , “Shot Noise” , “Poisson Noise”

Any scientific imaging system should aim to be “Photon Noise Limited”

Histogram shapes
Flatfield Histograms described by the
“Poisson Distribution”
Flatfield
sequence =0.3 P

=1.0
n
Increasing

=3.0
Probability P of a pixel
receiving n photons
=10 if mean illumination is
M.

=30 As M increases this

looks more and
more like a Gaussian
distribution.
Generic Cryostat Design
All cold elements of the camera need to be encased in a vacuum vessel (the
“Cryostat”) to reduce thermal load and avoid condensation.
For the E2V Silicon Carbide package (4k x 4k or 6k x 6k pixel devices).
Main design elements are shown below.

Crown Section
Contains detector, thermal servo (heater and temperature sensor), signal
distribution PCB and all associated cabling.
Liquid Nitrogen (LN2)
Radiative and conductive load on the LN2 tank need to be minimised to
maximise the “hold time” (This should be at least 16 hours).

Each litre of LN2 produces 750litres of N2 gas.

1W of cooling power consumes 22mL of LN2 per hour

Radiative Transfer
Copper LN2 can Radiative load P
0.03 Aluminium pressure
vessel
77K
0.05
298K
200mm

Stefan´s constant
Emissivity surface 1
Emissivity surface 2
Area
200mm Temperature surface 1
Temperature surface 2

P = 1.6Watts
5.6 x 10-8 W m-2 K-4
Emissivity of Ice =0.98. If Copper vessel becomes
coated in a film of ice, P = 4.1W
 Pumping Theory
Example 1 : Viscous flow regime.

Approx. 25% loss of effective

pump speed.

Example 2 : Molecular flow regime.

In Example 2, using a 4cm diameter

pipe would increase effective pump
=10 x loss of effective speed to:
pump speed.

Vacuum Outgassing
Release of adsorbed material from cryostat walls. Principally water and air.
Units of outgassing are mBar L s-1 cm-2 . Falls exponentially with time.

Typical values after 1 hour of pumping:

Surface area=1884 cm2
Stainless Steel Volume = 6.3 L

Aluminium
200mm

Example:
Pump a 6.3L Aluminium pressure vessel
for 1 hour. Expected outgassing rate after
pump is switched off=
200mm

Can be accelerated by baking