Lecture 5 - Photodetectors and Noise

Many different types of detector can be used to measure the amount of electro-
magnetic radiation that falls upon them. In this lecture I shall begin by giving a general
introduction to the various popular detectors, pointing out their specific applications.
I shall then look in more detail at three classes of detector which are of major impor-
tance in communication systems, photomultiplier tubes, semiconducting devices such as
pn-junctions and p-i-n diodes, and avalanche photodiodes. I shall also discuss the tech-
nology behind CCD detectors.

1 Thermal Detectors
In thermal detectors, radiation is absorbed in the active element, and this changes the
temperature of the device. The change in temperature then gives rise to a change in
some measurable physical property of the device, such as its resistance, which can then
be measured electronically. These detectors are generally insensitive to the wavelength
of the incident radiation, but usually have large areas to increase their sensitivity. These
factors make them relatively unimportant in the field of optoelectronic sensors, although
they do form the basis of most laser power meters. Typical examples are:

1.1 Thermoelectric detectors

These rely on the principle of the thermocouple (the Seebeck effect) whereby the heating
of one junction between two dissimilar metals relative to the other causes a current to flow
round a circuit that is proportional to the temperature difference between the junctions.
Their usefulness lies in their simplicity and their rugged construction.

1.2 The Bolometer

Here the incident radiation heats a fine wire or metallic strip causing a change in its elec-
trical resistance, which is then measured in a variety of ways. The performance of these
devices depends upon the temperature coefficient of resistance of the material. Thermis-
tors are similar, but rely on semiconducting elements such as oxides of manganese, cobalt
or nickel. Carbon resistance bolometers cooled to liquid helium temperature (4.2 K) have
proved successful in far-IR astronomy where very sensitive detectors are required.

1
1.3 Pneumatic Detectors
The receiving element in a pneumatic detector is placed in an airtight chamber. Radi-
ation falling on the element causes the air temperature inside the chamber to rise and
hence the air pressure to increase. This pressure change is then detected in a number of
ways, including the use of a flexible mirror, whose distortion moves a beam of light in a
controlled and measurable fashion (known as a Golay cell detector).

1.4 Pyroelectric Detectors

The incident radiation is absorbed in a ferroelectric material which has molecules with a
permanent electric dipole moment. The net electric polarisation present is temperature
dependent (below the a critical temperature characteristic of the material). When config-
ured as a capacitor in a circuit, changes in polarisation induced by temperature changes
in the material can be detected by measuring the change in the charge on the plates of
the capacitor. Pyroelectric detectors can be made with response times in the nanosecond
region and with a wavelength response extending out to ∼ 100 µm. They have proved
very useful as low cost, robust IR detectors in such uses as fire detection and intruder
alarms.

2 Photon Devices - Photoemissive Detectors

When photons with a wavelength less than a critical value are absorbed by a metal,
electrons are emitted (the photoelectric or photoemissive effect). The electrons must
have energies greater than the surface work function φ, and are emitted with kinetic
energies up to
E = hν − eφ.

No electrons will be emitted if hν < eφ (or λ0 > hc/eφ). This is illustrated in figure 5.1.
If inelastic collisions are present the only a fraction of the electrons with this energy may
escape. The ratio of the number of emitted electrons to the number of absorbed photons
is called the quantum yield or quantum efficiency.
Pure metals have low quantum efficiencies (∼ 0.1%), and high work functions.
Two types of photocathode are used in practice. The older classical types consist of
thin evaporated layers of alkali metals. They are often designated by an ‘S’ number.
They may be regarded as semiconductors where electrons must gain an energy of at least

2
Figure 5.1 Electron energy level diagram at a metal/vacuum interface illustrating
the photoelectric effect. To escape from inside the metal, an electron must gain
at least an energy eϕ, where ϕ is the work function.

Figure 5.2 The effective electron affinity of a semiconductor may be altered if band
bending takes place at the surface. Here holes trapped in surface states cause a
surface depletion region to be formed. The potential drop Vs across the depletion
region reduces the effective electron affinity for bulk electrons from χ to χ - Vs.
Eg + χ (where Eg is the energy gap and χ is the electron affinity) to escape from the
surface. The material NaKCsSb is a typical multi-alkali photocathode, known as S20, it
has Eg = 1.0 eV and χ ≈ 0.4 eV, and hence it should have a threshold photon energy of
∼ 1.4 eV, which is indeed close to that observed.
Newer photocathodes reduce the effective value of χ by making use of band bending
and are known as negative electron affinity or (NEA) types. The band bending is illus-
trated in figure 5.2, and is found to occur when there are states within the energy gap at
the semiconductor surface. A depletion region is formed, very similar to that found in a
p–n junction. The potential drop Vs across the depletion region leads to band bending as
shown. The effective electron affinity then becomes

χeff = χ − Vs .

If Vs > χ then we have a negative electron affinity and the effective work function for
bulk electrons is just Eg . In practice, NEA photocathodes are formed by evaporating
caesium oxide onto the semiconductor surface. Photocathodes using GaAs operate at
quite high quantum efficiencies right up to the energy gap of GaAs at ∼ 0.9 µm. The
quantum efficiencies of a number of the more common photocathode materials are shown
as a function of wavelength in figure 5.3.
Vacuum photodiodes are the simplest devices based on the above process. They
require large optical powers to produce relatively modest outputs, and therefore a much
more common and important device is the photomultiplier. In the photomultiplier the
photoelectrons are accelerated towards a series of electrodes (called dynodes) which are
maintained at successively higher potentials with respect to the cathode. On striking a
dynode surface, each electron causes the emission of several secondary electrons, which in
turn are accelerated towards the next dynode and continue the multiplication process. If,
for example, 5 electrons are emitted at each dynode, and there are 9 dynodes, then a gain
of G = 59 = 1.95 × 106 is possible. Thus photomultipliers offer significant amplification.
Four of the most common types of photomultiplier dynode configurations are illustrated
in figure 5.4.
Photomultipliers or PMTs are highly sensitive devices, but have the disadvan-
tage of being relatively bulky. It is possible to obtain PMTs with risetimes ∼ 1 ns. If
monochromatic radiation with a vacuum wavelength of λ0 and power Pλ is incident on a
photocathode then the number of photons Np incident per second is given by
Pλ P λ λ0
Np = =
hc/λ0 hc

3
 Figure 5.3 Quantum efficiency versus wavelength for a number of the more common
photocathode materials.

Figure 5.4 Dynode structures of four common types of photomultiplier: (a) venetian blind;
(b) box and grid; (c) linear focused and (d) circular cage focused. Typical electron trajectories
through the system are also shown.
and hence if the quantum efficiency of the photocathode is η then the current flowing
through the external circuit, iλ , is given by
ηePλ λ0
iλ =
hc
and the responsivity Rλ is given by
iλ ηeλ0
Rλ = = .
Pλ hc
Even when no radiation is falling on the photocathode, thermionic emission gives
rise to a dark current which often constitutes the main source of noise in PMTs. The
thermionic emission current iT for a cathode at temperature T of area A and work function
φ is given by the Richardson–Dushman equation (see Blakemore, Solid State Physics)
à !
eφ
iT = aAT 2 exp −
kB T
where a is a constant, which for pure metals is 1.2 × 10−6 Am−2 K−2 . Dark current can
therefore be reduced by cooling the PMT, typically to temperatures ∼ 240 K. Another
important source of noise in PMTs is shot noise. This arises from the random generation
and flow of charge carriers, and is related to the discrete nature of the electronic charge.
An exhaustive treatment of noise is beyond this lectures course, but Yariv’s book Optical
Electronics in Modern Communications has a complete discussion of the topic. I shall be
using some of the more well-known results of noise analysis without formal proof. Thus
the magnitude of the r.m.s. current fluctuations ∆is with frequencies between f and
f + ∆f is given by
∆is = (2ie∆f )1/2

where i is the current flowing (the sum of the dark current and signal current). Therefore
the minimum detectable signal power in the presence of a thermionic dark current iT is
given by
(2iT e∆f )1/2
Pmin = .
Rλ
Thus it is now possible to make an estimate of the minimum signal power. For a cathode
of area 1000 mm2 , a work function φ of 1.25 eV and a cathode temperature of 300 K
(giving kB T /e = 0.025 eV) we have for iT

iT = 1.2 × 106 × 10−3 × (300)2 exp(−1.25/0.025) = 2 × 10−14 A

At a wavelength of 500 nm and a quantum efficiency η of 0.25 the responsivity is

ηeλ0 0.25 × 1.6 × 10−19 × 0.5 × 10−6
Rλ = = −34 8
= 0.1 AW−1
hc 6.6 × 10 × 3 × 10
4
Fianlly, if we take a bandwidth of 1 Hz then the minimum detectable signal power is

(2 × 2 × 10−14 × 1.6 × 10−19 × 1)1/2

Pmin = = 8 × 10−16 W
0.1
This is equivalent to around 2000 photons. Thus if the photocathode is cooled, individual
photons can be detected by these devices.
Two other important noise sources for PMT tubes are multiplication noise,
which comes from the dynode chain, but is usually a small contribution to the total noise
( 20%) and Johnson or Nyquist noise. Johnson noise arises because of the thermal
agitation of the charge carriers within a conductor; the random nature of this motion
results in a fluctuating voltage appearing across the conductor. The r.m.s. value of this
voltage ∆VJ having frequency components between f and f + ∆f across a resistance R
at a temperature T is given by

∆VJ = (4kB T R∆f )1/2

In a PMT such noise will appear across a load resistor at the anode. In practice, Johnson
noise is also much smaller then the dark-current shot noise.
Image intensifiers are another important class of devices based on the PMT princi-
ple. They are intended to boost very low intensity optical images to the point where they
become useful. They can also act as wavelength down-converters in that they can convert
near-IR radiation into visible radiation. There are many different designs in use, but one
of the most popular recently is called the microchannel plate image intensifier or
MCP. The device is illustrated in figure 5.5. The MCP consists of a slab of insulator
(∼ 500 µm thick) with a high density of smaller diameter (∼ 15 µm) holes or channels in
it. The inner faces of the channel are made slightly conducting and a potential (∼ 1 kV)
is applied between opposite faces of the slab as shown. Electrons entering one of the
channels are accelerated down it and strike the walls soon after entering. As in the PMT,
secondary electrons are generated by the impact and the process is repeated down the
channel as illustrated. The electron shower then strikes the phosphor and a pixellated
image is produced.

3 Photon Devices - Photoconductive Detectors

In these semiconductor based detectors light whose energy is greater than that of the
bandgap causes the generation of electron-hole pairs. As long as the electron remains in

5
Figure 5.5 Electron multiplication in a microchannel plate image intensifier. When the primary
electrons enter the channel and strike the walls, secondary electrons are emitted that in turn
generate further secondaries. The channel thus acts as a miniature photomultiplier tube. On
emerging from the channel the electrons generate light by striking a phosphor screen.

Figure 5.6 (a) Geometry of a slab of photoconductive material. The slab of length L, width W
and thickness D has electrodes on opposite faces; radiation falls onto the upper face.
(b) Photoconductor bias circuit. Changes in the resistance of the photoconductor cause
changes in the voltage appearing across RL.
the conduction band, the conductivity of the semiconductor will be increased. This is
the phenomenon of photoconductivity, which is the basic mechanism operative in pho-
toconductive detectors. A typical detector geometry and circuit is shown in figure 5.6.
Any change in the conductivity of the detector results in an increased flow of current
around the circuit, which will increase the potential across the load resistor RL . Figure
5.7 shows the variation of the optical absorption coefficient α for several semiconductors.
The irradiance I(x) at a point a distance x into a semiconductor can thus be written as

I(x) = I(0) exp(−αx)

This means that the fraction of the incident irradiance which is actually absorbed in the
semiconductor is
η = (1 − R) × [1 − exp(−αD)]

where R is the intensity reflection coefficient and D is the thickness of the slab. The
wavelength response of photoconductive detectors should, in theory increase linearly up
to λg , although the variation of α with wavelength means that η often decreases faster at
short wavelengths.
In the visible region of the spectrum these devices are often used as light meters
in cameras (using CdS and CdSe detectors). They usually have high gains but poor
response times (∼ 50 ms). In the near IR lead sulphide (PbS) is often used as a detector
material, with a useful wavelength response from 1 µm to 3.4 µm. Typical response times
are ∼ 200 µs. Another important photoconductive detector in the far-IR is mercury
cadmium telluride (Hgx Cd1−x Te). This compund may be thought of as an alloy composed
of the semimetal HgTe and the semiconductor CdTe. Consequently, depending upon the
composition of the alloy, a semiconductor may be formed with a bandgap varying between
zero and 1.6 eV. Real detectors have peak sensitivities in the range 5 − 14 µm, and need
to be cooled. The detectors are useful as they can detect black body radiation from
live tissue. Early television cameras used devices called vidicons, which are electron
tubes where photoconductive elements form the target, which is then read by a scanning
electron gun. These have now largely been superseded by CCD detectors, as will be
discussed later.

6
Figure 5.7 Variation of the optical absorption coefficient a with wavelength for
several semiconductor materials.

Figure 5.8 Electron energy diagram showing the generation and subsequent
separation of an electron-hole pair by photon absorption within the depletion
region of a p-n junction.
4 Photon Devices - Junction Detectors
All junction detectors, as the name implies, are based in some way on the humble p–n
junction. When a p–n junction is formed in a semiconductor material, a region depleted
of mobile charge carriers is created that has a high internal electric field across it. If an
electron–hole pair is generated by photon absorption within this region the electric field
separates the charges as illustrated in figure 5.8.
The charge separation may be detected in three distinct ways. First, if the device
is left on open circuit an externally measurable potential will appear between the p and
n regions; this is the photovoltaic mode of operation. Secondly, in the photoamperic
mode a very low external resistance is connected between the external contacts and a
photogenerated current flows through it. Finally, the most usual way to operate the device
is in the photoconductive mode where a reverse bias is applied across the junction and the
resulting current flow through an external load resistor measured. The load resistor in
this case need not be as small as in the photoamperic mode. The photoconductive mode is
preferred in practical detectors because the external current flowing in the detector circuit
is directly proportional to the incident light irradiance over a relatively wide dynamic
range. It also offers a more rapid response than other modes. One disadvantage is the
presence of a dark current which, as in the PMT, gives rise to shot noise and limits the
ultimate sensitivity of the device. The responsivity in this mode can be written as
ηeλ ηλ
Rλ = = AW−1
hc 1.24
If we use a typical value for η of 0.8 and a wavelength of 0.85 µm in silicon, we get a value
for Rλ near to the peak of Rλ ∼ 0.55 AW−1 .
A typical structure for a p–n junction diode is shown in figure 5.9. The resulting
electric field variation within the depletion regions is shown in figure 5.10. For efficient
detection the electron–hole pairs should be generated within the depletion region. At
short wavelengths, where the absorption coefficient is relatively high, they will be gen-
erated close to the surface. Consequently, to achieve a good short wavelength response,
the p+ region should be made as thin as possible. Conversely, at the upper wavelength
range of the detector the absorption coefficient will be relatively small and a wide deple-
tion region is required for high detection efficiency. This is controlled by the reverse bias
voltage, which is limited by the diode breakdown voltage. Detection efficiency may also
be improved by providing an anti-reflection coating to the front surface of the detector,
consisting of a λ/4 thick layer of SiO2 .

7
 Figure 5.9 Typical silicon photodiode structure for photoconductive operation. A junction
is formed between heavily doped p-type material (p+) and fairly lightly doped n-type
material so that the depletion region extends well into the n-type material. The p+ layer
is made fairly thin. Metallic contacts can be made directly to the p+ material, but to obtain
an ohmic contact to the n-type material an intermediate n+ layer must be formed.

Figure 5.10 Electric field distribution within the p+ - n junction diode shown above, assuming
an abrupt diode structure.
 The bandwidth of the photodiode is related to the circuit used. There are three
factors which contribute to the circuit’s response time: (a) the time taken by the photo-
generated carriers to drift across the depletion layer, (b) the time taken by the carriers
to diffuse through the respective p or n materials to the contacts and (c) the RC time of
the electrical circuit.
For (a), the time depends upon both the drift velocity of the carriers and the
width of the layer. In Si the maximum velocity for the carriers is ∼ 5 × 104 ms−1 and
a typical depletion width is ∼ 2.5 µm, thus the response time becomes ∼ 50 ps and
the bandwidth is ∼ 20 GHz. For (b) the diffusion time must also be minimised. Since
diffusion velocities are generally lower than drift velocities the p and n regions must be
made quite narrow. Thicknesses of 0.5 µm are normally used. Finally, for (c), a typical
p–n junction capacitance is ∼ 25 pF, so for a 20 GHz bandwidth, we require
1
R< = 0.32 Ω.
2πf C
This is a small value for the resistance, necessitating the use of a carefully constructed
amplifier. If the more usual 50 Ω value is chosen, then the bandwidth becomes ∼ 100 MHz.
This is more than sufficient for the vast majority of applications (but these do not include
high-bit-rate optical communications!).

4.1 Noise in p–n junctions

What of noise in p–n junctions? The shot noise is given by the expression quoted earlier
for the PMT as
in = (2iD e∆f )1/2

where iD is the current flowing through the circuit. This will be the sum of two compo-
nents
iD = iP + id

where iP is the photogenerated current and id is the dark current, which is due to leak-
age of thermally generated charge carriers across the junction. It depends both on the
temperature of the device and on its area. The noise performance of the diode itself
is usually expressed in the form of the noise equivalent power (NEP), which is the
optical power input which would produce a photocurrent equal to the dark current, id .
Again, by analogy with the PMT we have
hν(2eid ∆f )1/2
NEP = .
eη

8
It is quoted as a power ‘per root hertz’, in other words, for unit bandwidth. A typial
value for the NEP for a Si photodiode is ∼ 10−14 WHz−1 , corresponding to dark currents
of ∼ 1 nA.
To get the whole noise picture the load resistor and the rest of the circuit discussed
above needs to be included. Thus the Johnson noise current in the resistor will be
à !1/2
4kB T ∆f
iR =
R
and hence the total circuit noise will be
à !1/2
1/2 4kB T ∆f
iN = (2id e∆f ) +
R
The relative values of these two terms depend upon the relative values of id and R. As
R is typically 50 Ω for fast circuits, and id ∼ 1 nA, thermal noise normally dominates.

4.2 p-i-n diodes

As a detector the p-i-n photodiode has a number of advantages over the p–n photodiode.
A p-i-n diode is a p–n junction with an intrinsic (usually lightly doped) layer snadwiched
between the p and n layers. It may be operated under a variety of bias conditions. The
energy band diagram, charge distribution and electric field distribution for a reverse-
biased p-i-n diode are illustrated in figure 5.11. This structure serves to extend the
width of the region supporting an electric field, in effect widening the depletion layer.
Photodiodes with p-i-n structure offer the following advantages:
(a) Increasing the width of the depletion layer of the device (where the generated carriers
can be transported by drift) increases the area available for capturing photons.
(b) Increasing the width of the depletion layer reduces the junction capacitance and
thereby the RC time constant. On the other hand, the transit time increases with the
width of the depletion layer.
(c) Reducing the ratio between the diffusion length and the drift length of the device
results in a greater proportion of the generated current being carried by the faster drift
process.
Response times in tens of ps, corresponding to bandwidths of ∼ 50 GHz, are
achievable. The responsivity of two commercially available p-i-n photodiodes is compared
with that of an ideal device in figure 5.12. It is interesting to note that the responsivity
maximum occurs for wavelengths substantially shorter than the bandgap wavelength.
This is because Si is an indirect gap material.

9
 Figure 5.11 The p-i-n photodiode structure, energy diagram, charge distribution
and electric field distribution. The device can be illuminated either perpendicularly
or parallel to the junction.

Figure 5.12 Responsivity versus wavelength for ideal and commercially available
Si p-i-n photodiodes.
 In fibre-optic communication systems, the wavelengths of interest are 1.3 µm and
1.5 µm. Thus there is a demand for photodiodes which have high sensitivities and excep-
tionally wide bandwidths. These are generally made from heterojunctions of semicon-
ductor alloys, generally grown on InP substrates. These include the ternary compound
Inx Ga1−x As and the quaternary compound Inx Ga1−x Asy P1−y . A typical structure is
shown in figure 5.13, where an i layer of In0.53 Ga0.47 As is sandwiched between a p+ layer
of InP and an n layer of InP. Because the radiation has to pass through a layer of InP
which has a bandgap wavelength of 0.92 µm, then no radiation below this wavelength will
be detected. A further advantage of this structure is that since no carriers of interest will
be generated in either the surface p+ layer or the lower n layer there will be no diffusion
carriers to the junction from outside the depletion region, thus improving the response
time. From figure 5.7 it can be seen that the absorption coefficient of In0.53 Ga0.47 As is
much higher near the bandedge than that of Si. This is because In0.53 Ga0.47 As is a di-
rect gap material. This means that thinner active layers (∼ 2 µm) can be used, again
improving the response speed. Typical values are ∼ 30 ps.

4.3 Avalanche photodiodes

An avalanche photodiode (APD) operates by converting each detected photon into a
cascade of moving carrier pairs. Weak light can then produce a current that is sufficient
to be readily detected by the electronics following the APD. Te device is a strongly
reverse-biased photodiode in which the junction electric field is large; the charge carriers
therefore accelerate, acquiring enough energy to excite new carriers by the process of
impact ionization. The process is illustrated in figure 5.14. A photon is absorbed
at point 1, creating and electron–hole pair. The electron accelerates under the strong
electric field, thereby increasing its energy with respect to the bottom of the conduction
band. Should the electron, whilst being scattered by phonons, acquire an energy larger
than Eg at any time, it has the opportunity to generate a second electron–hole pair by
impact ionization (point 2). The two electrons then accelerate, and may be the source
for further impact ionization. Holes also have a chance of impact ionizing, as shown at
point 3. Further details of this device are available in Physics of Semiconductor Devices
by Sze.

10
 Figure 5.13 Basic structure of a p-i-n heterojunction InGaAs detector.

Figure 5.14 Schematic representation of the multiplication process in an APD.

5 CCD detector arrays
In the 1970s the charge coupled device or CCD was invented at Bell Laboratories
in the U.S.A. The basic building block of this is the metal–oxide–semiconductor (MOS)
capacitor. This is formed by growing a layer of silicon dioxide (SiO2 ) on a p-type silicon
substrate; a metal electrode is then evaporated on top of the oxide layer (shown schemat-
ically in figure 5.15). The metal electrode is known as the gate and is biased positively
with respect to the silicon. Photogenerated electron–hole pairs within the silicon will
be separated, with the electrons being attracted to the surface of the Si under the gate,
where they will remain whilst the gate voltage is positive. The electrons are effectively
trapped within a potential well formed under the gate contact. The amount of trapped
charge will be proportional to the total integrated light flux falling onto the device during
the measurement period.
The problem then becomes how to ‘read out’ this charge sequentially along a line of
such detectors? This can be done by passing the charge from detector to detector. There
are several ways of achieving this in practice. I shall illustrate the basic idea by considering
a CCD where the gate potentials are supplied from three voltage lines (L1 , L2 , L3 ) each
one being connected to every third electrode (G1 , G2 , G3 ), as illustrated in figure 5.16.
Suppose that initially the potential of L1 is at some positive value Vg , whilst L1 and L2
are at zero potential. Photogenerated charge will be trapped under the G1 electrodes
in proportion to the amount of light falling on these elements (figure 5.16 (b)). After a
suitable integration time, the charge may be moved along the chain of MOS capacitors
by applying a repeated sequence of potentials to the gate supply lines. Thus, suppose a
voltage Vg is applied to L2 whilst maintaining L1 at Vg ; the charge initially under L1 will
now be shared between G1 and G2 (Figure 5.16 (c)). The potential of L1 is then reduced
to zero, and so all the charge that was initially under G1 is now under G2 (Figure 5.16
(d)). Thus it is possible to progressively move the charge along the line of MOS capacitors
from left to right. At the end of the line, the amount of charge arriving as a function of
time then provides a sequential scan of the G1 detector outputs.
Two-dimensional arrays based on the above one-dimensional design are possible,
known as frame transfer devices. Here the transfer registers feed into a readout register
running down the edge of the device. The contents of each line are read out in sequence
into the readout register so that the signal appearing at the end of the register represents
a line-by-line scan of the image. One problem with the above scheme is that a new light

11
Figure 5.15 MOS capacitor structure. When the gate is biased positively, photogenerated
electron-hole pairs become separated and the electrons then become trapped at the
SiO2 - Si boundary beneath the gate electrode.
Figure 5.16 (a) Basic CCD array composed of a line of MOS capacitors. (b) The initial
charge distribution within the potential wells when G1 has a positive voltage applied (Vg) and
G2 = G3 = 0. (c) G1 = G2 = Vg, G3 = 0 and the charge has now spread out under both
G1 and G2. (d) G1 = 0, G2 = Vg and G3 = 0. The charge that was initially under G1 has now
moved to the right to be under G2. Note that no charge should be generated under the gates
G2 and G3; these elements are therefore screened from incident light.

Figure 5.17 Interline transfer scheme for reading out photogenerated charge from a
two-dimensional CCD image sensor. Charge flow is indicated by arrows and areas
shielded from radiation by shading.
scan cannot be carried out until all of the charge has been transferred along the array. A
faster scanning rate my be achieved by making each array twice as long as before, with
the second half of the array being shielded from the incident radiation. Then, after a
charge image has been built up it is moved along each array into the shielded section
where it can be stored until it can be transferred into the readout register. Readout thus
take place whilst a new image is being built up. This is illustrated in figure 5.17.
Figure 5.18 shows a picture of a typical array. Pixel sizes can vary between ∼ 10 µm
and 50 µm, and arrays with over 1000 pixels are common. In modern CCDs almost no
charge is lost in the transfer along the registers. Thermal excitations are a source of
noise, however, and can limit the sensitivity of the device. In practice a CCD made with
transparent polysilicon electrodes is now the preferred design. This can limit the quantum
efficiency and so can be illuminated from below by employing a thinned substrate. The
quantum efficiency can then be up to 0.9, and if cooled to 200 K the device can have
almost no thermal noise. The only remaining source of noise then becomes that due to
the readout electronics and A/D converter. This can be as low as one electron per pixel.
Thus CCDs used for long integration exposures are the most sensitive detectors available
today and routinely find application in astronomy, Raman and luminescence spectroscopy,
and low light-level surveillance. At higher frame transfer rates (∼ 30 frames per second)
they are used in commercial TV and video cameras, as well as in still digital cameras.
When colour images are required, digital cameras use a masking technique, where
red, green and blue filters are placed over successive pixels. This means that the effective
resolution of the CCD is reduced by a factor of 9. As colour changes in images are
usually gradual, the camera electronics then interpolate the colour information from
adjacent pixels to reconstruct an image with the full pixel resolution of the CCD. This is
illustrated in figure 5.19.

12
Figure 5.18 (a) Corner of an array of Schottky-barrier photodiodes. Each pixel is
40µm × 80µm in size. (b) Cross-section of a single pixel in the CCD array. The light
shield prevents the generation of photocarriers in the CCD transfer gate and buried
channel. The guard ring minimises dark current spikes and the channel stop confines
the signal charge in the lateral direction.
Figure 5.19 Schematic of colour CCD used in commercial digital cameras.