Semiconductor Detector Systems Chapter-1 PDF

1
DETECTOR SYSTEMS OVERVIEW
All semiconductor detector systems include the same basic functions. The signal
from each sensor or sensor channel in a detector array must be amplified and
processed for storage and analysis. Some functions are clearly associated with
individual circuit blocks, but frequently circuit blocks perform multiple functions.
Figure 1.1 compares a “traditional” silicon detector system for charged par-
ticle spectroscopy with an integrated detector module. The left panel shows a
room-temperature silicon detector, removed from the vacuum chamber in which
it is operated. The detector is connected to a preamplifier through a vacuum
feedthrough mounted on a vacuum flange. The pulse shaper and detector bias
Fig. 1.1. Left: A “traditional” silicon detector system showing a single readout chan-
nel. The silicon sensor is the cylindrical object at the lower right. Right: A 512-chan-
nel detector module used for particle tracking. Three 2.5 cm wide × 6 cm long sen-
sors are ganged together and read out by four integrated circuits with 128 channels
each. A low-mass ribbon cable provides data and power connections to the external
readout electronics.
1
2 DETECTOR SYSTEMS OVERVIEW
INCIDENT SENSOR PREAMPLIFIER PULSE ANALOG TO DIGITAL

RADIATION SHAPING DIGITAL DATA BUS
CONVERSION
Fig. 1.2. Basic detector functions: Radiation is absorbed in the sensor and converted
into an electrical signal. This low-level signal is integrated in a preamplifier, fed to
a pulse shaper, and then digitized for subsequent storage and analysis.
supply reside in a NIM bin and the data acquisition system is a plug-in card in a
PC, which also provides data storage and data display. In single-channel systems
digitization and data storage are often combined in a single unit, a multichannel
analyzer, whereas more complicated systems utilize a bank of external digitiz-
ers read out through a data bus (CAMAC, VME, VXI, PCI, etc.) and fed to a
computer. Such systems are still in widespread use for high-resolution x-ray and
gamma spectroscopy.
In contrast, the right panel of Figure 1.1 shows a 512-channel detector module
from a high-energy physics experiment, CDF at FermiLab. The primary func-
tion of this detector is position sensing. Multiple layers of these detectors provide
space points to reconstruct particle trajectories. The silicon sensor, the pream-
plifier, pulse shaper, digital readout control, and signal bussing are combined in
one integrated unit, a detector module. The 512 channels of analog and digital
electronics are accommodated in four integrated circuits (ICs), each about 6 mm
in size (Kleinfelder et al. 1988).
Here the term detector becomes ambiguous, especially in experiments where
the “detector” consists of several detector subsystems – tracking, calorimetry,
muon detection – which in turn consist of many individual detector modules.
Whenever ambiguities might arise we’ll refer to the device that translates the
presence of a particle to an electrical signal as a sensor.
The sequence of detector functions is illustrated in Figure 1.2 and described
below.
1.1 Sensor
The sensor converts the energy deposited by a particle (or photon) to an elec-
trical signal. This can be achieved in a variety of ways, but in this context en-
ergy is absorbed in a semiconductor, for example silicon, which produces mobile
charge carriers – electron–hole pairs. An electric field applied to the sensor sweeps
the charge carriers to electrodes, inducing an electrical current. The number of
electron–hole pairs is proportional to the absorbed energy, so by integrating the
signal current one obtains the signal charge, which is proportional to energy. As
will be shown below, the sensor pulses can be quite short (of order nanoseconds
PREAMPLIFIER 3
SENSOR PULSE SHAPER OUTPUT
TP
Fig. 1.3. In energy measurements a pulse processor typically transforms a short sen-
sor current pulse to a broader pulse with a peaking time TP .
or less) and the spatial extent of the charge cloud is small (of order microns), so
semiconductor sensors can handle very high particle rates.
1.2 Preamplifier
The signal charge can be quite small, about 50 aC (5 · 10−17 C) for 1 keV x-
rays and 4 fC (4 · 10−15 C) in a typical high-energy tracking detector, so the
sensor signal must be amplified. The magnitude of the sensor signal is subject to
statistical fluctuations, and electronic noise further “smears” the signal. These
fluctuations will be discussed in detail in Chapters 2 and 3, but at this point
we note that the sensor and preamplifier must be designed carefully to minimize
electronic noise. A critical parameter is the total capacitance in parallel with the
input, i.e. the sensor capacitance and input capacitance of the amplifier. The
signal-to-noise ratio increases with decreasing capacitance. The contribution of
electronic noise also relies critically on the next stage, the pulse shaper.
1.3 Pulse shaper

In semiconductor detector systems the primary function of the pulse shaper is
to improve the signal-to-noise ratio. Although we are considering signal pulses,
i.e. time-varying signals, the signal power is also distributed in frequency space,
quantified by the pulse’s Fourier transform. The frequency spectra of the signal
and the noise differ, so one can improve the signal-to-noise ratio by applying a
filter that tailors the frequency response to favor the signal, while attenuating
the noise. Changing the frequency response also changes the time response, the
pulse shape, so this function is called pulse shaping. As will be shown below,
improving the signal-to-noise ratio commonly implies reducing the bandwidth,
which increases the duration of the pulse (Figure 1.3).
Usually, we are not interested in measuring just one pulse, but many pulses
in succession and often at a very high rate. Too large a pulse width will lead to
pile-up of successive pulses, as shown in Figure 1.4 (left). A system that measures
the peak amplitude will give an erroneous result for the second pulse. Pile-up
can be ameliorated by reducing the pulse width, as shown in the second panel
of Figure 1.4.
Figure 1.5 shows how the pulse transformation shown in Figure 1.3 can be
accomplished. The preamplifier is configured as an integrator, which converts
AMPLITUDE
AMPLITUDE
TIME TIME
Fig. 1.4. Amplitude pile-up occurs when two pulses overlap (left). Reducing the
shaping time allows the first pulse to return to the baseline before the second pulse
arrives.
the narrow current pulse from the sensor into a step impulse with a long decay
time. A subsequent CR high-pass filter introduces the desired decay time and
an RC low-pass filter limits the bandwidth and sets the rise time. This will be
discussed in more detail in Chapter 4. Shapers can be much more complex, using
multiple integrators to improve pulse symmetry, for example. However, common
to all shapers are operations that constrain the upper frequency bound, which
sets the rise time, and the lower frequency bound, which determines the pulse
duration. When designing a system it is necessary to find a balance between
the conflicting requirements of reducing noise and increasing speed. Sometimes
minimum noise is crucial, sometimes rate capability is paramount, but usually a
compromise between the two must be found.
Although the primary measure of the signal energy is the charge, when the
pulse shape is the same for all signal magnitudes, the pulse amplitude or “pulse
height” is equivalent (hence the frequently used term “pulse height analysis”).
The pulse height spectrum is the energy spectrum. This is convenient, since
SENSOR CURRENT INTEGRATOR HIGH-PASS FILTER LOW-PASS FILTER

“DIFFERENTIATOR” “INTEGRATOR”
td ti
is
-A
e-t /td
Fig. 1.5. Components of a pulse shaping system. The signal current from the sensor
is integrated to form a step impulse with a long decay. A subsequent high-pass
filter (“differentiator”) limits the pulse width and the low-pass filter (“integrator”)
increases the rise-time to form a pulse with a smooth cusp.
DIGITIZER 5
Fig. 1.6. The theoretical “optimum” shaper output (left) and a practical pulse
(right), which attains its maximum for a measurable time.
analog-to-digital converters (ADCs) measure voltage or current amplitude. How-

ever, this imposes an additional requirement on the pulse shaper; the pulse shape
should be compatible with the digitizer. Since the digitizer has a finite response
time, the maximum signal amplitude should be maintained for a commensurate
time, so the shaper output should have a smooth maximum. This is worth re-
membering, since the filter that theoretically “optimizes” signal-to-noise ratio
for many detectors is a cusp, where the peak amplitude is attained for only
an infinitesimally short time, as shown in Figure 1.6. Clearly, determining the
amplitude of this pulse in a realistic system is fraught with uncertainties.
Sometimes the shaper is hidden; “charge sensing” ADCs are often used to
digitize short pulses from photomultiplier tubes. Internally, the input stage inte-
grates the input pulse and translates the signal charge to a voltage level, which
is held for the duration of the digitization. This is also a form of pulse shaping.
Very sophisticated shapers have been developed to optimize noise and rate capa-
bility, and also to reduce sensitivity to variations in sensor pulse shape. However,
in many applications, shapers can be quite simple. Since all amplifiers have a
limited bandwidth, every amplifier is a pulse shaper. Frequently, rather sophis-
ticated pulse shaping can be implemented by tailoring the bandwidths of the
amplifiers needed anyway to increase the signal level.
1.4 Digitizer
Analog-to-digital conversion translates a continuously varying amplitude to dis-
crete steps, each corresponding to a unique output bit pattern. First developed
for use in radiation detection, analog-to-digital conversion today is a mainstream
technique and ADCs with a wide range of characteristics are available. A con-
ceptually simple ADC is shown in Figure 1.7. The signal is fed in parallel to
a bank of comparators with monotonically increasing thresholds, provided by a
resistor voltage divider. When the pulse height exceeds a certain threshold, all
comparators with lower thresholds fire and a decoder translates the hit pattern
to a more convenient (e.g. binary) form. This technique is very fast, but requires
many comparators, as the number of comparators determines the resolution. For
example, 256 comparators can provide a full scale range of 1 V with 3.9 mV reso-
lution. In the age of vacuum tubes or discrete transistors this technique was not
very practical, as the space required for many precision comparators was pro-
Vref COMPARATORS
INPUT ENCODER
R
R
DIGITIZED
OUTPUT
R
Fig. 1.7. A conceptually simple technique for analog-to-digital conversion utilizes a

bank of comparators with increasing threshold levels. The address of the highest
level comparator responding to a signal is encoded to provide a binary output.
hibitive. However, in monolithically integrated circuits it is quite feasible, but

in practice power dissipation and chip size constrain the obtainable resolution.
Generally, increasing circuit speed requires more power, so ADCs trade off reso-
lution vs. speed. More sophisticated conversion techniques have been developed
to provide high resolution (as high as 24 bits) with fewer circuit elements, but
at the expense of conversion time. Generally, speed and resolution are opposing
parameters, as are speed and power. Although a bit pattern appears unambigu-
ous, ADCs are not perfect. Analog-to-digital conversion techniques with their
strengths and flaws are discussed in Chapter 5.
1.5 Electro–mechanical integration

The ability to combine many sensor channels in a small volume brings with it
the need to implement many connections, both within a detector module and
also to connect modules to the “outside world”. One must remove the heat due
to electrical power dissipation, control “cross-talk” (unwanted coupling between
different channels), provide precise mechanical positioning, and deal with a host
of other problems that straddle the realms of electronic and mechanical design.
To illustrate some of these problems, consider vertex detection in high-energy
physics. A powerful tool in identifying interesting events is the detection of sec-
ondary vertices. A particle formed in the primary collision, a B meson, for exam-
ple, decays after a brief time to form new particles, whose tracks form a vertex
displaced from the primary collision point. The formation of the initial particle
ELECTRO–MECHANICAL INTEGRATION 7
Fig. 1.8. Axial view of a collider event. Most tracks originate from the primary
interaction point in the center. A reconstructed neutrino is shown as a dashed
track, as it is not directly detected. Two B mesons are emitted toward the right
and decay in flight. The decay products originate from displaced vertices, which are
a few mm distant from the primary vertex. Concentric arrays of position-sensitive
detectors, shown schematically and not to scale, provide track coordinates at two
radii.
is inferred by reconstructing the trajectories of the secondaries and detecting the

displaced vertex.
Figure 1.8 illustrates the detection of displaced vertices. Segmentation of the
concentric detector layers provides both position resolution and the ability to
separate adjacent tracks. When the track density is not too high, high resolution
in the rϕ plane alone is sufficient for pattern recognition and track reconstruction.
Basic requirements for vertex detection can be derived from this simple tracking
system with two layers at radii r1 and r2 and resolutions of σ1 at r1 and σ2 at
r2 . The impact parameter resolution
2 2
σ1 r2 σ2 r1 1
σb2 ≈ + = (σ1 r2 )2 + (σ2 r1 )2 . (1.1)
r2 − r1 r2 − r1 (r2 − r1 )2
The position resolution at the inner radius is weighted by the outer radius, so
precision at the inner radius is paramount. If the two layers have equal resolution
σ1 = σ2 = σ, this result can be rewritten as
σ 2 2 2
b 1 1
≈ + . (1.2)
σ 1 − r1 /r2 r2 /r1 − 1
The geometrical impact parameter resolution is limited by the ratio of the outer
to inner radius, so it is desirable to measure the first space point at as small a
radius as possible. The obtainable impact parameter resolution improves rapidly

from σb /σ = 7.8 at r2 /r1 = 1.2 to σb /σ = 2.2 at r2 /r1 = 2 and attains values <
1.3 at r2 /r1 > 5. For σ = 10 µm and r2 /r1 = 2, σb ≈ 20µm. Thus, the inner layer
requires a high-resolution detector, which also implies a high-density electronic
readout with associated cabling and cooling, mounted on a precision support
structure. All of this adds material, which imposes an additional constraint.
The obtainable vertex resolution is affected by angular deflection due to mul-
tiple scattering from material in the detector volume. The scattering angle
r
0.0136[GeV/c] x x
Θrms = 1 + 0.038 · ln , (1.3)
p⊥ X0 X0
where p⊥ is the particle momentum, x the thickness of the material, and X0

the radiation length (see Particle Data Group 2004 for a concise summary). As
noted above, the position resolution at inner radii is critical, so it is important to
minimize material close to the interaction. Typically, the first layer of material
is the beam pipe.
Consider a Be beam pipe of x = 1 mm thickness and R = 5 cm radius. The
radiation length of Be is X0 = 35.3 cm, so x/X0 = 2.8·10−3 and at p⊥ = 1 GeV/c
the scattering angle Θrms = 0.56 mrad. This corresponds to σb Θrms = 28 µm,
which in this example would dominate the obtainable resolution. Clearly, any
material between the interaction and the measurement point should be mini-
mized and the first measurement should be at as small a radius as possible. This
exercise shows how experimental requirements drive the first detector layers to
small radii, which increases the particle flux (hits per unit area) and radiation
damage.
The need to reduce material imposes severe constraints on the sensor and
electronics, the support structures, and the power dissipation, which determines
the material in the cooling systems and power cabling. Since large-scale arrays
combine both analog and digital functions in the detector module, special tech-
niques must be applied to reduce pickup from digital switching without utilizing
massive shielding. Similar constraints apply in other applications, x-ray imagers,
for example, where Compton scattering blurs the image.
Subsequent chapters will provide detailed discussions of the relevant physics
and design parameters. In the spirit of a “road map” the remainder of this
chapter summarizes the key aspects of semiconductor detector systems.
1.6 Sensor structures I

1.6.1 Basic sensor
Semiconductor detectors are basically ionization chambers. In the simplest con-
figuration an absorbing medium is subtended by a pair of electrodes with an
applied voltage, as illustrated in Figure 1.9. Absorbed radiation liberates charge
pairs, which move under the influence of an applied field and induce an electrical
current in the external circuit.
SENSOR STRUCTURES I 9
isig
Vbias
isig
Fig. 1.9. Charge collection in a simple ionization chamber.
1.6.2 Position sensing

The electrodes of the sensor can be segmented to provide position information.
Now the magnitude of the signal measured on a given electrode depends on
its position relative to the sites of charge formation, as shown in Figure 1.10.
Segmenting one electrode into strips, as shown in the left panel of Figure 1.11,
provides position information in one dimension. Angled tracks will deposit charge
on two or more strips. Evaluating the ratio of charge deposition allows interpola-
tion to provide position resolution better than expected from the electrode pitch
alone. We’ll return to this later. A second orthogonal set of strips on the opposite
face gives two-dimensional position readout, shown in the second panel of Figure
1.11.
In a colliding-beam experiment the strip pitch (center-to-center distance) is
typically 25 – 100 µm and lengths range from centimeters to tens of centimeters,
usually aligned parallel to the beam axis to provide rϕ coordinates. The maxi-
mum strip length per sensor is limited by wafer size (10 – 15 cm for detector-grade
Si), so multiple sensors are ganged to form longer electrodes. Practical detectors
PARTICLE PARTICLE
TRACK TRACK
Fig. 1.10. Segmenting the sensor electrode provides position information. Angled
tracks deposit charge on two or more electrodes.
Fig. 1.11. Subdividing an electrode into strips provides one-dimensional position

sensing (left). Subdividing both electrodes to form orthogonal strips provides
two-dimensional imaging.
have used strips as long as 30 or 40 cm, limited by electronic noise and the hit
rate per strip.
Two-dimensional position sensing using crossed strips is simple, but has prob-
lems at high hit densities. Each hit generates an x- and a y-coordinate. However,
n tracks generate n x-coordinates and n y-coordinates, simulating n2 hits of
which n2 − n are fake. The “ghosts” can only be exorcised with additional infor-
mation to eliminate coordinates not consistent with tracks, clearly a formidable
task in a mixture of stiff and soft tracks with low-momentum particles loop-
ing in a magnetic field. A compromise solution that is often adequate utilizes
“small-angle stereo”, where the strips subtend a small angle, rather than 90◦ .
Small-angle stereo is illustrated in Figure 1.12. The area subtended by two
sensing elements (strips) of length L1 and L2 arranged at an angle 90◦ is A =
L1 L2 , so a hit in a given strip can form combinations with hits on all of the
transverse strips – the probability of “ghosting” is maximal. However, if the
p1
W
a
p2
L
Fig. 1.12. Small-angle stereo reduces the area subtended by strips that could pro-
vide a coincident signal. The width W of the shaded area subject to confusion is
L(p2 /p1 ) tan α + p2 .
SENSOR STRUCTURES I 11
angle α subtended by the two strip arrays is small (and their lengths L are
approximately equal), the capture area
p2
A ≈ L2 tan α + Lp2 . (1.4)
p1
Consider a given horizontal strip struck by a particle. To determine the longitu-
dinal coordinate, all angled strips that cross the primary strip must be checked
and every hit that deposits charge on these strips adds a coordinate that must
be considered in conjunction with the coordinate defined by the horizontal strip.
Since each strip captures charge from a width equal to the strip pitch, the exact
width of the capture area is an integer multiple of the strip pitch. The probability
of multiple hits within the acceptance area, and hence the number of “ghosts”, is
reduced as α is made smaller, but at the expense of resolution in the longitudinal
coordinate.
1.6.3 Pixel devices
To obtain unambiguous two-dimensional information the sensor must provide
fine segmentation in both dimensions, which can be achieved either by geomet-
rical or electronic segmentation. Charge coupled devices (CCDs), random access
pixel devices, and silicon drift chambers represent different approaches to ob-
taining nonprojective two-dimensional information. The conceptually simplest
implementation is shown in Figure 1.13. The sensor electrodes are patterned as
a checkerboard and a matching two-dimensional array of readout electronics is
connected via a two-dimensional array of contacts, for example solder bumps.
In this scheme the pixel size is limited by the area required by each electronic
readout cell. Pixel sizes of 30 – 100 µm are practical today, depending on the
complexity of the circuitry required in each pixel. Figure 1.13 also shows that
the readout IC requires more area than the pixel array to accommodate the
SENSOR
CHIP
READOUT
CHIP
BUMP
BONDS
READOUT WIRE-BOND PADS FOR

CONTROL DATA OUTPUT, POWER,
CIRCUITRY AND CONTROL SIGNALS
Fig. 1.13. Schematic view of a hybrid pixel detector. A pixellated sensor chip is
connected to a matching array of readout amplifiers by a two-dimensional array of
solder bumps. The readout chip extends beyond the sensor chip to accommodate
readout and control circuitry in addition to wire bonds for external connections.
readout control and driver circuitry and additional bond pads for the external
connections. Since multiple readout ICs are needed to cover more than several
cm2 , this additional area constrains designs that require full coverage. Examples
for integrating multiple readout ICs and sensors are discussed in Chapter 8.
Implementing this structure monolithically would be a great simplification
and some work has proceeded in this direction. Before describing these structures,
it is useful to discuss some basics of semiconductor detectors.
1.7 Sensor physics

1.7.1 Signal charge
All of the configurations discussed above differ only in the structures at the
surface of the sensor. Common to them is that the charge pairs are formed in
the sensitive volume and the average signal charge
E
Qs = e, (1.5)
Ei
where E is the absorbed energy, Ei the energy required to form a charge pair, and
e the electronic charge. In solids the absorbed energy must exceed the bandgap
to form mobile charge carriers. In Si the gap energy is 1.12 eV, so photons with
greater energy, i.e. wavelengths less than 1.1 µm, can be detected. At higher
energies (> 50 eV) the additional constraint of momentum conservation becomes
significant and the ionization energy Ei = 3.6 eV. As will be discussed in Chapter
2, the ionization energy Ei is proportional to the bandgap, so higher bandgap
materials yield less signal charge.
For a charged particle track traversing the sensor, the energy loss E – and
hence the signal charge Qs – will increase with sensor thickness. Minimum ioniz-
ing particles average about 80 electron–hole pairs per µm path length in silicon.
For x-rays absorbed by the photoelectric effect, the deposited energy is fixed, but
the sensor must be sufficiently thick to provide good efficiency. For gamma-rays
above 100 keV Compton interactions dominate, so the sensor volume must be suf-
ficiently large to accommodate multiple sequential interactions (for a discussion
see Knoll 2000).
When a low-energy x-ray is absorbed by the photoelectric effect, the charge
deposition is localized, with a charge cloud whose extent is determined by the
range of the ejected photoelectron. A charged particle traversing the sensor forms
charge pairs along the track with a radial extent of order µm. The signal is formed
when the liberated charge carriers move, which changes the induced charge on
the sensor electrodes. This will be treated quantitatively in Chapter 2, so at this
point we’ll simply note that when all signal charges have reached their respective
electrodes, the change in induced charge, i.e. the integrated signal current, is Qs .
To establish the electric field a potential is applied between the electrodes to
accelerate the charge carriers. As the carriers move through the medium they
scatter. After a short equilibration time (of order ps in Si) carrier transport
SENSOR PHYSICS 13
becomes nonballistic and the velocity does not depend on the duration of ac-
celeration, but only on the magnitude of the local electric field (see Sze 1981).
Thus, the velocity of carriers at position x depends only on the local electric field
E(x), regardless of where they originated and how long they have moved. The
carrier velocity
~
~v (x) = µE(x) , (1.6)
where µ is the mobility. For example, in Si the mobility is 1350 V/cm · s2 for
electrons and 450 V/cm · s2 for holes. As an estimate to set the scale, applying
30 V across a 300 µm thick absorber yields an average field of 103 V/cm, so the
velocity of electrons is about 1.4 · 106 cm/s and it will take about 20 ns for an
electron to traverse the detector thickness. A hole takes three times as long.
1.7.2 Sensor volume

To establish a high field with a small quiescent current, the conductivity of the
absorber must be low. Signal currents are typically of order µA, so if in the above
example the quiescent current is to be small compared to the signal current, the
resistance between the electrodes should be 30 MΩ. In an ideal solid the re-
sistivity depends exponentially on the bandgap. Increasing the bandgap reduces
the signal charge, so the range of suitable materials is limited. Diamond is an
excellent insulator, but the ionization energy Ei is about 6 eV and the range of
available thickness is limited. In semiconductors the ionization energy is smaller,
2.9 eV in Ge and 3.6 eV in Si. Si material can be grown with resistivities of order
104 Ω cm, which is too low; a 300 µm thick sensor with 1 cm2 area would have a
resistance of 300 Ω, so 30 V would lead to a current flow of 100 mA and a power
dissipation of 3 W. On the other hand, high-quality single crystals of Si and Ge
can be grown economically with suitably large volumes, so to mitigate the effect
of resistivity one resorts to reverse-biased diode structures.
The conductivity of semiconductors is controlled by introducing dilute con-
centrations of impurities into the crystal, a process called doping. Let the semi-
conductor be of atomic number Z. If the dopant is of atomic number Z + 1,
one of the shell electrons is only lightly bound and can be thermally excited
into the conduction band, so electrons are available as mobile charge carriers. If
the atomic number of the dopant is Z − 1, one of the bonds lacks an electron,
but only little energy is needed to “borrow” an electron from a nearby atom.
Thus, the unfilled bond moves and acts like a positive mobile charge, a “hole”.
To form a diode, one can start with material doped to provide mobile electrons,
“n-type” material. By introducing a Z − 1 dopant from the surface, a region can
be formed with holes as mobile carriers, “p-type” material. This forms a “pn-
junction”. When a voltage is applied with positive polarity on the n-side and
negative on the p-side (reverse bias), the electrons on the n-side and the holes on
the p-side are drawn away from the junction. Thus, the region adjacent to the
pn-junction is depleted of mobile charge and forms an insulator, over which the
applied voltage builds up the desired electric field, as illustrated in Figure 1.14.
p n
Fig. 1.14. Adjoining regions of p- and n-type doping form a pn-junction (top). The
charge of the mobile electrons and holes (circled) is balanced by the charge of the
atomic cores, so charge neutrality is maintained. When an external potential is
applied with positive polarity on the n-side and negative polarity on the p-side
(bottom), the mobile charges are drawn away from the junction. This leaves a net
space charge from the atomic cores, which builds up a linear electric field in the
junction. This is treated analytically in Chapter 2.
Note that even in the absence of an externally applied voltage, thermal diffu-
sion forms a depletion region. As electrons and holes diffuse from their original
host atoms, a space charge region is formed and the resulting field limits the
extent of thermal diffusion. As a result, every pn-junction starts off with a non-
zero depletion width and a potential difference between the p- and n-sides, the
“built-in” potential Vbi .
Figure 1.15 shows the cross-section of a typical detector diode. The pn-
junction is formed by introducing the dopant at the upper surface. The detector
junction is in the middle. Similarly doped regions to the left and right indicate a
guard ring, which surrounds the detector diode to isolate it from the edge of the
wafer. Mechanical damage at the edge leads to very large leakage currents. The
guard ring, biased at the same potential as the detector electrode, captures the
edge currents and also forms a well-defined electrical boundary for the detector
diode (the active area ends midway between the detector electrode and the guard
ring). Metallization layers (typically aluminum) deposited on the electrodes pro-
vide electrical contact. The intermediate silicon surface is protected by a layer of
SiO2 that provides a well-controlled interface to the silicon lattice. In detectors
the surface side of the junction is usually much more heavily doped than the
substrate material, so the resulting asymmetric junction depletes into the bulk.
Appendix A provides more details on detector structures and fabrication.
SENSOR PHYSICS 15
OXIDE JUNCTION CONTACT GUARD RING
~ 1 mm
300 mm
~ 1 mm
n BULK OHMIC CONTACT p+ DOPING
Fig. 1.15. Detector diodes are usually asymmetric, with a highly doped layer at the
surface and a lightly doped bulk. With reverse bias the junction depletes into the
bulk. SiO2 layers protect the silicon surface and a guard ring isolates the diode from
the edges of the chip.
A reverse bias voltage Vb yields the depletion width

r
2ε(Vb + Vbi )
wd = , (1.7)
Ne
where N is the dopant concentration in the bulk and ε the dielectric constant
(11.9ε0 for Si). The “built-in” junction potential Vbi in detector diodes is typically
about 0.5 V. When the depletion width is less then the silicon thickness, the diode
is “partially depleted”. When wd extends to the back contact the diode is “fully
depleted”.
The depleted junction volume is free of mobile charge and thus forms a ca-
pacitor, bounded by the conducting p- and n-type semiconductor on each side.
The capacitance
s
A εeN
C=ε =A . (1.8)
wd 2(Vb + Vbi )
For bias voltages Vb Vbi

1
C∝√ . (1.9)
Vb
In technical units
C ε pF 1
= ≈1 . (1.10)
A wd cm W
A Si diode with 100 µm thickness has a capacitance of about 1 pF/mm2 . This ap-
plies to a detector whose electrodes are large compared to the depletion thickness.
In strip and pixel detectors the fringing capacitance to neighboring electrodes
usually dominates. The interstrip capacitance depends on the ratio of electrode
width w to strip pitch p. For typical geometries the interstrip capacitance Cs per
cm length l follows the relationship (Demaria et al. 2000)

Cs w + 20µm pF
= 0.03 + 1.62 . (1.11)
l p cm
Typically, the interstrip capacitance is about 1 pF/cm. The backplane capaci-

tance
pl
Cb ≈ εε0 . (1.12)
w
Since the adjacent strips confine the fringing field lines to the interstrip bound-
aries, the strip appears as an electrode with a width equal to the strip pitch.
Corrections apply at large strip widths (Barberis et al. 1994).
Ideally, reverse bias removes all mobile carriers from the junction volume, so
no current can flow. However, thermal excitation can promote electrons across
the bandgap, so a current flows even in the absence of radiation, hence the
term “dark current”. The probability of electrons surmounting the bandgap is
increased strongly by the presence of impurities in the lattice, as they introduce
intermediate energy states in the gap that serve as “stepping stones”. As derived
in Appendix F the reverse bias current depends exponentially on temperature
T,
Eg
IR ∝ T 2 exp − , (1.13)
2kT
where Eg is the bandgap energy and k the Boltzmann constant, so cooling the
detector can reduce leakage substantially. The ratio of leakage currents at tem-
peratures T1 and T2
2
IR (T2 ) T2 Eg T1 − T2
= exp − . (1.14)
IR (T1 ) T1 2k T1 T2
In Si (Eg =1.12 eV) this yields a ten-fold reduction in leakage current when the
temperature is lowered by 14 ◦ C from room temperature.
1.7.3 Charge collection

How quickly electrons and holes are swept from the depletion region is deter-
mined by the electric field. To simplify the following equations we’ll set V ≡
Vb + Vbi . At low reverse bias the field in the depletion region initially has a
triangular profile
r
eN 2N e x x
|E(x)| = (wd − x) = V · 1− ≡ Emax · 1 − (1.15)
ε ε wd wd
up to the voltage where the depletion width wd equals the thickness of the bulk
d, corresponding to the depletion voltage
SENSOR PHYSICS 17
V< Vd V > Vd
E E
Emax
Emax Emin
x x
w d d
Fig. 1.16. Electric field distributions in a partially depleted detector (left) and a
detector operated with overbias (right).
N ed2
Vd = . (1.16)
2ε
Increasing the bias voltage V beyond this point (“overbias”, often called “overde-
pletion”) increases the field uniformly and evens out the field profile
2Vd x V − Vd
|E(x)| = 1− + . (1.17)
d d d
Then the maximum field is (V + Vd )/d and the minimum field (V − Vd )/d.
Figure 1.16 illustrates the electric field distributions in partial depletion and
with overbias.
When radiation forms electron–hole pairs, they drift under the influence of
the field with a velocity v = µE. The time required for a carrier to traverse the
full detector thickness, the collection time, is

d2 V + Vd + 2Vbi
tc = log , (1.18)
2µV V − Vd
where the collection time for electrons or holes is obtained by using the appro-
priate mobility. At full depletion or beyond, the collection time can be estimated
by using the average field E = V /d, so
d d d2
tc ≈ = = (1.19)
v µE µV
and charge collection can be sped up by increasing the bias voltage. In partial
depletion, however, the collection time is independent of bias voltage and de-
termined by the doping concentration alone, as d2 /V remains constant. This is
discussed in Chapter 2.
In practice the dopant concentration N of silicon wafers is expressed as the

resistivity of the material ρ = (eµN )−1 , as this is readily measurable. Using this
parameter and introducing technical units yields the depletion voltage
" #
Ω · cm d2
Vdn = 4 2 · ρ − Vbi (1.20)
(µm) n
for n-type material and

" #
Ω · cm d2
Vdp = 11 · − Vbi (1.21)
(µm)2 ρp
for p-type material. The resistivity of silicon suitable for tracking detectors (or
more precisely, the highest resistivity available economically) is 5 – 10 kΩ cm.
Note that in 10 kΩ cm n-type Si the built-in voltage by itself depletes 45 µm
of material. Detector wafers are typically 300 µm thick. Hence, the depletion
voltage in n-type material is 35 – 70 V for the resistivity range given above.
Assuming 6 kΩ cm material (Vd = 60 V) and an operating voltage of 90 V, the
collection times for electrons and holes are 8 ns and 27 ns, respectively. Electron
collection times tend to be somewhat longer then given by eqn 1.15 since the
electron mobility decreases at fields > 103 V/cm (see Chapter 2 and Sze 1981)
and eventually the drift velocity saturates at 107 cm/s. At saturation velocity the
collection time is 10 ps/µm. In partial depletion, as noted above, the collection
time is independent of voltage and depends on resistivity alone. For electrons
the collection time constant
h ns i
τcn = ρε = 1.05 ·ρ . (1.22)
k Ω · cm
To increase the depletion width or speed up the charge collection one can
increase the voltage, but ultimately this is limited by the onset of avalanching.
At sufficiently high fields (greater than about 105 V/cm in Si) electrons acquire
enough energy between collisions that secondary electrons are ejected. At even
higher fields holes can eject secondary electrons, which in turn can eject new
secondaries, and a self-sustaining charge avalanche forms (see Chapter 2). This
phenomenon is called “breakdown” and can cause permanent damage to the sen-
sor. In practice avalanching often occurs at voltages much lower than predicted
by eqn 1.17, since high fields can build up at the relatively sharp edges of the
doping distribution or electrode structures. When controlled, charge avalanching
can be used to increase the signal charge, as discussed in Chapter 2. In detect-
ing visible light, the primary signal charge is quite small, so this technique is
most often applied in photodiodes to provide internal gain and bring the signal
above the electronic noise level (avalanche photodiodes or APDs). APDs must
be designed carefully to prevent breakdown and also to reduce additional signal
fluctuations introduced by the avalanche process. Bias voltage and temperature
both affect the gain strongly, so they must be kept stable.
SENSOR PHYSICS 19
1.7.4 Energy resolution

The minimum detectable signal and the precision of the amplitude measurement
are limited by fluctuations. The signal formed in the sensor fluctuates, even for a
fixed energy absorption. Generally, sensors convert absorbed energy into signal
quanta. In a scintillation detector absorbed energy is converted into a number of
scintillation photons. In an ionization chamber energy is converted into a number
of charge pairs (electrons and ions in gases or electrons and holes in solids). The
absorbed energy divided by the excitation energy yields the average number of
signal quanta
E
N= . (1.23)
Ei
This number fluctuates statistically, so the relative resolution
√ r
∆E ∆N FN F Ei
= = = . (1.24)
E N N E
The resolution improves with the square root of energy. F is the Fano factor,
which comes about because multiple excitation mechanisms can come into play
and reduce the overall statistical spread. For example, in a semiconductor ab-
sorbed energy forms electron–hole pairs, but also excites lattice vibrations –
quantized as phonons – whose excitation energy is much smaller (meV vs. eV).
Thus, many more excitations are involved than apparent from the charge sig-
nal alone and this reduces the statistical fluctuations of the charge signal. For
example, in Si the Fano factor is 0.1. The Fano factor is explained in Chapter 2.
In most applications, the intrinsic energy resolution of semiconductor sensors
is so good that external contributions determine the overall fluctuations. How-
ever, for low-energy x-rays signal charge fluctuations are significant, whereas in
gamma-ray detectors electronic noise tends to determine the obtainable energy
resolution. For minimum ionizing charged particles, it is the statistics of energy
loss. Since the energy deposited by minimum ionizing particles varies according
to a Landau–Vavilov distribution (Figure 1.17) with σQ /Qs ≈ 0.2 in 300 µm
of Si, the inherent energy resolution of the detector is negligible. Nevertheless,
electronic noise is still important in determining the minimum detectable signal,
i.e. the detection efficiency.
1.7.5 Position resolution
The position resolution of the detector is determined to first order by the elec-
trode geometry. The size and shape of the electrodes is limited by the size of a
wafer, on the one hand (10 or 15 cm diameter for detector grade material), and
the resolution capability of IC fabrication technology on the the other (∼ 1 µm).
In practice the lower bound is set by the readout electronics, which in the small-
est dimension tend to require 20 – 50 µm overall width. Most commonly, sensors
for tracking applications have strip electrodes. The strips are usually 8 – 12 µm
wide, placed on a pitch of 25 – 50 µm, and 6 – 12 cm long. Frequently, multiple
sensor wafers are ganged to form longer electrodes.
NUMBER OF EVENTS
2 3 4 5 6 7 8 9 10
DEPOSITED CHARGE (fC)
Fig. 1.17. Measured energy loss distribution of 1.5 MeV/c electrons in a silicon de-
tector. The dashed line is a Vavilov theory calculation. (Wood et al. 1991. Figure
courtesy of P. Skubic)
It is important to note that despite the gaps between electrodes, the detectors
still remain 100% efficient. The field lines remain parallel in the detector until
near the surface, where they bend along the surface and end on the electrode.
Hence, the electrical segmentation is determined by the electrode pitch, rather
than the width. Since the response function is essentially box-like, the position
resolution of a single detector is equal to the strip pitch p. However, for tracks
randomly aligned with respect to a strip, the differences between the measured
and the true positions have a Gaussian distribution with the standard deviation
Zp/2
2 x2 p2
σ = dx = , (1.25)
p 12
−p/2
√
so the root mean square (rms) resolution is the strip pitch divided by 12. The
same mechanism leads to “sampling noise” in image processing or when digitizing
an analog waveform, and is discussed in Section 5.2.
To first order the electrons and holes simply follow the field lines on which
they originated and end on a certain electrode. In reality, however, they are also
subject to thermal diffusion, which spreads the charge cloud transversely as the
charges drift through the detector, with an rms width
√
σy = 2Dt . (1.26)
Since the diffusion constant is linked to the mobility by the Einstein relation
kT
D= µ (1.27)
e
SENSOR PHYSICS 21
TRANSVERSE DIFFUSION
-15 -10 -5 5 10 15 [mm]
PARTICLE TRACK
Fig. 1.18. Transverse diffusion distributes charge on multiple strip electrodes, here
shown on a 20 µm pitch. The charge division boundaries are indicated by dashed
lines. By evaluating the distribution of charge on the electrodes, the position reso-
√
lution can be improved beyond the geometric resolution σ = p/ 12.
and the collection time is inversely proportional to the carrier mobility, the trans-
verse diffusion is the same for electrons and holes. Using the average field ap-
proximation E = V /d the transverse diffusion
s
√ kT d2
σy = 2Dt ≈ 2 , (1.28)
e Vb
which is independent of mobility, giving the same result for electrons and holes.
For d = 300 µm, T = 300 K and Vb = 100 V the transverse diffusion σy ≈ 7 µm.
At first glance this might seem to degrade the obtainable position resolution,
but it can, in fact, be turned to advantage, since transverse diffusion spreads
charge to neighboring strips. As illustrated in Figure 1.18, one can evaluate the
charge distribution over a central strip and its neighbors to improve the position
resolution beyond the limit given by strip geometry. Since the fractional charge
terminating on the neighboring strip is determined by superimposed Gaussian
distributions (Lüth 1990), whose integral falls off rapidly for deviations beyond
several standard deviations, this technique is practical only for a rather limited
range of strip pitches. In the interpolation regime the position resolution is in-
versely proportional to signal-to-noise ratio. Kenney et al. (1993) have applied a
weighted interpolation algorithm to rectangular pixels of 34 µm × 125 µm with
S/N = 55. In the direction of the 34 µm pitch the measured resolution was
2.2 µm, whereas in the direction of the 125 µm pitch the interpolation could only
be applied to the outer 25 µm√regions to yield 5.3 µm resolution. The resolution
in the central region was 75/ 12 µm.
R R R R
Css Css Css Css
E Cb Cb Cb Cb
PARTICLE -V
TRACK
Fig. 1.19. Intermediate “floating” electrodes can be used to reduce the effective read-
out pitch (left). The equivalent circuit (right) shows how the signal current induced
on the “floating” electrodes is transferred to the readout amplifiers through the
intermediate capacitive dividers formed by the strip-to-strip capacitance Css and
the backplane capacitance Cb .
The range of charge interpolation can be extended by introducing interme-

diate strips that are not connected to readout channels (Figure 1.19). The bias
resistors keep all strips at the same quiescent potential, but the time constant
formed by the bias resistance and the strip capacitance is made so large that
the potential of a “floating” strip can change individually in response to signal
charge. The charge induced on the “floating” strips is coupled capacitively to
its neighbors. The readout amplifiers must have a low input impedance, so that
the signal current from a given electrode will divide inversely proportional to
the effective coupling capacitance. It is crucial that all electrodes be at the same
quiescent potential, to ensure uniform charge collection efficiency. Connecting
the bias resistors as shown in Figure 1.19 ensures that each electrode is biased at
the input voltage of the amplifiers. The biasing resistors must be large to reduce
electronic noise, as will be discussed in Chapter 4, but still small enough that
the detector leakage current does not alter the electrode voltages significantly.
For simplicity, first assume that the backplane capacitance Cb is zero. Then
the capacitances coupling the central strip to the two amplifiers are formed by
two interstrip capacitances Css in series on each side. Thus, the signal current
will divide equally. For the strip to the left of center, the coupling capacitance
is Css to the left-hand amplifier and Css /3 to the right-hand amplifier, so the
left-hand amplifier will receive 3/4 of the signal.
This technique can also be used to reduce the number of readout channels.
However, a portion of the signal charge remains on the backplane capacitance
Cb of each strip in the signal path. For charge induced on the central strip, the
charge transferred to one of the two amplifiers
SENSOR PHYSICS 23
10000
ABSORPTION COEFFICIENT (cm-1)

1000
100 PHOTOELECTRIC
10
1
COMPTON
PAIR
0.1
RAYLEIGH
0.01
0.001
0.001 0.01 0.1 1 10 100 1000
PHOTON ENERGY (MeV)
Fig. 1.20. The photon absorption coefficient µ vs. energy in silicon. At low energies
photoelectric absorption dominates. Above 100 keV Compton scattering takes over
and at high energies pair production dominates. The fraction of photons that have
interacted within a distance x is 1 − exp(µx).
Qa 1 1 1 1
= · 1 2 ≈ · . (1.29)
Qs 2 1 + 2(Cb /Css ) + 2 (Cb /Css ) 2 1 + 2 (Cb /Css )
For Cb /Css = 0.1, a typical ratio for strip detectors, the exact expression yields
Q1a /Qs = 0.41, so the backplane capacitance incurs an 18% loss in the summed
signal from both amplifiers. However, the capacitance presented to the ampli-
fier inputs is also smaller than in a fully instrumented system, so interpolation
via floating strips can reduce overall power dissipation with respect to a fully
instrumented readout. Double-track resolution, however, will be determined by
the readout pitch.
In attempting to optimize position resolution, other effects must also be con-
sidered. The energy deposited by minimum ionizing particles fluctuates along
the track, so tracks impinging at the same position will show varying centroids
in the induced charge. A further limit on the position resolution is imposed by
energetic delta-electrons formed along the track trajectory, which can skew the
charge centroid appreciably (Bedeschi et al. 1989).
In x-ray imaging applications where photoelectric absorption dominates, the
resolution is limited by the range of the emitted photoelectron, as it deposits
energy along its path. The binding energy Eb ≈ 2 keV, so the photoelectron’s
energy
Ek = Ephoton − Eb . (1.30)
For a 20 keV photon the photoelectron’s range is about 5 µm, whereas for 100
keV the range is about 80 µm. As can be seen in Figure 1.20 Compton scatter-
ing is about equally probable at 100 keV and dominates up to about 10 MeV.
However, given the maximum practical silicon detector thickness of several mm

reasonable efficiency only obtains to 30 or 40 keV. Materials with higher atomic
number are necessary for higher energies. Alternative materials are discussed
briefly in Chapter 2. At high energies pair production can be used to determine
the direction of gamma rays by detecting the emitted electrons and positrons in
a silicon strip tracker (Chapter 8).
1.8 Sensor structures II – monolithic pixel devices

In the early years of large-scale semiconductor detectors the monolithic inte-
gration of large scale sensors with electronics was viewed as the “holy grail”.
Clearly, it is an appealing concept to have a 6 × 6 cm2 detector tile that com-
bines a strip detector and 1200 channels of readout electronics with only the
power and data readout as external connections. The problem was perceived at
the time to be the incompatibility between IC and detector fabrication processes
(see Appendix A). Development of an IC-compatible detector process allowed
the monolithic integration of high-quality electronics and full depletion silicon
sensors without degrading sensor performance (Holland and Spieler 1990), subse-
quently extended to full CMOS circuitry (Holland 1992). Nevertheless, a simple
yield estimate shows that this isn’t practical. In the conventional scheme read-
ing out ∼ 1200 channels with a 50 µm readout pitch requires 10 ICs with 128
channels each. These devices are complex, so their yield is not 100%. Even when
assuming 90% functional yield per 128-channel array, the probability of ten adja-
cent arrays on the wafer being functional is prohibitively small. The integration
techniques are applicable, however, to simpler circuitry and have been utilized in
monolithic pixel detectors (Snoeys et al. 1992). The oldest and most widespread
wafer-scale monolithic imaging device is the charge coupled device.
1.8.1 Charge coupled devices

The classic high-resolution pixel array is the charge coupled device (CCD), which
combines the charge readout with the sensors. Figure 1.21 illustrates the princi-
ple. In signal acquisition mode the pixels function as small ionization chambers.
To transfer the signal charge to the readout amplifier, additional electrodes are
appropriately biased to shift the charge to the adjacent electrode. By applying
the appropriate sequence of pulses, the signal charge is sequentially transferred
to the output electrode, which in turn is connected to a readout amplifier. This
structure allows small pixel sizes, about 10 µm, and provides full coverage. The
drawback is that the readout is sequential, so larger arrays require more readout
time. Since charge is commonly transferred over thousands of pixels, the charge
transfer efficiency η from one pixel to the next must be very close to unity. After
transferring through n pixels the signal arriving at the output node is attenu-
ated by η n , but modern fabrication techniques provide practically 100% charge
transfer over ∼ 104 pixels. Pixels are read out sequentially, column by column as
shown in Figure 1.21. Typically, a single amplifier reads out the entire array. Low
noise militates against fast clocking, so readout times are long. This is discussed
SENSOR STRUCTURES II – MONOLITHIC PIXEL DEVICES 25
PARTICLE
TRACK
T1 T2 T3 T4 T5 T6 t
T1
T2
PIXEL
ARRAY
T3
T4
OUTPUT
AMPLIFIER T5
T6
SERIAL OUTPUT REGISTER
Fig. 1.21. Upper right: schematic cross-sectional view of a CCD. Voltages are applied
to the electrodes according to the timing diagram at the upper left. The potential
sequence shifts the charge from the track to the right. Three electrodes comprise one
pixel, but all charge from the track subtended by the pixel is drawn to the pixel’s
left-most electrode. Six clock periods shift the charge to the neighboring pixel.
The pixels are read out sequentially (bottom left). Charge is transferred down the
column and then horizontally. The charge is deposited on a storage capacitor and
transferred to the readout line by the output amplifier.
in Chapter 4. Large arrays are commercially available; the SLD detector (Abe
et al. 1997) used 16 × 80 mm2 devices with 20 µm pixels. The sensitive depth
was 20 µm, so minimum ionizing particles yielded a broad charge distribution
peaking at about 1200 e. Electronic noise was 100 e. The thin depletion depth
reduces the signal, but limits transverse diffusion and provides excellent position
resolution. The readout rate was 5 MHz and four readout amplifiers were used
to speed up the readout. CCDs are in widespread use, but high-energy physics
and x-ray detection require specialized devices.
1.8.2 Silicon drift chambers

An ingenious structure that provides the functionality of a CCD without discrete
transfer steps is the silicon drift chamber (Gatti and Rehak 1984, Rehak et
TRACK
-V
-V 1 0V
p+ p+
n
p+
-V 1 +V 2
-V
Fig. 1.22. Principle of a silicon drift chamber. The n-type bulk is depleted from
both surfaces by a series of p+ electrodes, biased to provide a positive potential
gradient along the center axis of the detector. Holes drift to the p electrodes, whereas
electrons are transported parallel to the surface and then attracted to the collection
electrode, where the signal is read out.
al. 1985). In this device a potential trough is established in the bulk, so that
the signal charge is collected in the trough and then drifts towards the readout
electrode (Figure 1.22). The position is derived from the time it takes for a signal
charge to move to the output, so the detector requires a time reference. When a
pulsed accelerator or pulsed x-ray tube is used, the start time is readily available.
With random rates, as with radioactive sources, the time reference must be
derived from the sensor. This will be discussed in Chapters 3 and 4. Although
originally proposed as a position-sensing device, the Si drift chamber’s other
useful application is energy spectroscopy. Since this structure collects charge
from a large area onto a small collection electrode, the capacitance presented to
the readout amplifier is small (order 10 – 100 fF), so the electronic noise can be
very low. This can be exploited in x-ray detection and in photodiodes. Various
drift detector topologies are described by Lutz (1999).
1.8.3 Monolithic active pixel sensors

Neither CCDs nor silicon drift devices can be fabricated using standard IC fabri-
cation processes. The doping levels required for diode depletion widths of 100 µm
or more are much lower than used in commercial integrated circuits. The pro-
cess complexity and yield requirements of the readout electronics needed for
most application dictate the use of industry-standard fabrication processes. In
contrast to detectors, where the entire thickness of the silicon wafer is utilized
for charge collection, integrated electronics utilize only a thin layer, of order µm,
at the surface of the silicon. The remainder of the typically 500 µm thick wafer
provides mechanical support, but also serves to capture deleterious impurities,
through gettering processes described in Appendix A. IC substrate material –
typically grown by the Czochralski method – has both crystalline defects and
impurities, whereas detector grade material utilizes float-zone material, which is
SENSOR STRUCTURES II – MONOLITHIC PIXEL DEVICES 27
VISIBLE LIGHT
n METALLIZATION
p EPI
p + BULK
ELECTRONICS
SENSOR
TRACK
p-WELL WITH CIRCUITRY
n WELL (~1 µm)
p EPI (~1-10 µm)
p+ BULK
(300 - 500 µ m)
Fig. 1.23. An active pixel sensor that integrates sensors and electronics monolith-
ically. The left shows an implementation for visible light. Electrons are collected
directly by the sensor electrode formed by the n-well. Light penetrates only a short
distance, so the portion of the epi-layer covered by the electronics is insensitive. The
right shows an alternative layout for high-energy particles. Electrons formed by a
track traversing the electronics diffuse towards the n-well electrodes. This structure
provides 100% sensitive area, but diffusion transport leads to long collection times.
dislocation-free and achieves the very low impurity levels needed for high resis-
tivity. High-quality transistors also require low defect densities, but much higher
doping levels than detectors, so a thin high quality layer is epitaxially grown
on the Czochralski substrate (referred to as the “epi-layer”). The doping levels
are still quite high, typically corresponding to 1 – 10 Ω cm resistivity (103 times
lower than in typical detectors), so breakdown limits depletion widths to several
µm or less.
Visible light (400 – 700 nm) in silicon is practically fully absorbed in a thick-
ness of 0.5 – 7 µm, so thin depletion layers are usable, also because diffusion from
non-depleted silicon also adds to the charge signal. Driven by the digital camera
market and other commercial applications, there is widespread activity in the
application of conventional IC fabrication processes to optical imaging (Fossum
1997). These devices, called CMOS imagers or active pixel sensors, utilize a por-
tion of the pixel cell as a sensor, as illustrated in Figure 1.23. Each pixel includes
an active region (the sensor) with adjacent amplifier and readout circuitry. Met-
allization layers provide connections between the sensor and the electronics, and
between the components in the electronics cell. Light impinging on the sensor
is detected, but is blocked by the metallization. Typical “fill factors”, the ratio
of light-sensitive area to pixel area, are 20 – 30%, but this can be improved as
smaller feature sizes shrink the electronics.
High-energy particles, on the other hand, traverse both the metallization and
the transistors, so devices have been developed that seek to utilize the epi-layer
in the entire pixel as the sensor region, as shown in the right panel of Figure 1.23
(Deptuch et al. 2003, Turchetta et al. 2003, Kleinfelder et al. 2004). Since the
depletion layers are thin, this device relies on diffusion for a substantial portion
of the recovered charge. Thus, the recovered signal charge is much smaller than
in fully depleted detectors, about 1000 e compared to 22 000 e for a 300 µm thick
sensor. Diffusion is channeled laterally by the potential well formed in the epi-
layer, as it is lightly doped with respect to the substrate and the p-wells that
accommodate the electronics. Since the pixel capacitance is small, electronic
noise levels can also be low.
However, relying on diffusion increases the collection time to about 100 ns,
which can still provide the time resolution required at high-luminosity colliders,
but radiation damage will degrade the carrier lifetime (see Appendix F) to order
1 – 10 ns after a relatively short time at high luminosity and the small radii where
pixel detectors are needed. Incomplete charge collection also limits the usability
of these devices in applications that require good energy resolution, e.g. x-ray
spectroscopy, although they may be acceptable for counting measurements.
The conceptually simplest form of an active pixel array is a matrix of tran-
sistors. During image acquisition all transistors are inactive and signal charge
is stored on their input capacitance. Control electrodes are bussed by row and
outputs by column. During readout each transistor is addressed individually by
selecting the appropriate row and all columns read out simultaneously. This
structure has been implemented by monolithically integrating the transistors
(called DEPFETs) in a high-resistivity substrate (Kemmer and Lutz 1987). This
arrangement allows the readout of individual pixels, but unlike more complex
active pixel devices can’t signal which pixels to read out. When reading out full
image frames the performance of this structure is comparable to a CCD with a
fully parallel readout. Understanding the limits of this technique requires some
additional background, so we’ll return to it in Chapter 8.
Some active pixel designs replicate the fully sequential readout used for
CCDs. This is a good match to digital photography, where every pixel carries in-
formation. The electronic circuitry in each pixel cell is quite simple and readout
can be slow, so the circuitry does not occupy much area. Slow readout applied
to charged particle detection also allows simple circuitry and facilitates low elec-
tronic noise. However, in sparse data environments with high event rates, such
as high luminosity hadron colliders, “smart pixels” that signal the presence of a
hit and then allow the selective readout of struck pixels sorted by time stamp are
necessary. This requires both fast response, to allow time stamping, and local
threshold discrimination, buffering, and readout logic. This drives up circuit com-
ELECTRONICS 29
plexity substantially, so the “real estate” occupied by the electronics increases,

both in the pixel and in the common control and readout circuitry. This will
be illustrated in Chapter 8. Comparison of various technologies requires careful
scrutiny that the adopted architecture and circuit design match the intended
purpose and not some simpler situation.
1.9 Electronics
Electronics are a key component of all modern detector systems. Although ex-
periments and their associated electronics can take very different forms, the same
basic principles of the electronic readout and the optimization of signal-to-noise
ratio apply to all.
The purpose of pulse processing and analysis systems is to
1. Acquire an electrical signal from the sensor. Typically this is a short current
pulse.
2. Tailor the time response of the system to optimize
(a) the minimum detectable signal (detect hit/no hit),
(b) energy measurement,
(c) event rate,
(d) time of arrival (timing measurement),
(e) insensitivity to sensor pulse shape,
or some combination of the above.
3. Digitize the signal and store for subsequent analysis.
Position-sensitive detectors utilize the presence of a hit, amplitude measure-
ment, or timing, so these detectors pose the same set of requirements. Generally,
these properties cannot be optimized simultaneously, so compromises are neces-
sary.
In addition to these primary functions of an electronic readout system, other
considerations can be equally or even more important, for example radiation
resistance, low power (portable systems, large detector arrays, satellite systems),
robustness, and – last, but not least – cost.
1.10 Detection limits and resolution

In addition to signal fluctuations originating in the sensor, the minimum detec-
tion limit and energy resolution are subject to fluctuations introduced by the
electronics. The gain can be controlled very precisely, but electronic noise in-
troduces baseline fluctuations, which are superimposed on the signal and alter
the peak amplitude. Figure 1.24 (left) shows a typical noise waveform. Both the
amplitude and time distributions are random.
When superimposed on a signal, the noise alters both the amplitude and
time dependence. Figure 1.24 (right) shows the noise waveform superimposed on
a small signal. As can be seen, the noise level determines the minimum signal
whose presence can be discerned.
TIME TIME
Fig. 1.24. Waveforms of random noise (left) and signal + noise (right), where the
peak signal is equal to the rms noise level (S/N = 1). The noiseless signal is shown
for comparison.
In an optimized system, the time-scale of the fluctuations is comparable to

that of the signal, so the peak amplitude fluctuates randomly above and below
the average value. This is illustrated in Figure 1.25, which shows the same signal
viewed at four different times. The fluctuations in peak amplitude are obvious,
but the effect of noise on timing measurements can also be seen. If the timing
signal is derived from a threshold discriminator, where the output fires when the
signal crosses a fixed threshold, amplitude fluctuations in the leading edge trans-
late into time shifts. If one derives the time of arrival from a centroid analysis, the
timing signal also shifts (compare the top and bottom right figures). From this
TIME TIME
TIME TIME
Fig. 1.25. Signal plus noise at four different times, shown for a signal-to-noise ratio
of about 20. The noiseless signal is superimposed for comparison.
DETECTION LIMITS AND RESOLUTION 31
NORMALIZED COUNT RATE

Qn
0.78
FWHM= 2.35 Q n
0.5
Q s /Q n
Fig. 1.26. Repetitive measurements of the signal charge yield a Gaussian distribu-
tion whose standard deviation equals the rms noise level Qn . Often the width is
expressed as the full width at half maximum (FWHM), which is 2.35 times the
standard deviation.
one sees that signal-to-noise ratio is important for all measurements – sensing
the presence of a signal or the measurement of energy, timing, or position.
1.10.1 Electronic noise

Electronic noise originates as both velocity or number fluctuations. This is dis-
cussed in detail in Chapter 3. Velocity fluctuations arise from thermal excitation.
The spectral density of the noise power can be derived directly as the long wave-
length limit in Planck’s theory of black-body radiation (see Chapter 3). At the
frequencies of interest here the spectral density is independent of frequency; the
spectrum is “white”. Number fluctuations occur when charge carriers are injected
into a sample independently of one another. Thermionic emission or current flow
through a semiconductor pn-junction are common examples. This is called “shot
noise” and also has a white spectrum.
In electronic circuits the noise sources can be modeled either as voltage or
current sources. Generally, the frequency spectra of the signal and the noise are
different. Typically, the noise spectra extend over a greater frequency band than
the signal, so by shaping the frequency response of the system one can optimize
the signal-to-noise ratio. The amplitude distribution of the noise is Gaussian,
so superimposing a constant amplitude signal on a noisy baseline will yield a
Gaussian amplitude distribution whose width equals the noise level (Figure 1.26).
To analyze the contributions of electronic noise, let’s consider a typical detec-
tor front-end as shown in Figure 1.27. The sensor is represented by a capacitance
Cd , a relevant model for most detectors. Bias voltage is applied through resistor
Rb and the signal is coupled to the preamplifier through a blocking capacitor
Cc . The series resistance Rs represents the sum of all resistances present in the
input signal path, e.g. the electrode resistance, any input protection networks,
and parasitic resistances in the input transistor. The preamplifier provides gain
and feeds a pulse shaper, which tailors the overall frequency response to optimize
DETECTOR BIAS
Cb
PREAMPLIFIER PULSE SHAPER
BIAS Rb
RESISTOR
OUTPUT
Cc Rs
DETECTOR Cd
Fig. 1.27. Typical detector front-end circuit.
signal-to-noise ratio while limiting the duration of the signal pulse to accommo-
date the signal pulse rate. Even if not explicitly stated, all amplifiers provide
some form of pulse shaping, due to their limited frequency response.
The equivalent circuit for the noise analysis (Figure 1.28) includes both cur-
rent and voltage noise sources. The leakage current of a semiconductor detector,
for example, fluctuates due to electron emission statistics. This “shot noise” ind
is represented by a current noise generator in parallel with the detector. Resis-
tors exhibit noise due to thermal velocity fluctuations of the charge carriers. This
noise source can be modeled either as a voltage or current generator. Generally,
resistors shunting the input act as noise current sources and resistors in series
with the input act as noise voltage sources (which is why some in the detector
community refer to current and voltage noise as “parallel” and “series” noise).
Since the bias resistor effectively shunts the input, as the capacitor Cb passes
current fluctuations to ground, it acts as a current generator inb and its noise
current has the same effect as the shot noise current from the detector. Any
other shunt resistances can be incorporated in the same way. Conversely, the
series resistor Rs acts as a voltage generator. The amplifier’s noise is described
DETECTOR BIAS SERIES AMPLIFIER +

RESISTOR RESISTOR PULSE SHAPER
Rs ens
ena
Rb
Cd inb ina
ind
Fig. 1.28. Equivalent circuit for noise analysis of the detector front-end in Figure
1.27.
fully by a combination of voltage and current sources at its input, shown as ena
and ina .
Shot noise and thermal noise have a “white” frequency distribution, i.e. the
spectral densities are constant with the magnitudes
i2nd = 2eId
i2nb = 4kT /Rb
e2ns = 4kT Rs
where e is the electronic charge, Id the detector bias current, k the Boltzmann
constant, and T the√ temperature. Typical√ amplifier noise parameters ena and ina
are of order nV/ Hz and fA to pA/ Hz. Trapping and detrapping processes
in resistors, dielectrics and semiconductors can introduce additional fluctuations
whose noise power frequently exhibits a 1/f spectrum. The spectral density of
the 1/f noise voltage is
Af
e2nf = , (1.31)
f
where the noise coefficient Af is device specific and of order 1010 – 1012 V2 .
A portion of the noise currents flows through the detector capacitance, re-
sulting in a frequency-dependent noise voltage in /ωCd , which is added to the
noise voltages in the input circuit. Since the individual noise contributions are
random and uncorrelated, they add in quadrature. The total noise at the output
of the pulse shaper is obtained by integrating over the full bandwidth of the
system.
1.10.2 Amplitude measurements
Since radiation detectors are typically used to measure charge, the system’s noise
level is conveniently expressed as an equivalent noise charge Qn , which is equal
to the detector signal that yields a signal-to-noise ratio of one. The equivalent
noise charge is commonly expressed in Coulombs, the corresponding number of
electrons, or the equivalent deposited energy (eV). For a capacitive sensor
C2
Q2n = i2n Fi TS + e2n Fv + Fvf Af C 2 , (1.32)
TS
where C is the sum of all capacitances shunting the input. Note that the voltage
noise contributions increase with capacitance. The shape factors Fi , Fv , and Fvf
depend on the shape of the pulse determined by the shaper. TS is a characteristic
time, for example the peaking time of a semi-Gaussian pulse (Figure 1.3) or the
prefilter integration time in a correlated double sampler (discussed in Chapter
4). The shape factors Fi , Fv are easily calculated,
Z∞ Z∞ 2
1 TS dW (t)
Fi = [W (t)]2 dt and Fv = dt , (1.33)
2TS 2 dt
−∞ −∞
where for time invariant pulse shaping W (t) is simply the system’s impulse re-
sponse (the output signal seen on an oscilloscope) with the peak output signal
104
EQUIVALENT NOISE CHARGE (e)

103 TOTAL
1/f NOISE
CURRENT NOISE VOLTAGE NOISE
2
10
0.01 0.1 1 10 100
SHAPING TIME (µs)
Fig. 1.29. Equivalent noise charge vs. shaping time of a typical detector system.
normalized to unity. For more details see the papers by Goulding (1972), Radeka
(1968, 1974), and Goulding and Landis (1982).
A pulse shaper formed by a single differentiator and integrator with equal
time constants τd = τi = τ ≡ TS as in Figure 1.5 has Fi = Fv = 0.9 and
Fvf = 4, independent of the shaping time constant. The overall noise bandwidth,
however, depends on the time constant, i.e. the characteristic time TS . The
contribution from noise currents increases with shaping time, i.e. pulse duration,
whereas the voltage noise decreases with increasing shaping time. Noise with a
1/f spectrum depends only on the ratio of upper to lower cutoff frequencies
(integrator to differentiator time constants), so for a given shaper topology the
1/f contribution to Qn is independent of TS . Increased detector capacitance
shifts the voltage noise contribution upward and the noise minimum to longer
shaping times. Pulse shapers can be designed to reduce the effect of current noise,
e.g. mitigate radiation damage. Increasing pulse symmetry tends to decrease Fi
and increase Fv (e.g. to 0.45 and 1.0 for a shaper with one CR differentiator and
four cascaded integrators).
For the circuit shown in Figures 1.27 and 1.28

4kT C2
Q2n = 2eId + + i2na Fi TS + 4kT Rs + e2na Fv d + Fvf Af Cd2 . (1.34)
Rb TS
As the shaping time TS is changed, the total noise goes through a minimum,
where the current and voltage contributions are equal. Figure 1.29 shows a typical
example. At short shaping times the voltage noise dominates, whereas at long
shaping times the current noise takes over. The noise minimum is flattened by
the presence of 1/f noise. Increasing the detector capacitance will increase the
voltage noise and shift the noise minimum to longer shaping times (Figure 4.29).
For quick estimates one can use the following equation, which assumes a field
effect transistor (FET) amplifier (negligible ina ) and a simple CR-RC shaper
with time constants τ (equal to the peaking time).
2 " #
2 e2 5 e kΩ τ 4 e2 ns C2
Qn = 12 Id τ + 6 · 10 + 3.6 · 10 2 2 e2n
nA · ns ns Rb (pF) (nV) /Hz τ
For a given amplifier (i.e. en ), noise is improved by reducing the detector ca-
pacitance and leakage current, judiciously selecting all resistances in the input
circuit, and choosing the optimum shaping time constant.
The noise parameters of the amplifier depend primarily on the input device.
Chapter 6 treats this in detail. In field effect transistors the noise current contri-
bution is very small, so reducing the detector leakage current and increasing the
bias resistance will allow long shaping times with correspondingly lower noise.
In bipolar transistors the base current sets a lower bound on the noise current,
so these devices are best at short shaping times. In special cases where the noise
of a transistor scales with geometry, i.e. decreasing noise voltage with increasing
input capacitance, the lowest noise is obtained when the input capacitance of the
transistor is equal to the detector capacitance, albeit at the expense of power
dissipation. Capacitive matching is useful with FETs, but not bipolar transis-
tors, as discussed in Chapter 6. In bipolar transistors the minimum obtainable
noise is independent of shaping time, but only at the optimum collector current
IC , which does depend on shaping time:
r
C p kT p Fv 1
Q2n,min = 4kT √ Fi Fv at IC = C βDC , (1.35)
βDC e F i TS
where βDC is the direct current gain. For a CR-RC shaper and βDC = 100,

e √ µA · ns C
Qn,min ≈ 250 √ · C at IC = 260 · . (1.36)
pF pF TS
Practical noise levels range from < 1 e for CCDs at long shaping times to
≈ 104 e in high-capacitance liquid argon calorimeters. Silicon strip detectors typ-
ically operate at ≈ 103 e, whereas pixel detectors with fast readout can provide
noise of order 100 e.
1.10.3 Timing measurements
In timing measurements the slope-to-noise ratio must be optimized, rather than
the signal-to-noise ratio alone, so the rise time tr of the pulse is important. The
“jitter” σt of the timing distribution
σn tr
σt = ≈ , (1.37)
(dS/dt)ST S/N
where σn is the rms noise and the derivative of the signal dS/dt is evaluated
at the trigger level ST . To increase dS/dt without incurring excessive noise the
PREAMPLIFIER SHAPER ANALOG PIPELINE ADC
TOKEN IN
ANALOG SIGNAL PROCESSING
SPARSIFICATION
ANALOG SIGNAL PROCESSING CONTROL
DIGITAL
ANALOG SIGNAL PROCESSING CONTROL DATA OUT
OUTPUT
ANALOG SIGNAL PROCESSING DRIVERS TOKEN OUT
ANALOG SIGNAL PROCESSING
TEST PULSE GENERATOR, DACs, R/ W POINTERS, etc.
Fig. 1.30. Circuit blocks in a representative readout IC. The analog processing chain
is shown at the top. Control is passed from chip to chip by token passing.
amplifier bandwidth should match the rise-time of the detector signal. The 10 –
90% rise time of an amplifier with bandwidth fu is
2.2 0.35
tr = 2.2τ = = . (1.38)
2πfu fu
For example, an oscilloscope with 350 MHz bandwidth has a 1 ns rise time. When
amplifiers are cascaded, which is invariably necessary, the individual rise times
add in quadrature q
tr ≈ t2r1 + t2r2 + . . . + t2rn . (1.39)
Thus, reducing the risetime of the electronics beyond the risetime of the sensor
signal will increase the electronic noise more rapidly than improve the signal
risetime. Time resolution improves with signal-to-noise ratio, so minimizing the
total capacitance at the input is also important. At high signal-to-noise ratios
the time jitter can be much smaller than the rise time. When a simple threshold
discriminator is used the timing signal will shift with pulse amplitude (“walk”),
but this can be corrected by various means, either in hardware or software.
Timing measurements are discussed in Chapter 4 and by Spieler (1982).
1.11 Subsystems
1.11.1 Circuit integration and bussing
A detector array combines the sensor and the analog signal processing circuitry
together with a readout system. Figure 1.30 shows the circuit blocks in a repre-
sentative readout IC. Individual sensor electrodes connect to parallel channels of
SUBSYSTEMS 37
STRIP DETECTOR
IC3 IC2 IC1
TOKEN
PASSING
CONTROL BUS
DATA BUS
Fig. 1.31. Multiple ICs are ganged to read out a strip detector. The right-most chip
IC1 is the master. A command on the control bus initiates the readout. When IC1
has written all of its data it passes the token to IC2. When IC2 has finished it
passes the token to IC3, which in turn returns the token to the master IC1.
analog signal processing circuitry. Data are stored in an analog pipeline pending
a readout command. Variable write and read pointers are used to allow simulta-
neous read and write. The signal in the time slot of interest is digitized, compared
with a digital threshold and read out. Circuitry is included to generate test pulses
that are injected into the input to simulate a detector signal. This is a very useful
feature in setting up the system and is also a key function in chip testing prior to
assembly. Analog control levels are set by digital-to-analog converters (DACs).
Multiple ICs are connected to a common control and data output bus, as shown
in Figure 1.31. Each IC is assigned a unique address, which is used in issuing
control commands for setup and in situ testing. Sequential readout is controlled
by token passing. IC1 is the master, whose readout is initiated by a command
(trigger) on the control bus. When it has finished writing data it passes the token
to IC2, which in turn passes the token to IC3. When the last chip has completed
its readout the token is returned to the master IC, which is then ready for the
next cycle. The readout bit stream begins with a header, which uniquely identi-
fies a new frame. Data from individual ICs are labeled with a chip identifier and
channel identifiers. Many variations on this scheme are possible. As shown, the
readout is event oriented, i.e. all hits occurring within an externally set exposure
time (e.g. time slice in the analog buffer in Figure 1.30) are read out together.
In colliding-beam experiments only a small fraction of beam crossings yields
interesting events. The time required to assess whether an event is potentially
interesting is typically of order microseconds, so hits from multiple beam cross-
ings must be stored on-chip, identified by beam crossing or time-stamp. Upon
receipt of a trigger the interesting data are digitized and read out. This allows
use of a digitizer that is slower than the collision rate. It is also possible to read
out analog signals and digitize them externally. Then the output stream is a se-
Fig. 1.32. Detector modules combining silicon strip sensors and electronics, shown
in a protective enclosure. Multiple modules are mounted on a stave that also in-
corporates the signal and power busses. The right-hand panel shows the ceramic
hybrid that combines six readout ICs (top row) with the power supply and data
bussing. Bypass capacitors are visible as small rectangles beneath the ICs. A flex
ribbon cable connects the hybrid to the control and data acquisition system and
also provides power. (Photographs courtesy of M. Garcia-Sciveres and C. Haber.)
quence of digital headers and analog pulses. An alternative scheme only records
the presence of hits. The output of a threshold comparator signifies the presence
of a signal and is recorded in a digital pipeline that retains the crossing number.
When reading out pulse height information, either in analog or digitized
form, the “smearing” of pulse height information by electronic noise will be
clearly visible in the output data. In a “binary” readout the presence of noise
is not so obvious. With a large signal-to-noise ratio the threshold can be set
high, so predominantly true hits will appear in the output stream. However,
when sensitivity is of the essence, the threshold will be set as low as possible.
If set too low, the comparator will fire predominantly on noise pulses. If set
too high, noise hits are suppressed, but efficiency for desired events will suffer.
Thus, a compromise threshold is chosen that will provide high efficiency with an
acceptable rate of noise hits. In any case, since the “tails” of the noise distribution
extend to infinity, the output of every binary system is contaminated by noise
pulses. Only the ratio of noise hits to true hits will be different and depend on
the signal-to-noise ratio. This is discussed quantitatively in Chapter 4.
1.11.2 Detector modules, services, and supports

The outputs of the ICs must be transferred to the off-detector electronics. To
provide this interface the readout ICs are mounted on a substrate, which accom-
modates the signal bussing between ICs, control signals, and power supply busses.
SUBSYSTEMS 39
Important components are bypass capacitors and filter networks to block exter-
nal interference from the readout ICs, but also to keep digital switching spikes
from propagating through the power supply lines. Figure 1.32 shows an assembly
of multiple detectors with readout circuitry, mounted on a stave that also inte-
grates the data and power busses. The right-hand panel of Figure 1.32 shows the
electronics unit (called a “hybrid”, as it combines multiple technologies). Each
integrated circuit includes 128 channels of front-end circuitry, analog pipeline,
analog-to-digital conversion, and readout logic and driver with on-chip zero sup-
pression (sparsification), so that only struck channels are read out. This hybrid
utilizes a multilayer ceramic substrate to integrate the readout ICs, associated
capacitors, and interconnects. Power, control, and data lines are implemented as
polyimide ribbon cables. Figure 1.33 shows a closeup of ICs mounted on a hy-
brid using a flexible kapton substrate (Kondo et al. 2002), described in Chapter
8 (Section 8.6.5). The wire bonds connecting the IC to the hybrid are clearly
visible. Channels on the IC are laid out on a ∼ 50 µm pitch and pitch adapters
fan out to match the 80 µm pitch of the strip detector. The space between chips
accommodates bypass capacitors and connections for control busses carrying
signals from chip to chip. Other examples are discussed in Chapter 8.
In large systems optical links are often chosen to eliminate cross-coupling
from other lines, but properly designed differential cable drivers and receivers
can also provide high noise immunity. Optical links require additional interface
Fig. 1.33. Close-up of ICs mounted on a hybrid utilizing a flexible polyimide sub-
strate (Kondo et al. 2002). The high-density wire bonds at the upper edges connect
via pitch adapters to the 80 µm pitch of the silicon strip detector. The ground
plane is patterned as a diamond grid to reduce material. (Photograph courtesy of
A. Ciocio.)
ICs, as silicon is not well-suited for optical emitters. In some designs all drivers
are accommodated on the detector module, whereas in others multiple detector
modules interface through a common driver module.
1.11.3 Data acquisition

Signals are bussed from the detector modules to a readout module, which in-
cludes data buffering. Depending on the complexity of the systems, several al-
ternatives exist for interfacing to the data acquisition computer. Small systems
can interface directly via a plug-in card in a PC. In large systems VME or PCI
bus interfaces are frequently used. In the past the nuclear instrumentation com-
munity designed interfaces such as CAMAC or FastBus, but now the availability
of suitable standard industry interfaces has displaced these community-specific
interfaces. The interface modules that accept the data from the detector module
and transfer it to the computer data acquisition bus are usually custom designed
and typically contain buffer memory, FPGA-based logic, and local processors
to preprocess the data. Digital interfacing utilizes standard techniques known
to many engineers and scientists, so it will not be covered in this book. For an
overview of data acquisition systems in high-energy physics see Butler (2003).
1.12 Further reading

The following chapters go into detail on the topics laid out in this introduction,
but emphasize aspects relevant to large scale semiconductor systems. Books by
G. Knoll (2000) and C. Grupen (1996) provide excellent general introductions to
radiation detection and techniques. S.M. Sze (1981) gives a concise description of
semiconductor physics and a comprehensive treatment of semiconductor devices.
The following chapters and appendices include references to more specialized
texts.
References
Abe, K. et al. (1997). Design and performance of the SLD vertex detector: a
307 Mpixel tracking system. Nucl. Instrum. and Meth. A400 (1997) 287–343
Barberis, E. et al. (1994). Capacitances in silicon microstrip detectors. Nucl.
Instr. and Meth. A342 (1994) 90–95
Bedeschi, F. et al. (1989). CDF silicon detector prototype test beam results.
IEEE Trans. Nucl Sci. NS-36 (1989) 35–39
Butler, J. (2003). Triggering and data acquisition general considerations, in
Instrumentation in Elementary Particle Physics, AIP Conf. Proc. 674 (2003)
101–129
Demaria, N. et al. (2000). New results on silicon microstrip detectors of CMS
tracker. Nucl. Instr. and Meth. A447 (2000) 142–150
Deptuch, G. et al. 2003. Development of monolithic active pixel sensors for
charged particle tracking. Nucl. Instr. and Meth. A511 (2003) 240–249
Fossum, E.R. (1997). CMOS image sensors: electronic camera on a chip. IEEE.
Trans. Electron Dev. ED-44/10 (1997) 1689–1698
REFERENCES 41
Gatti, E. and Rehak, P. (1984). Semiconductor drift chamber – an application

of a novel charge transport scheme. Nucl. Instr. and Meth. 225 (1984) 608–614
Goulding, F.S. (1972). Pulse shaping in low-noise nuclear amplifiers: a physical
approach to noise analysis. Nucl. Instr. and Meth. 100 (1972) 493–504
Goulding, F.S. and Landis, D.A. (1982). Signal processing for semiconductor
detectors. IEEE Trans. Nucl. Sci. NS-29/3 (1982) 1125–1141
Grupen, C. (1996). Particle Detectors. Cambridge University Press, Cambridge.
ISBN 0-521-55216-8, QC787.G6G78
Holland, S. and Spieler, H. (1990). A monolithically integrated detector–pre-
amplifier on high-resistivity silicon. IEEE Trans. Nucl. Sci. NS-37 (1990)
463–468
Holland, S. (1992). Properties of CMOS Devices and circuits fabricated on high-
resistivity, detector-grade Silicon. IEEE Trans. Nucl. Sci. NS-39/4 (1992)
809–813
Kemmer, J. and Lutz, G. (1987). New detector concepts. Nucl. Instr. and Meth.
A253 (1987) 365–377
Kenney, C. et al. (1993). Performance of a monolithic pixel detector. Nucl. Phys.
B (Proc. Suppl.) 32 (1993) 460 and First test beam results from a monolithic
silicon pixel detector. Nucl. Instrum. and Meth. A326 (1993) 144–149
Kleinfelder, S.A. et al. (1988). A flexible 128 channel silicon strip detector in-
strumentation integrated circuit with sparse data readout. IEEE Trans. Nucl.
Sci. NS-35 (1988) 171–175
Kleinfelder, S.A. et al. (2004). Novel integrated CMOS sensor circuits. IEEE
Trans. Nucl. Sci. NS-51/5 (2004) 2328–2336
Knoll, G.F. (2000). Radiation Detection and Measurement(3rd edn). Wiley, New
York, ISBN 0-471-07338-5, QC787.C6K56 1999
Kondo, T. et al. (2002). Construction and performance of the ATLAS silicon
microstrip barrel modules. Nucl. Instr. and Meth. A485 (2002) 27–42
Lüth, V. (1990). Spatial Resolution of Silicon Detectors. SLAC BaBar Note #54
9-7-90
Lutz, G. (1999). Semiconductor Radiation Detectors. Springer Verlag, Berlin,
1999. ISBN 3-5406-4859-3
Particle Data Group (2004). Review of Particle Physics. Phys. Lett. B592
(2004) 1–1109 and at http://pdg.lbl.gov
Radeka, V. (1968). Optimum signal-processing for pulse-amplitude spectrom-
etry in the presence of high-rate effects and noise. IEEE Trans. Nucl. Sci.
NS-15/3 (1968) 455–470
Radeka, V. (1974). Signal, noise and resolution in position-sensitive detectors.
IEEE Trans. Nucl. Sci. NS-21 (1974) 51–64
Rehak, P. et al. (1985). Semiconductor drift chambers for position and energy
measurements. Nucl. Instrum. and Meth. A235 (1985) 224–234
Snoeys, W. et al. (1992). A new integrated pixel detector for high energy
physics. IEEE Trans. Nucl. Sci. 39 (1992) 1263–1269
Spieler, H. (1982). Fast timing methods for semiconductor detectors. IEEE

Trans. Nucl. Sci. NS-29 (1982) 1142–1158
Sze, S.M. (1981). Physics of Semiconductor Devices(2nd edn). Wiley, New York
1981. ISBN 0-471-05661-8, TK7871.85.S988
Turchetta, R. et al. (2003). Monolithic active pixel sensors (MAPS) in a VLSI
CMOS technology. Nucl. Instr. and Meth. A501 (2003) 251–259
Wood, M.L. et al. (1991). Charge correlation measurements of double-sided
direct-coupled silicon microstripe detectors. Supercollider 3 (J. Nonte, ed).
Proc. 3rd Annual International Industrial Symposium on the Super Collider,
Atlanta, Georgia, Mar 13–15, 1991. pp. 903–926. Plenum Press, New York.
ISBN 0-3064-4037-7, QC787.P7.I57

Semiconductor Detector Systems Chapter-1 PDF

Uploaded by

Copyright:

Available Formats

Semiconductor Detector Systems Chapter-1 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Semiconductor Detector Systems Chapter-1 PDF

Uploaded by

Copyright:

Available Formats

1

DETECTOR SYSTEMS OVERVIEW

INCIDENT SENSOR PREAMPLIFIER PULSE ANALOG TO DIGITAL

SENSOR PULSE SHAPER OUTPUT

1.3 Pulse shaper

SENSOR CURRENT INTEGRATOR HIGH-PASS FILTER LOW-PASS FILTER

analog-to-digital converters (ADCs) measure voltage or current amplitude. How-

Fig. 1.7. A conceptually simple technique for analog-to-digital conversion utilizes a

hibitive. However, in monolithically integrated circuits it is quite feasible, but

1.5 Electro–mechanical integration

is inferred by reconstructing the trajectories of the secondaries and detecting the

radius as possible. The obtainable impact parameter resolution improves rapidly

where p⊥ is the particle momentum, x the thickness of the material, and X0

1.6 Sensor structures I

Fig. 1.9. Charge collection in a simple ionization chamber.

1.6.2 Position sensing

Fig. 1.11. Subdividing an electrode into strips provides one-dimensional position

READOUT WIRE-BOND PADS FOR

1.7 Sensor physics

1.7.2 Sensor volume

OXIDE JUNCTION CONTACT GUARD RING

n BULK OHMIC CONTACT p+ DOPING

A reverse bias voltage Vb yields the depletion width

For bias voltages Vb  Vbi

Typically, the interstrip capacitance is about 1 pF/cm. The backplane capaci-

1.7.3 Charge collection

In practice the dopant concentration N of silicon wafers is expressed as the

for n-type material and

1.7.4 Energy resolution

-15 -10 -5 5 10 15 [mm]

Css Css Css Css

The range of charge interpolation can be extended by introducing interme-

ABSORPTION COEFFICIENT (cm-1)

However, given the maximum practical silicon detector thickness of several mm

1.8 Sensor structures II – monolithic pixel devices

1.8.1 Charge coupled devices

1.8.2 Silicon drift chambers

1.8.3 Monolithic active pixel sensors

p-WELL WITH CIRCUITRY

n WELL (~1 µm)

p EPI (~1-10 µm)

plexity substantially, so the “real estate” occupied by the electronics increases,

1.10 Detection limits and resolution

In an optimized system, the time-scale of the fluctuations is comparable to

NORMALIZED COUNT RATE

1.10.1 Electronic noise

Fig. 1.27. Typical detector front-end circuit.

DETECTOR BIAS SERIES AMPLIFIER +

EQUIVALENT NOISE CHARGE (e)

CURRENT NOISE VOLTAGE NOISE

PREAMPLIFIER SHAPER ANALOG PIPELINE ADC

TEST PULSE GENERATOR, DACs, R/ W POINTERS, etc.

IC3 IC2 IC1

1.11.2 Detector modules, services, and supports

1.11.3 Data acquisition

1.12 Further reading

Gatti, E. and Rehak, P. (1984). Semiconductor drift chamber – an application

Spieler, H. (1982). Fast timing methods for semiconductor detectors. IEEE

You might also like

For bias voltages Vb Vbi