Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-1

Chapter 6 Pulse Processing

6.1. Introduction
Most radiation detectors require pulse (or signal) processing electronics so that energy or time
information involved with radiation interactions can be properly extracted. The objective of this
chapter is to study and understand the general principles of pulse processing in radiation
detection. There are two types of signal pulses in radiation measurements: linear and logic
pulses. A linear pulse is a signal pulse carrying information through its amplitude and shape. A
logic pulse is a signal pulse of standard size and shape that carries information only by its
presence or absence. Generally, linear pulses are produced by radiation interactions and then
converted to logic pulses.

It has become common practice to manufacture most pulse processing electronics in standard
modules so that they can fit into a housing called a bin or crate, which occupies 19 inch width.
Most popular standards are the Nuclear Instrument Module (NIM) and Computer Automated
Measurement and Control (CAMAC). Commercial modular electronics are usually manufactured
according to one of these standards. The convenient features of these standards are:

- The basic dimensions for the bin and modules are specified.
- Only the bin is connected to the ac power and generates all the dc supply voltages required by
modules contained within that bin.
- The connector interface between the module and bin is standardized both electrically and
mechanically.
- Specifications are included for the polarity and span of both linear and logic pulses.

6.2. Common components

A. Preamplifier
A preamplifier is the first component in a signal processing chain of a radiation detector. The
charge created within a detector is collected by the preamp. In spite of its name, the preamp does
not act as an amplifier (just means “before”, i.e. “pre” the amplifier), but acts as an interface
between the detector and the pulse processing electronics that follow. The main function of a
preamplifier is to extract the signal from the detector without significantly degrading the intrinsic
signal-to-noise ratio. Therefore, the preamplifier is located as close as possible to the detector,
and the input circuits are designed to match the characteristics of the detector. Two important
requirements of the preamp are:
- To terminate the capacitance quickly to maximize the signal-to-noise ratio.
The cable length between the preamp and the detector is maintained at minimum as well due to
the same reason.
- To have a low output impedance: i.e. to provide a low impedance source for the amplifier. Of
course, it should provide a high impedance load for the detector.
The schematic diagram of a RC feedback charge sensitive preamp is shown in Fig. 6.1. The
detector high voltage bias is fed through the preamp in general except for scintillation detectors.
When a preamp is ac-coupled as shown in Fig. 6.1, a single cable is connected between the
preamp and the detector, and is used for both high voltage bias to the detector and signal
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-2

extraction from the detector. A coupling capacitor should be provided between the detector and
preamp circuits in this configuration while it is eliminated if the dc-coupled configuration is
adopted instead.

Fig. 6.1. Typical RC feedback charge sensitive preamp.

Fig. 6.2. Signal pattern of a resistive
feedback charge sensitive preamp.

In the charge sensitive preamp, charge from the detector is collected on the feedback capacitor
Cf over a period of time, effectively integrating the detector current pulse. As the charge is
collected the voltage on the feedback capacitor rises, producing a step change in voltage. The
output voltage is then proportional to the total integrated charge as long as the time constant RfCf
is sufficiently longer than the duration of the input pulse. In other words, the output pulse height
is in proportion to the energy deposited by a radiation interaction. As shown in Fig. 6.2, the
output pulse shape is characterized by a short rise time determined by the charge collection
characteristics of the detector, and long decay time (~ 100 s).

In normal operation at ordinary counting rates, the rising step caused by each detector event rides
on the exponential decay of a previous event due to the long decay time, and the preamp output
does not get a chance to return to the baseline. This does not create a serious problem since the
significant information of the output pulse is in its rising edge and the shaping amplifier is
capable of extracting the pulse height from the rising edge of each pulse. As the counting rate
increases, the pile up of pulses on the tails of previous pulses increases, and the excursions of the
preamp output move farther away from the baseline. The dc power supply voltages eventually
limit the excursions, and determine the maximum counting rate that can be tolerated without
distortion of the output pulses. When the maximum counting rate condition is met, the preamp
becomes saturated and no pulses will be output. The second limitation is that the feedback
resistor Rf has an intrinsic noise (Johnson noise) associated with it. The noise can be minimized
by selecting a higher Rf value, which is limited because

- Simple increase of Rf  may lead to too long a time constant

- Keeping time constant by reducing Cf  affects linearity of the preamp
The two shortcomings of the RC feedback preamp can be potentially relieved if the feedback
resistor Rf is eliminated. Without the feedback resistor, the pulses of charge from the detector are
simply accumulated on the feedback capacitor. As shown in Fig. 6.3, the output voltage then
grows in staircase pattern with each upward step corresponding to a separate pulse. Some
method must be provided to reset the preamp when the staircase approaches the maximum
allowable voltage. A popular way is to reset with a transistor reset circuit as in a transistor reset
preamp.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-3

Fig. 6.3. The output voltage of a transistor reset preamp.

B. Detector bias supply and pulse generators

Radiation detectors require the application of an external high voltage for proper operation and
high voltage supplies used for this purpose are conventionally called detector bias supplies. The
important characteristics of detector bias supplies are:
- The maximum voltage level and its polarity
- The maximum current available from the supply
- The degree of regulation against long-term drifts due to changes in temperature or line voltage
- The degree of filtering to eliminate ripple at power line frequency or other sources.
In the case of PMTs connected to scintillation detectors, bias supplies should cover up to 3 kV
with a current of a few mA. The high voltage output must also be well regulated to prevent gain
shift in the PMT. High voltage supplies for germanium semiconductor detectors may cover up to
5 kV for large size models.

An electronic pulse generator (or pulser) is always required in the initial setup and calibration of
radiation spectroscopy systems. A tail pulse generator with adjustable rise and decay times is the
most common and its output is fed to the test input on preamps so that the pulser output can be
used for adjusting and testing shaping or timing parameters. If the output amplitude is constant
with a high accuracy, the amplitude distribution measured by a pulse analysis system gives the
electronic noise level present in the system. For normal pulsers, the interval between pulses is
uniform and periodic. However, some tests require random generation of pulses. Randomly
spaced pulses can be generated using either the noise signal from an internal component or an
external trigger pulse derived from a random signal source like another radiation detector.

6.3. Pulse counting systems

A. Integral discriminator
In order to count the pulses reliably, the preamp output signal must be shaped and amplified by a
shaping amplifier and then the shaped linear pulses must be converted into logic pulses. The
integral discriminator is the simplest unit that does this operation and consists of a device that
produces a logic output pulse only when the linear input pulse height exceeds a threshold, i.e.
discriminator level. The logic output pulse is normally produced shortly after the leading edge of
the linear pulse crosses the discriminator level. Integral discriminators are designed to accept
shaped linear pulses in the 0 ~ 10 V range.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-4

B. Differential discriminator (single channel analyzer)

A single channel analyzer (SCA) produces a logic output
pulse only when the linear input pulse height lies between two
independent discriminator levels. In some units, the lower
level discriminator (LLD) and upper level discriminator
(ULD) are independently adjustable while in other units, the
LLD is labeled the E level and the window width or difference
between levels is labeled E.
Fig. 6.4. SCA principle.
In normal SCAs, the time of appearance of the logic pulse is not closely coupled to the actual
event timing, and therefore these logic pulses will give imprecise results when used for timing
measurements. Timing SCAs are made so that the logic pulse can be much more closely
correlated with the actual arrival time of the linear pulse by incorporating the time pick-off
methods described later in this chapter.

C. Counter and timer

As the final step in a counting system, the logic pulses must be accumulated and their number
recorded over a period of time. A counter is used for this purpose and increments one count each
time a logic pulse is presented to its input.

Counters are operated in one of two modes usually: preset time or preset count. In the preset time
mode, the counting period is controlled by an internal or external timer. In the preset count mode,
the counter accumulates pulses until the total number of counts reaches the preset value of counts.
The function of a timer is simply to start and stop the accumulation period for an electronic
counter or other recording device.

6.4. Pulse height analysis systems

A. Shaping (spectroscopy) amplifier
For pulse-height or energy spectroscopy, the linear, pulse shaping amplifier performs several
essential functions. Its primary role is to magnify the amplitude of the preamplifier output pulse
from the mV range into the 0.1 ~ 10 V range. This makes possible accurate pulse height
measurements with a peak-sensing analog-to-digital converter (ADC) or SCA. In addition, the
amplifier shapes the pulses to optimize the energy resolution, and to minimize the risk of overlap
between successive pulses. Most amplifiers also incorporate a baseline restorer to ensure that the
baseline between pulses is held rigidly at ground potential in spite of changes in counting rate or
temperature.

Frequently, the requirement to handle high counting rates is in conflict with the need for
optimum energy resolution. For most radiation detectors, achieving the optimum energy
resolution requires long pulse widths. On the other hand, short pulse widths are essential for high
counting rates. In such cases, a compromise pulse width must be selected so that the
spectroscopy system can be optimized. In this section, various techniques available for pulse
shaping in the linear amplifier are described.

The linear, pulse-shaping amplifier must accept the output pulse shapes provided by the
preamplifier and change them into the pulse shapes suitable for optimum energy spectroscopy.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-5

The output for each pulse consists of a rapidly rising step, followed by a slow exponential decay.
It is the amplitude of the step that represents the energy of the detected radiation.

Fig. 6.5. Output pulse shapes from a RC feedback preamp (a) and a shaping amplifier (b).

Before amplification, the pulse-shaping amplifier must replace the long decay time of the preamp
output pulse with a much shorter decay time. Otherwise, the acceptable counting rate would be
seriously restricted. Fig. 6.5 demonstrates this function using the simple example of a shaping
amplifier. The energy information represented by the amplitudes of the steps from the
preamplifier output has been preserved, and the pulses return to baseline before the next pulse
arrives. This makes it possible for a peak sensing ADC to determine the correct energy by
measuring the pulse amplitude with respect to the baseline. With the shorter pulse widths at the
amplifier output, much higher counting rates can be tolerated before pulse pile-up causes
significant distortion in the measurement of the pulse heights above baseline.

Delay-line pulse shaping

Shaping amplifiers employing delay-line pulse shaping are well suited to the pulse processing
requirements of scintillation detectors. The propagation delay of distributed or lumped delay
lines can be combined into suitable circuits to produce an essentially rectangular output pulse
from each step-function input pulse. For pulse pile-up prevention, this shaping method is close to
ideal because an immediate return to baseline is obtained.

With scintillation detectors, the signal-to-noise ratio of the preamplifier and amplifier
combination is seldom a limitation on the energy resolution. The energy resolution of
scintillation detectors is governed by the statistics of scintillation light production and the
conversion to photoelectrons at the photo cathode. However, for detectors having no internal
gain, delay-line shaping is not appropriate, because the signal-to-noise ratio is inferior to that
obtained with other shaping methods.

There are many circuits that can be used for delay-line shaping, and Figs. 6.6 and 6.7 are typical
examples. The step pulse from the preamplifier is inverted, delayed, and added back to the
original step pulse. The result is a rectangular output pulse with a width equal to the delay time
of the delay line. In practice, the value of the resistor labeled 2RD is made adjustable over a small
portion of its nominal value to allow compensation for the exponential decay of the input pulse.
With proper adjustment, the output pulse will return to baseline promptly without undershoot.
The main advantage of delay line shaping is a rectangular output pulse with fast rise and fall
times. In fact, the falling edge of the pulse is a delayed mirror image of the rising edge. These
characteristics make delay line pulse shaping ideal for timing and pulse-shape discrimination
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-6

applications with scintillation detectors at low or high counting rates.

By following one delay-line shaper with a second, a doubly differentiated delay-line shape is
obtained, as illustrated in Fig. 6.7. The result is an output pulse with a bipolar shape. The
double-delay-line (DDL) shaping is ideal for scintillation detectors with high counting rates.
The baseline shift caused by high counting rates is eliminated in the bipolar shape. This benefit is
gained at the expense of doubling the pulse width.

DDL shaping is often useful for simple zero-crossover timing with scintillation detectors.
Double-delay-line shaping is not a good choice for detectors having a substantial preamplifier
noise. Its signal-to-noise ratio is worse than single-delay-line shaping, and much worse than
semi-Gaussian shaping.

Fig. 6.6. Single delay line (SDL) shaping and Double delay line (DDL) shaping.

CR-RC pulse shaping

The simplest concept for pulse shaping is the use of a CR circuit followed by an RC circuit. In
this shaping, the preamp signal first passes through a CR shaper and then RC shaper. Fig. 6.7
shows the circuit diagrams. Although this elementary filter is rarely used, it encompasses the
basic concepts essential for understanding the higher-performance, active filter techniques.

Fig. 6.7. CR and RC filters.

For circuit analysis, the Laplace and Fourier transformations are essential. The fundamentals of
these transforms are as follows.

 Laplace trasnform: defined as L[ f (t )]  F ( s )   f (t )e st dt for a time function f(t)
0
df (t ) t 1
Time differentiation and integration: L[ ]  sL[ f (t )]  f (0) , L[  f (t )dt ]  L[ f (t )]
dt 0 s
The Laplace transforms for some functions are given at the end of the chapter.

 Fourier trasnform: F [ f (t )]  F (i )   f (t )e i t dt  a special case of the Laplace transform !
0
 Transfer function: In general, the input-output relation of a linear circuit device is described by
a differential equation
dVout (t ) d nVout (t ) dVin (t ) d mVin (t )
a0Vout (t )  a1      an  b V
0 in (t )  b1      bm
dt dt n dt dt m
(Vin(t): input, Vout(t): output)
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-7

Taking the Laplace transform gives

b0  b1s  b2 s 2    bm s m
Vout ( s )  Vin ( s )  G ( s )Vin ( s ) ( Vout ( s )  L[Vout (t )],Vin ( s)  L[Vin (t )] )
a0  a1s  a2 s 2    an s n
The function G(s) is called the transfer function of the device. As the Laplace transform of the
delta function is “1”, the transfer function G(s) can be alternatively defined as the impulse (delta
function) input response. The time domain output vout(t) can be obtained through the inverse
transform as
Vout (t )   G (t  t )Vin (t ) dt  ( G (t )  L1[G ( s)] )

In general, the function G(s) can be converted into

( s  z1 )( s  z 2 )    ( s  z m )
G(s) 
( s  p1 )( s  p2 )    ( s  pn )
The roots s = p1, p2, … pn are defined as the poles and s = z1, z2, … zm are defined as the zeros of
G(s). For a cascade of circuit components and a system with a feedback component, the transfer
functions can be obtained by
Vin Vout Vin(s) E(s) Vout(s)
G1(s) G2(s) G3(s) G1(s)
+
- G1 ( s )
G(s) 
Vout ( s ) G2(s) 1  G1 ( s )G2 ( s )
G ( s)   G1 ( s )G2 ( s )G3 ( s ) B(s)
Vin ( s )
Fig. 6.8. Transfer functions for a cascade of components and a feedback case.
 Responses of the CR and RC shapers
1) CR shaping
Q(t ) dVin (t ) i (t ) dVout (t )
Vin (t )   Vout (t )   
C dt C dt
dV (t ) dV (t )
by Vout (t )  i (t ) R and   RC ,  in  Vout (t )   out
dt dt
Assuming the zero initial condition, taking Laplace transform leads to
s
Vout ( s )  Vin ( s )  GCR ( s )Vin ( s )
1  s
As the signal from the preamp has a long time constant and the shaped pulse width is much
smaller than the preamp time constant, we will neglect the preamp signal decay and assume the
preamp signal as a step function for convenience.

For the step function input

V0 (t  0) V0
Vin (t )   Vin ( s)  L[Vin (t )] 
0 (t  0) s
the output signal becomes

Vout ( s )  V0  Vout (t )  V0 e t / 
1  s
To figure out the noise filtering performance of the CR shaper, the transfer function in the
frequency (Fourier) domain is required. This function can be obtained by replacing s with i
from the Laplace domain transfer function GCR(s):
i 
GCR (i )   GCR (i ) 
1  i 1   2 2
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-8

2) RC shaping
dQ (t ) dV (t )
Vin (t )  i (t ) R  Vout (t ) and i (t )   C out
dt dt
dVout (t ) 1
 Vin (t )    Vout (t )  Vout ( s )  Vin ( s )  GRC ( s )Vin ( s )
dt 1  s
Output signal for the step function input:
1 V0
Vout ( s )   Vout (t )  V0 (1  e t /  )
1  s s
Frequency domain transfer function:
1 1
GRC (i )   GRC (i ) 
1  i 1   2 2
1.0 1.0

RC CR
0.8 0.8

CR

abs [G(i)]
0.6 0.6
Vout(t)/Vo

RC
CR-RC
0.4 0.4
CR-RC

0.2 0.2

0.0 0.0
0 1 2 3 4 5 0.01 0.1 1 10 100
t/ 

Fig. 6.9. CR, RC and CR-RC time domain Fig. 6.10. Absolute value of the frequency
responses for a step function input. domain transfer function for CR, RC and CR-RC.

The time domain responses of the CR and RC shapers for a step function input are shown in Fig
6.9. By the CR filter, the decay time of the pulse is shortened. If the time constant is made
sufficiently small, the output voltage is almost proportional to the time derivative of the input
wave form (CR differentiator). The RC filter makes the rising edge of the pulse stretched. If the
time constant is made sufficiently large, the output signal is almost proportional to the integral of
the input signal (referred to as a RC integrator).

The absolute values of the frequency domain transfer functions are shown as a function of the
frequency in Fig. 6.10. At the frequency 0 = 1/, the output signal is about 1/2  0.71 level of
the input signal for both CR and RC filters. The CR filter attenuates low frequency signals ( <
0) while passes the high frequency signals (high-pass filter). This improves the signal-to-noise
ratio by attenuating the low frequencies, which contain a lot of noise and very little signal. The
RC filter shows the opposite trend (low-pass filter). The RC filter improves the signal-to-noise
ratio by attenuating high frequencies, which contain excessive noise.

To attenuate both low and high frequency noises and make the pulse shape convenient for
analysis, both CR and RC filters must be employed. Based on individual transfer functions, the
transfer function of the CR-RC filter is given by
1  1s
GCR RC ( s )  (  1 : CR time constant,  2 : RC time constant)
(1   2 s ) (1   1s )
The time domain output signal for the step input is obtained as
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-9

V0 1
Vout (t )  ( e t /  1  e t /  2 )
1   2
Typically, the differentiation time constant is set equal to the integration time constant. In that
case, the frequency domain transfer function become
i 
GCR RC (i )   GCR  RC (i ) 
(1  i ) 2
1   2 2
As shown in Fig. 6. 10, both high and low frequency noises are efficiently filtered out by the CR-
RC filter. The output response for a step input is given by
V0
Vout (t )  te  t / 

The time constant  is called shaping time. The output pulse rises slowly and reaches its
maximum at τ. This time interval, the time taken for the signal leading edge to rise from zero to
maximum, is defined as the peaking time. Another conveniently used time interval is the rise
time, which is defined as the time taken for the signal leading edge to rise from 10 to 90% of
maximum.

For semiconductor detectors, the electronic noise at the preamplifier input makes a noticeable
contribution to the energy resolution. This noise contribution can be minimized by choosing the
appropriate amplifier shaping time constant. Fig. 6.11 shows the effect. At short shaping time
constants, the series noise component of the preamplifier is dominant. This noise is typically
caused by thermal noise in the channel of the field-effect transistor, which is the first amplifying
stage in the preamplifier. At long shaping time constants the parallel noise component at the
preamplifier input dominates. This component arises from noise sources that are effectively in
parallel with the detector at the preamplifier input (e.g., detector leakage current, thermal noise in
the preamplifier feedback resistor).

The total noise at any shaping time constant is the square root of the sum of the squares of the
series and parallel noise contributions. Consequently, the total noise has a minimum value at the
shaping time constant where the series noise is equal to the parallel noise. This time constant is
called the noise corner time constant. The time constant for minimum noise will depend on the
characteristics of the detector, the preamplifier, and the amplifier pulse shaping network.

For silicon charged-particle detectors, the minimum noise usually occurs at time constants in the
range from 0.5 to 1 µs. Generally, minimum
noise for semiconductor detectors is achieved
at much longer time constants (in the range
from 4 to 20 µs). Such long time constants
impose a severe restriction on the counting
rate capability. Practically, energy resolution
is often compromised by selecting shorter
shaping time constants in order to handle
higher counting rates.

Fig. 6.11. Optimization of the shaping time for resolution.

Pole-zero cancellation (Tail cancellation)
Up to this point, we approximated the preamp output pulse as a step function, however, in a real
preamp pulse, the falling tail is a long exponential decay instead of a simple step function.
Consequently, there is a small amplitude undershoot starting at about 7τ. This undershoot decays
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-10

back to baseline with the long time constant of the preamplifier. At medium to high counting
rates, a substantial fraction of the amplifier output pulses will ride on the undershoot from a
previous pulse. The apparent pulse amplitudes measured for these pulses will be significantly
lower, which deteriorates the energy resolution.

Most shaping amplifiers incorporate a pole-zero cancellation circuit to eliminate this undershoot.
The benefit of pole-zero cancellation is
improved peak shapes and resolution in the
energy spectrum at high counting rates. Fig.
6.12 illustrates the pole-zero cancellation
circuit, and its effect. The preamplifier signal
is applied to the input of the normal CR
differentiator circuit in the amplifier. The
output pulse from the differentiator exhibits
the undesirable undershoot. To cancel the
undershoot, the variable resistor Rpz is added
in parallel with the capacitor CD, and adjusted.
The result is an output pulse exhibiting a
simple exponential decay to baseline with the Fig. 6.12. Pole-zero cancellation circuit.
desired differentiator time constant. This circuit is termed a pole-zero cancellation circuit
because it uses a zero in the transfer function, expressed in Laplace transform, of the shaping
circuit to cancel a pole present in the input pulse. Exact pole-zero adjustment is critical for good
energy resolution. The circuit analysis can be done as follows.

By the current conservation,

d 1 1 1 1 1
C [Vin (t )  Vout (t )]  [Vin (t )  Vout (t )]  Vout (t )  [Cs   ] Vout ( s )  [Cs  ] Vin ( s )
dt RPZ R RPZ R RPZ
Vi
For the exponential function input Vin ( s )  , the output becomes
s  1/ i
1
s
Vi RPZ C
Vout ( s ) 
1 1 1
s  s
RPZ C  i
The cancellation requirement leads to the condition RPZ C D   i .
Semi-Gaussian pulse shaping
If a single CR high-pass filter is followed by several stages of RC integration, the output pulse
shape becomes close to Gaussian. Amplifiers shaping in this way are called semi-Gaussian
shaping amplifiers. Its output pulse is given by
t
Vout (t )  ( ) n e t / 

where, n represents the number of the integrators. The peaking time in this case is equal to n.
Hence, if the time constants are same, the peaking time of a CR-(RC)2 circuit is twice as long as
that for a simple CR-RC circuit. When the time constants are adjusted to make equal peaking
times, the CR-(RC)2 circuit gives the more symmetric shape as shown in Fig. 6.13, which makes
a faster return to the baseline. The signal-to-noise ratio of the CR-(RC)2 filter is also better.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-11

0.5
CR-RC ( = 4 s)
2
CR-(RC) ( = 4 s)
0.4 2
CR-(RC) ( = 2 s)
4
CR-(RC) ( = 1 s)
0.3
V(t)

0.2

0.1

0.0 Fig. 6.14. Semi-Gaussian shaping with active filters.

0 4 8 12 16 20 24
Time [s]

Fig. 6.13. Pulse shape of the CR-(RC)2 circuit.

The block diagram in Fig. 6.14 shows a more practical circuit for semi-Gaussian shaping. The
integrator components, i.e. (RC)n, are now replaced with active circuit components, which
incorporate transistors and diodes. The function of the active filter is similar to that of the passive
RC network.

Baseline restorer
To ensure good energy resolution and peak position stability at high counting rates, the higher-
performance spectroscopy amplifiers are entirely dc-coupled (except for the CR differentiator
network located close to the amplifier input). As a consequence, the dc offsets of the earliest
stages of the amplifier are magnified by the amplification gain to cause a large and unstable dc
offset at the amplifier output. A baseline restorer is required to remove this dc offset, and to
ensure that the amplifier output pulse rides on a baseline that is tied to ground potential.

Pile-up rejection
0.6 0.6
2 [s] shift

0.4 0.4 6 [s] shift

V(t)

V(t)

10 [s] shift

0.2 0.2

0.0 0.0
0 4 8 12 16 20 24 0 6 12 18 24
Time [s] Time [s]

Fig. 6.15. Pulse pile-up. Fig. 6.16. Various pile-up patterns for the CR-(RC)2
shaper with 2 s shaping time.
When two incident particles arrive at the detector within the width of the shaping amplifier
output pulse, their respective amplifier pulses pile up to form an output pulse of distorted height
(Fig. 6.15). Depending on the time difference between two pulses, the pile-up pattern
significantly changes as shown in Fig. 6.16. When the second pulse comes relatively late (6 s
and 10 s shift cases) and rides on the falling tail of the first pulse, the rising edge and the height
of the first pulse are not distorted, so that the first event can be processed without problem. In
contrast, when the second pulse arrives relatively early (2 s shift case) and rides on the rising
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-12

edge of the first pulse, the two pulses form a single pulse with a slower rising edge. In this case,
both detection events are discarded. A pile-up rejector can be used to prevent further processing
of these distorted pulses. Practically, it is implemented by adding a “fast” pulse shaping amplifier
with a very short shaping time constant in parallel with the “slow” shaping amplifier (Fig. 6.17).
In the fast amplifier, the signal-to-noise ratio is compromised in favor of improved pulse-pair
resolving time. A fast discriminator is set above the much higher noise level at the fast amplifier
output as shown in the figure (c). The falling edge of the fast discriminator output triggers an
inspection interval TINS that covers the slow amplifier pulse width TW.

Fig. 6.18. Influence of the pile-up on spectral

shape.
Fig. 6.17. Pile-up detection and rejection.

If a second fast discriminator pulse from a pile-up pulse arrives during the inspection interval, an
inhibit pulse is generated like (e). The inhibit pulse is used in the associated peak-sensing ADC
or multichannel analyzer to prevent analysis of the piled-up event. As demonstrated in Fig. 6.18,
the pile-up rejector can substantially reduce the pile-up continuum at high counting rates.

B. Peak sensing ADC and histogramming memory

A peak sensing analog-to-digital converter (ADC) measures the height of an analog pulse and
converts that value to a digital number. The digital output is a proportional representation of the
analog pulse height at the ADC input. For sequentially arriving pulses, the digital outputs from
the ADC are fed to a dedicated memory, or a computer, and sorted into a histogram. This
histogram represents the spectrum of input pulse heights. The dynamic range of the ADC
operation is consistent with the range of the spectroscopy amplifier output, i.e. 0 ~ 10 V.

Although the peak sensing ADC is mainly used for energy spectroscopy, it can be used for time
spectroscopy as well. When the output of a time-to-amplitude converter is connected to the ADC
input, the histogram represents the time spectrum measured by the time-to-amplitude converter.
The combination of the peak sensing ADC, the histogramming memory, and a display of the
histogram forms a multichannel analyzer (MCA). If a computer is employed to display the
spectrum, as usually done in these days, the combination of the ADC and the histogramming
memory is called a multichannel buffer (MCB).
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-13

Wilkinson ADC

Fig. 6.19. Signals in the Wilkinson ADC. Fig. 6.20. Wilkinson ADC operation.
The operation of the Wilkinson ADC is illustrated in Figs. 6.19 and 6.20. The lower-level
discriminator (LLD), is used to recognize the arrival of the amplifier output pulse. In general, the
LLD is set just above the noise level to prevent the ADC from spending time analyzing noise.
When the input pulse rises above the LLD, the input linear gate is open and the rundown
capacitor is connected to the input. Then, the capacitor is forced to charge up so that its voltage
follows that of the rising input pulse.
When the input signal has reached its maximum amplitude and begins to fall, the linear gate is
closed and the capacitor is disconnected from the input. At this point, the voltage on the
capacitor is equal to the maximum amplitude of the input pulse. Following detection of the input
pulse peak, the rundown capacitor is disconnected from the input and connected to a constant
current generator to keep a linear discharge of the capacitor voltage. At the same time, the
address clock with a high frequency is connected to the address counter and the clock pulses are
counted for the duration of the capacitor discharge. When the voltage on the capacitor reaches
zero, the counting of the clock pulses ceases.

Since the time for linear discharge of the capacitor is proportional to the original input pulse
height, the number Nc recorded in the address counter is also proportional to the pulse height.
Therefore, in the Wilkinson ADC, the A-D conversion time becomes longer as the input pulse
height increases. During the memory cycle, the address Nc is located in the histogramming
memory, and one count is added to the contents of that location. The value Nc is usually referred
to as channel number. The total number of channels is defined as conversion gain and is
selectable from 256 (for low resolution applications) to 16,384 channels (for high resolution
requirements).

The dead time of an MCA is composed of the ADC processing time and the memory cycle (or
storage) time. For the Wilkinson ADC, the processing time is variable depending on the input
pulse height as described above. The processing time is also dependent on the conversion gain
and the larger conversion gains require longer processing. The processing time per channel is the
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-14

period of the address clock. For a typical clock frequency of 100 MHz, this time is 10 ns per
channel. The dead time  of a Wilkinson ADC is given by
  N c / f c  Ts
The dead time depends on the clock frequency fc, the channel number Nc and the memory
storage time Ts. The storage time is 0.5 ~ 2 µs typically. The advantage of Wilkinson ADCs is
excellent linearity (nonlinearity < 1 %). The disadvantage is the long conversion time for large
A-D conversion gains.

Successive-Approximation ADC
Analog input
Analog
summing

DAC ADC

a b
Random number Digital
generator subtraction
b-a

Fig. 6.22. Sliding scale principle.

Fig. 6.21. Operation of a successive-approximation ADC.
The successive-approximation ADC is illustrated in Fig. 6.21. During the rise of the analog input
pulse, the switch S1 is closed and the voltage on capacitor C1 tracks the rise of the input signal.
When the input signal reaches maximum height, S1 is opened, leaving C1 holding the maximum
voltage of the input signal. After detection of the input pulse peak, the ADC begins its
measurement process.

First, the most significant bit of the digital-to-analog converter (DAC) is set. If the comparator
determines that the DAC output voltage is greater than the signal amplitude Vs, the most
significant bit is reset. If the DAC output voltage is less than Vs, the most significant bit is left in
the set condition. Subsequently, the same test is made by adding the next most significant bit.
This process is repeated until all bits have been tested. The bit pattern set in the register driving
the DAC at the end of the test is a digital representation of the analog input pulse height. This
binary number Nc is the address of the memory location to which one count is added to build the
histogram representing the pulse-height spectrum. If the ADC has n bits ( 2n channels), n test
cycles are required to complete the analysis, and this is the same for all pulse heights.

Although successive-approximation ADCs are available with the number of bits required for
high-resolution spectroscopy, their linearity is not good. This problem is overcome by adding the
sliding scale linearization as shown in Fig. 6.22. For each input signal, a random analog voltage
is generated and added to it before pulse height analysis. If the generated random number is m,
this results in the ADC reporting the analysis m channels higher than normal. By digitally
subtracting the number m at the output of the ADC, the digital representation is brought back to
its normal value. Due to its random nature, the added pulse averages the analysis of each input
pulse height over adjacent channels (typically, 256 channels or 8-bit) in the successive
approximation ADC. This improves the nonlinearity significantly ( < 1 %). The advantages of
the successive-approximation ADC with sliding scale linearization are low differential
nonlinearity, and a fast conversion that is independent of the pulse height.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-15

6.5. Digital pulse processing

With current analog pulse processing systems the preamp signal from the detector is shaped,
filtered and amplified by a shaping amplifier, and then digitized by a peak sensing ADC at the
very end of the analog signal processing chain. In digital pulse processing (DPP) systems, the
detector signal form is digitized with a sampling (or digitizing) ADC immediately after the
preamplifier. The digitized signal pulse is then shaped digitally and the pulse height is extracted.
The digital processor is the key element doing this operation and either a field programmable
gate array (FPGA) or a digital signal processor (DSP) can be employed. After extraction of the
pulse height, one count is added to the memory address corresponding to the pulse height as in
analog pulse processing.

Sampling Digital
Preamp
ADC processor
Fig. 6.23. Block diagram of the digital pulse processing.
DPP allows implementation of signal filtering functions that are not possible through traditional
analog signal processing. Digital filter algorithms require considerably less overall processing
time, so that the resolution remains fairly constant over a large range of count rates whereas the
resolution of analog systems typically degrades rather rapidly as the counting rate increases. As a
result, DPP will provide a much higher throughput without significant resolution degradation.
Improved system stability is another potential benefit of DPP techniques. The detector signal is
digitized much earlier in the signal processing chain, which minimizes the drift and instability
associated with analog signal processing.
Fig. 6.24 shows the output pulse of the preamplifier following a detection event and its digitized
form. Since the signal has been digitized, it is no longer continuous. Instead it is a string of
discrete values (Vin [1], Vin [2], …, Vin [i],…). The first step is to apply an appropriate filter as
in the analog pulse processing.

300 300
Preamp output [mV]
Preamp output [mV]

200 200

100 100

0 0

14 15 16 17 18 19 20 14 15 16 17 18 19 20
Time [s] Time [s]

Fig. 6.24. Signal pulse from a preamp and its digitized form.

The simplest digital filter would be the moving average filter, which is defined as
1 L 1
Vav1[i ]   Vin [i  j ]
L j 0
for the ith input data point. Since this filter takes an average value for a data length L, the high
frequency noise component is filtered out and the output is much smoother than the original data.
Fig. 6.25 shows an example of the moving average filter with L = 20.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-16

150
Raw data
Moving Average Filter (20 points)
100

V(t) [mV]
50

-50
16 18 20 22 24
Time [s]

Fig. 6.25. The moving average filter with L = 20.

Although the moving average filter partly attenuates the noise, its performance is not enough for
pulse height analysis compared to the analog pulse shaping. The shaped pulse has a long tail and
the low frequency noise components are not attenuated yet. A more popular type of the digital
filter for detector pulse processing is the trapezoidal filter. Its principle is as follows.

For the ith input data point Vin[i],

1) Compute the average value for the next L data points as shown in Fig. 6.26:
1 L1
Vav1[i ]   Vin [i  j ]
L j 0
The interval tL is the time interval corresponding to the data length L.
2) Make a separation gap of G data points and compute another average for the data length L.
1 L1
Vav 2 [i ]   Vin [ L  G  i  j ]
L j 0
The interval tG is the time interval corresponding to the data length G.
3) The output signal Vout[i] corresponding to the input point Vin[i] is computed by
Vout [i ]  Vav 2 [i ]  Vav1[i ]

150 150

tL
Preamp output [mV]

100 100
V(t) [mV]

50 50 tG
Vin[i]

0
tL tL
0
tL
tG
-50 -50
8 9 10 11 12 9 10 11 12 13
Time [s] Time [s]

Fig. 6.26. Trapezoidal filter parameters. Fig. 6.27. Trapezoidal filter output.

When this operation is applied to all input data points, the output pulse shape becomes
trapezoidal as the name implies (Fig. 6.27). The output pulse has a peaking time equal to tL, a
flat top equal to tG, and a symmetrical fall time equal to tL. The total width of the output pulse
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-17

is 2tL + tG. Hence, both tL and tG are employed as free parameters to adjust the output pulse
shape as the shaping time  is used in the semi-Gaussian analog shaping as a shaping parameter.
Fig. 6.28 shows three trapezoidal filter outputs with different shaping conditions.

Fig. 6. 29 compares a CR-(RC)2 analog filter with  = 0.5 s and an equivalent trapezoidal filter
with tL = 1.0 s and tG = 0.5 s. The peaking time of the CR-(RC)2 shaping is 2 and the
output pulses are not symmetric, so that the total pulse width is about 6 (3 times of the peaking
time). In contrast, the trapezoidal filter output pulse shows a sharp termination on completion of
its total width 2tL + tG. The flat top region of the trapezoidal pulse is helpful for enhancing the
detector charge carrier collection when the peaking time is shorter than the collection time of a
fraction of charge carriers. Moreover, the analog pulse processing chain requires another step for
pulse height analysis (i.e. AD conversion) while the digitized pulse height is already available in
the trapezoidal pulse. Consequently, the processing speed in DPP is faster and DPP is preferred
in high counting rate measurements.

Another attractive feature of the digital pulse processing is its simplicity and flexibility for pulse
timing. The arrival time or the detailed information on the rising part of a pulse is required when
the pulse processing system is intended for coincidence, pulse shape analysis, particle tracking
etc. In analog pulse processing, sophisticated timing modules must be set up in addition to the
pulse height analysis electronics. In the digital processing, the detailed waveform of the detection
signal is already digitized, so that timing operations can be implemented in the digital processor
without additional units. Moreover, the algorithms for timing or pulse shape analysis can be
modified easily.

150 150

tL=1 s, tG=0.5 s

tL=1 s, tG=1 s
100 100
V(t) [mV]
V(t) [mV]

2
CR-(RC) , = 0.5 s
50 50

0 0

tL=0.5 s, tG=0.5 s tL= 1 s, tG= 0.5 s

-50 -50
9 10 11 12 13 8 10 12 14 16
Time [s] Time [s]

Fig. 6.28. Trapezoidal filter outputs for different Fig. 6.29. Comparison between the CR-(RC)2 and
shaping parameters. the trapezoidal filters.
There is an important fact to remember. The DPP systems are not entirely digital. An analog
preamp is required to convert the detector charge to a voltage signal and some additional analog
front end conditioning is required to match the preamp output signal to the input of the sampling
ADC. Thus, the stability and integrity of the analog front end electronics is still important in
achieving good system performance.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-18

6.6. Pulse timing

In many applications, information on the accurate arrival time of a quantum of radiation in the
detector is required. For timing information, detector pulses are usually processed quite
differently than the pulse height analysis.

Time pick-off is the fundamental operation generating a logic pulse to indicate the time of
occurrence of an input linear pulse. The leading edge of the logic pulse corresponds to the time
of occurrence. Modular electronics doing this function are time pick-off units or triggers.

To achieve the best timing performance, an accurate time pick-off is most important. There are
two cases of inaccuracy in time pick-off: time jitter and amplitude walk. Time jitter is usually
induced by random fluctuations in the signal pulse size and shape. Amplitude walk is the effect
induced by the variable amplitudes of input pulses. The relative importance of two categories
depends on the dynamic range of the input pulse height.

Leading edge timing

Leading edge timing is the simplest time pick-off method and generates the output timing logic
pulse when an input pulse crosses a fixed discrimination level. This method is easy to implement
and is effective when the dynamic range of the input pulses is not large.

The effect of time jitter on leading edge timing is shown in Fig. 6.30. The random fluctuations
can make the output logic pulse be generated at different times with respect to the centroid of the
pulse. The amplitude walk in leading edge timing is illustrated in Fig. 6.31. The two pulses have
identical arrival times but their output logic pulses are significantly different in timing. If this
effect is too serious, the leading edge timing cannot be adopted for the applications requiring
high accuracy timing.

Fig. 6.30. Time jitter in leading edge timing.

Fig. 6.31. Amplitude walk in leading edge timing.

Even if the input amplitude is constant, time walk can be generated when the pulse shapes of the
rising part are different. This situation is usually met for germanium semiconductor detectors. In
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-19

germanium detectors, the charge carrier collection time is pretty dependent on the interaction site,
which accordingly makes the pulse shape depend on the interaction position.

Crossover timing
If a bipolar instead of a unipolar input pulse is used for timing, the output logic pulse can be
generated at the zero crossover point, which is defined as “crossover timing”. In this timing, the
crossover point is independent of the pulse amplitude, so that the amplitude walk can be
significantly reduced. DDL shaping is the simplest way to make a bipolar shape and therefore, is
preferred in crossover timing.

Constant fraction timing

The principle of constant fraction timing is illustrated in Fig. 6.32. The preamp output is
multiplied by the fraction f that is to correspond to the intended fraction of full amplitude. The
input signal is inverted and delayed for a time greater than the rise time as shown. The sum of the
two wave forms makes the “Signal for timing”. The time that this pulse crosses the zero axis is td
+ ftr, which is independent of the pulse amplitude and corresponds to the time at which the pulse
reaches the fraction f of the amplitude.

Time spectroscopy system

An example of a time spectroscopy system is shown in Fig. 6. 33. The source emits more than
two radiation quanta in cascade. The timing signal from detector 1 triggers the time-to-amplitude
converter (TAC) and then the signal from the other detector defines the stop time after a proper
delay. The amplitude of the TAC output is proportional to the difference between start and stop
times. The pulse height is analyzed with the MCA. Thus, the accumulated spectrum represents a
time spectrum. A simpler configuration can be made when a time-to-digital converter (TDC) is
employed instead of the TAC.

1 Inverted &
delayed Signal for timing
V(t)

-1
Attenuated (fVin(t))
Vin(t)
-2
Fig. 6.33. A time spectroscopy system.
-3
0 1 2 3 4
Time [s]
Fig. 6.32. Constant fraction timing principle.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-20

References
1. P.W. Nicholson, Nuclear Electronics, John Wiley & Sons, London, 1973.
2. W. Blum, W. Riegler, L. Rolandi, Particle Detection with Drift Chambers – 2nd edition (Chapter 6),
Springer, 2008.
3. G.F. Knoll, Radiation Detection and Measurement - 3rd edition (Chapters 16 to 18), John Wiley & Sons,
1999.
4. Amplifiers – Introduction, ORTEC, URL: http://www.ortec-online.com/.
5. CAMAC ADCs, Memories, and Associated Software, ORTEC, URL: http://www.ortec-online.com/.
6. V.T. Jordanov, G.F. Knoll, Nucl. Intstr. and Meth. A 345 (1994) 337.
7. J.B. Simões, C.M.B.A. Correia, Nucl. Intstr. and Meth. A 422 (1999) 405.
8. R. Grzywacz, Nucl. Intstr. and Meth. B 204 (2003) 649.
9 W.K. Warburton, M. Momayezi, B. Hubbard-Nelson, W. Skulski, Appl. Radiat. Isot. 53 (2000) 913.
10. User’s manual Digital Gamma Finder (DGF), Version 3.04, January 2004, X-Ray Instrumentation
Associates, URL: http://www.xia.com/
11. Performance of digital signal processors for gamma spectrometry, Application Note, Canberra
Industries, URL: http://www.canberra.com/.
Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-21

Appendix 1. Laplace transforms for some functions

Time function Laplace transform

 (t ) 1
1
Unit step u s (t )
s
1
t
s2

tn 1
n 1
n! s
1
e t
s 

sin t
s 2
2
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-22

Problems
1. A detector preamp signal is shaped by the following circuit. For a given preamp signal form Vin(t),
Sketch the expected signal patterns at the step A and the step B. The time constants are R1C1 = R2C2 = 2
s.

1.5

1.0

Vin(t) [V]
0.5

0.0

-0.5
0 2 4 6 8 10
Time [s]
1.5

A B 1.0
V(t) [V]

C1 R2 0.5 Step A
R1
C2 0.0

-0.5
0 2 4 6 8 10
Time [s]
1.5

1.0
V(t) [V]

0.5 Step B
0.0

-0.5
0 2 4 6 8 10
Time [s]

2. Suppose a proportional counter is connected to the

following circuit. Sketch the output signal shapes of
three detection events (horizontal axis: time t, vertical
axis: V(t) ) and briefly explain.

3. A step function input signal (voltage V0) is

connected to the following circuit. Sketch the
output signal shape and briefly explain the
reason.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-23

4. For a circuit given below,

(a) Write down the circuit equation between the input voltage Vin(t) and the output
Vout(t).

(b) For the equation obtained in (a), take the Laplace transform.

(c) Using the result obtained in (b), find Vout(t) for a step input voltage.

(d) Using the result obtained in (b), find the transfer function in the frequency domain and its magnitude.
Sketch the magnitude as a function of the angular frequency.

5. (a) For a CR-(RC)2 shaping amplifier, derive the 0% to 100% rise time of the output Vout(t) in case of a
step function input (amplitude V0).

(b) An output pulse shape of a CR-(RC)2 shaping amplifier is 6

given. Find the shaping time constant. 5

Voltage [V]
3

0
0 4 8 12 16 20 24
Time [s]

6. For the following circuit, find the relation between Vin(s) and
Vout(s).

7. The time domain output of the CR-RC shaper for a step input (amplitude V0) is given by
V
Vout (t )  0 1 (e t / 1  e t /  2 )
1   2
(a) Find the output Vout(t) when the time constants 1 and 2 become identical.

(b) Find the 0% to 100% rise time for the case (a).

(c) Sketch the shape of the Vout(t) with respect to time for a time constant of 1 s.

8. A Wilkinson peak sensing ADC is used for a radiation spectroscopy system. The dynamic range of the
ADC operation is 0 ~ 10 V.
(a) The AD conversion gain is set at 1,000 channels. For a 4 V signal, it took 2 s for the ADC to do
A-D conversion. Find the clock frequency of the ADC.

(b) The conversion gain is increased to 8,000 channels. To make this change in conversion gain, which
parameter of the ADC circuit should be changed ? Find the A-D conversion time of the same 4 V signal
for this case.
 Med Phys 4RA3, 4RB3/6R03 Radioisotopes and Radiation Methodology 6-24

9. Two pulses given below are fed to a successive approximation peak-sensing ADC. The dynamic range
is 0-10 V.
Suppose the number of the iteration cycles is set to 3. 6
1
(a) For the signal “1”, find the ADC output for each iteration cycle. 5

Briefly explain the reason. 4

2

Voltage [V]
3

0
0 10 20 30 40
Time [s]
(b) For the signal “2”, find the ADC output for each iteration cycle and briefly explain the reason.

10. The signal from a preamp is digitized using a 20 MHz 150

sampling ADC and filtered by the following algorithm:

Vout [i ]  Vav 2 [i ]  Vav1[i ] , 100

Output [arb. unit]

L 1 L 1
1 1
Vav1[i ]  
L j 0
Vin [i  j ] , Vav 2 [i ]  Vin [C  i  j ]
L j 0 50

th
Here, Vin[i] denotes i sampled data point. An output
0
pulse shape from this filter is given.

(a) Find the value “L”. -50

5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0
Time [s]

(b) Find the value “C”.

11. Suppose an input signal Vin(t) with a rise time tr is connected to a time pickoff unit.
As shown in the figure, signal shapes (1), (2), (3) denote: 3
(1): Vin(t) is attenuated by a fraction f. (0<f<1)
(2): Vin(t) is inverted and then delayed by td. 2
(3): Sum of the signals “(1)” and “(2)”.
1
(3)
Find the time that the signal “(3)” crosses the zero axis in (2)
V(t)

0
terms of f, tr and td.
(1)
-1

Vin(t)
-2

-3
Time