NI PULSE SHAPING

Nuclear Instrumentation: Lecture 2

Pulse Processing: Pulse Shaping

2.1 PULSE SHAPING

The shapes of signal pulses from detectors are usually changed or shaped by the signal
conditioning or processing elements of the data acquisition system. It is very common, for
example, to shape the output pulses of the preamplifier in the amplifier.
To assure complete charge collection from a detector, preamplifier circuits are normally
adjusted to provide a long decay time for the pulse (typical decay times are on the order of 50
µs). Since the pulses occur at random
times (radioactive decay is a random
process) they will sometimes overlap a) saturation
(especially if the count rate is large). In
such circumstances, a pulse train such as
shown in Figure 2.1(a) may occur. The
amplitudes of the pulses carry the basic
voltage

information, the charge deposited in the b)

detector (which often is proportional to the
energy of the original radiation). Hence,
the pile-up, saturation and subsequent non
time
linear response shown in Figure 2.1(a) is
Figure 2.1 (a) Typical pulse train from a preamplifier.
very undesirable. (b) Shaped pulse train.
Shaping the pulses to produce a pulse
train such as shown in Figure 2.1(b) can
alleviate the pile-up problem. With one exception, the pulses have been shaped in such a way
that their total lengths have been reduced without affecting the pulse amplitude. Such shaping is
normally carried out in a linear amplifier, usually using a variety of RC shaping networks. In this
lecture, the operation of some of the commonly used pulse shaping networks will be described.

L2-1
NI PULSE SHAPING

2.2 VOLTAGE AND CURRENT VERSUS TIME

It is useful to begin with a brief review of the behaviour of circuits, which contain resistive
and capacitive elements and their response to an AC signal (e.g. a pulse from a detector).

2.2.1 Discharging

Consider the simple circuit shown in Figure 2.2, where a charged capacitor C is connected
across a resistor R. The time evolution of the voltage is described by the differential equation,

C V0
-t/ τ
V = V0 e
voltage

R I 37%

time
τ
Figure 2.2 Discharge of a capacitor across a resistor.

dV V
C = I = − (2.1)
dt R

dV V
To solve Equation (2.1): Rewrite it as + = 0 , where τ = RC.
dt τ
+ ⎟ = (Ve t / τ ) = 0 .
⎛ dV V ⎞ d
Multiply by et/τ: e t / τ ⎜
⎝ dt τ ⎠ dt
Integrate: (Ve ) = K and, if V = V0 at t = 0, K = V0. Hence,
t /τ

the solution to Equation (2.1) is:

V = V0 e − t / RC (2.2)

where V0 is the initial voltage across the capacitor.

L2-2
NI PULSE SHAPING

Therefore, a charged capacitor placed across a resistor will discharge, as shown in Figure 2.2,
with a time constant τ = RC.
For R in ohms and C in farads, the product RC is in seconds. For example, a 1 µF capacitor
placed across a 1kΩ resistor has a time constant of 1 ms.

2.2.2 Charging

Figure 2.3 shows a slightly different situation where a resistor and capacitor are connected
across a battery (voltage source). At time t = 0, the switch is closed, connecting the battery. The
equation for the current is

dV VS − V
I =C = (2.3)
dt R

This equation can be solved in the same way as Equation (2.1) and is shown in the box below:

dV V VS
To solve Equation (2.3): Rewrite it as + = , where τ = RC.
dt τ τ
⎛ dV V ⎞ d
Multiply by et/τ: e t / τ ⎜ + ⎟= ( V
)
Ve t / τ = S e t / τ .
⎝ dt τ ⎠ dt τ
Integrate: Ve (
t /τ
) t /τ
= VS e + K and, if V = 0 at t = 0, K = -VS . Hence,

the solution to Equation (2.3) is

V = VS (1 - e-t/RC ) (2.4)

R
VS
67%
VS C V - t/ τ
voltage

V = VS ( 1 - e )

τ time

Figure 2.3 Response of an RC circuit to a sudden voltage change

L2-3
NI PULSE SHAPING

As can be seen, after a long time ( t » 5RC) the voltage approaches its final (equilibrium)
value.
5RC rule: A capacitor charges or decays to within 1% of its final value after about five time
constants.

2.3 CR DIFFERENTIATOR OR HIGH-PASS FILTER

A basic CR differentiator circuit is sketched in Figure 2.4(a). The input and output voltages
are related by the equation

Q
Vi = + V0 (2.5)
C

where Q is the charge on the capacitor at time t. Differentiating gives

dVi I dV0 (2.6)

= +
dt C dt

where the current I = dQ/dt.

Using V0 = IR and τ = RC we get

dV 0 dV
V0+ τ = τ i (2.7)
dt dt

This equation can be solved to give a general solution. However, useful insight is obtained by
considering two limiting situations where the time constant τ is very much less than or very
much greater than T, the duration of the pulse.

τ « T:
If τ (= RC) is made sufficiently small, the second term on the left hand side can be neglected
and the output voltage is proportional to the time derivative of the input, hence the name
differentiator!
To meet this requirement, the time constant needs to be short compared to the duration of
the input pulse (or to the Fourier frequency components of the pulse). However, care needs to
be taken not to `load' the input by making RC too small (at the transition, the change in voltage
across the capacitor is zero, so R is the load seen by the input).

L2-4
NI PULSE SHAPING

τ » T:
At the opposite extreme of a long time constant, the first term can be neglected giving

dV 0 dV
τ ≈τ i (2.8)
dt dt

Integrating this and setting the constant of integration equal to zero gives

V0 ≅Vi (2.9)

In this limit, the network will pass the waveform without distortion.
The CR differentiator is a high-pass filter.
ƒ High frequency components of pulses (edges) are passed without distortion.
ƒ Low frequencies are attenuated away and any dc component is not passed.
The effect of a high-pass CR filter on different input waveforms is shown in Figure 2.4.

Input Output

C (b) Vi=V V0 = V exp(-t/RC )

Input Output
Vi R V0
Equal
(c) areas
(a)
T

(d) RC >> T

(e) RC << T

(f) DC level
shift

Figure 2.4 High-pass CR filter (differentiator): (a) basic circuit; (b) step input; (c) single (square) pulse (RC = T); (d)
single pulse (RC » T); (e) single pulse (RC « T); (f) pulse train.

L2-5
NI PULSE SHAPING

ƒ For a step input, the output is

-t/τ
V0=Vie , (2.10)

which approximately represents the shaping of a rapidly rising signal pulse by a single
CR differentiator. Note that the fast leading edge is not differentiated because τ is not
small compared to its rise time. Therefore, the leading edge is simply passed through
while the shaping consists of differentiating away the long tail.
ƒ For a single square pulse [curve (c)], the fast rise and fall are passed undistorted and the
DC signal falls towards zero. Note that the areas above and below the baseline are equal,
which is a result of the fact that any dc Fourier component of the input is not passed.
ƒ Curve (d) shows that if the time constant is long compared with T, the pulse is passed
with little distortion.
ƒ By contrast, if RC « T [curve (e)], the pulse is strongly differentiated and the output is
large only when dVi/dt is large.
ƒ A train of pulses is passed [curve (f)], shaped according to the value of RC relative to T,
but there is no baseline shift, i.e. the dc component of the input signal is not passed.

The response of a high-pass filter to a rising input voltage Vi = kt, is shown in Figure 2.5.

Vi = kt
Signals

V0
A
∆
B kRC
t
Time

Figure 2.5 Response of a high-pass CR filter to a rising ramp input.

L2-6
NI PULSE SHAPING

ƒ At large t (» RC), the output reaches a saturation value equal to τ(dVi /dt).
ƒ Initially, (for t « RC), the output (solid curve) follows the input.
ƒ Any real pulse rises from zero to its peak in a finite time Tr and there will be a difference
between the output and input depending on the ratio of τ to Tr. This difference is called
the ballistic deficit (shown as ∆ in Figure 2.5). If τ > 5×Tr , ∆ < 1% of the signal height.

Differentiators are useful for detecting leading and trailing edges in pulses. For example, in
digital circuitry one sometimes sees networks like that shown in Figure 2.6. In this case, the
differentiator generates spikes at the transitions of the input pulse and the output buffer converts
these into short square pulses.

2.4 THE RC INTEGRATOR OR LOW-PASS FILTER

A passive RC network acts as an integrator when configured as shown in Figure 2.7. The
circuit equation is now

b
Vi V0
Input C Output
R

Vi Input

b - RC

V0 Output

Figure 2.6 Example of the use of a differentiator in digital circuitry.

V i = IR + V 0 . (2.11)

Since I (= dV/dt) represents the rate of charging of the capacitor. This can be rewritten as

L2-7
NI PULSE SHAPING

dV 0
Vi= τ +V 0
dt
which can be rearranged as

dV 0 V 0 V i . (2.12)
+ =
dt τ τ

Again, consider the extreme situations when RC is very large or very small (compared with
the pulse duration, for example).
τ » T:
When τ (= RC) is large, only the first term on the left-hand side matters and we see that the
output voltage is proportional to the integral of the input voltage, hence the name `integrator'.

1
V 0 = ∫ V i dt (2.13)
τ

τ « T:
At the opposite extreme of very small time constants (low frequencies), only the second term
is significant and once again

V0 ≅Vi (2.14)

so that the network will pass the signal without change.

The effect of a low-pass RC filter on different input waveforms is shown in Figure 2.7.

Input Output
R (b) Vi=V
Input Output V0 = V [1- exp(-t/RC ) ]
Vi C V0

(c)
(a)
T

(d) RC >> T

(e) RC << T

(f)

Figure 2.7 Low-pass RC filter (integrator): (a) basic circuit; (b) step input; (c) single (square) pulse (RC = T); (d)
single pulse (RC » T); (e) single pulse (RC « T); (f) pulse train.

L2-8
NI PULSE SHAPING

For a step input [curve (b)], the fast rise is not passed and the output begins to increase
only after the input has reached its maximum dc value. The capacitor C charges up at a rate
dependent upon τ and, eventually, the output becomes equal to the input.
ƒ For a single, square pulse [curve (c)], the output rises as C charges and then it
falls after the end of the pulse.
ƒ If RC » T, the output reaches a low value [curve (d)], because it varies inversely
as 1/τ, according to Equation (2.13).
ƒ By contrast, if RC « T [curve (e)], the pulse is passed with little distortion,
according to Equation (2.14).
ƒ A train of pulses is passed [curve (f)], shaped according to the value of RC
relative to T. In this case, the baseline shift is passed.

The response of a low-pass filter to a voltage that is rising for a time T, is shown in Figure 2.8.

RC << T
Vi = kt
Signals

V0

RC >> T

t
Time T

Figure 2.8 Low-pass RC filter response to a linear rising ramp input.

ƒ If t « RC, after a short delay, the output follows the input, according to Equation
(2.14)

ƒ If t » RC, the output is proportional to the integral of the input, according to

Equation (2.13).

The integrator is used extensively in analog computation. It is a useful circuit that finds
applications in control systems, analog/digital conversion, and in waveform generation.

L2-9
NI PULSE SHAPING

2.5 CR-RC SHAPING

The combination of a CR differentiator and RC integrator is commonly used as a pulse shaper

in linear amplifiers.
The output of a single CR differentiating circuit has several unwelcome features for pulse
processing systems.
ƒ The sharp top makes any subsequent pulse height analysis difficult (the maximum
is only held for a short time).
ƒ All high-frequency noise components are passed through by the circuit, implying
that the signal-to-noise ratio is usually poor.

If an RC integrating stage follows the CR stage both of these features can be much improved.
Such a combination, shown in Figure 2.9 is one of the most widely used methods for shaping
preamplifier pulses.

C1 R2

C2 Vo
Vi R1

Figure 2.9 A CR-RC shaping network.

The triangular symbol is an operational amplifier (op amp), which in this configuration has
infinite input impedance and zero output impedance. The op amp serves to isolate the two
individual networks (impedance isolation) so that neither influences the operation of the other.
The response of this CR-RC network to a step voltage of amplitude Vi is

V i τ1 -t/ τ1 -t/ τ 2
V0= (e - e ) (2.15)
τ1 - τ2

where τ1 and τ2 are the time constants of the differentiating and integrating circuits,
respectively. For interested students, the derivation of this equation is given in the box below.
Figure 2.10 plots the response of Equation (2.12) for several different combinations of time

L2-10
NI PULSE SHAPING

constants.

Derivation of Equation (2.15):

The first part of the circuit is a high-pass filter. The output voltage is given by Equation
(2.10): V S = V i e-t/τ1 , where τ1 = R1C1.
The second part is a low-pass filter, whose behavior is given by Equation (2.12) with Vi =
dV 0 V 0 V S
VS, i.e. + = , where τ2 = R2 C2.
dt τ2 τ2
dV V d V exp(t/τ 2 )
Multiply by exp(t/τ2): exp(t/τ 2 ) 0 + 0 exp(t/τ 2 ) = (V0 exp(t/τ 2 )) = S .
dt τ2 dt τ2

Substitute for VS:

d
(V0 exp(t/τ 2 )) = V i exp[t(1/τ 2 − 1/τ 1 )] .
dt τ2
V τ exp[t(1/τ 2 − 1/τ 1 )]
Integrate: V0 exp(t /τ2 ) = i 1 +K.
(τ 1 - τ 2 )
Setting the condition: V0 = 0 when t = 0, K = −Viτ1/(τ1-τ2) from which we get Equation
(2.15).

Figure 2.10 Response of a CR-RC network to a step voltage. Curves are shown for several different combinations
of time constants. Curves are labelled τ1+τ2.

L2-11
NI PULSE SHAPING

In some applications (a common situation in nuclear pulse processing), CR-RC shaping is

carried out using equal time constants (τ say). For this particular case, the solution is
t -t/τ
V0=Viτ e . (2.16)

This response is also illustrated in Figure 2.10.

Derivation of Equation (2.16):

V i τ1 -t / τ 2 -t (1/ τ1−1 / τ 2 )
If τ1 = τ2 = τ, rewrite Equation (2.15) as V 0 = e (e - 1) and expand the
τ1 - τ2
first term in brackets to first order in t:
V i τ1 -t / ⎡ ⎛ 1 1 ⎞ ⎤ V i τ e-t/τ ⎡ t ( τ1 - τ2) ⎤
V 0 = - e τ 2 ⎢1 − t ⎜⎜ τ − ⎟⎟ − 1⎥ = - , giving Equation (2.16).
τ1 τ2 ⎣ ⎝ 1 τ2 ⎠ ⎦ τ1 τ2 ⎢⎣ τ 2 ⎥⎦

The choice of shaping time constants is determined primarily by the charge collection time in
the detector. As always, there are competing factors that need to be considered:

ƒ In order to minimize pile-up effects, the time constants should be kept short so
that the pulse returns to baseline as quickly as possible.

ƒ On the other hand, when the time constants become comparable to the rise time of
the pulses from the preamplifier, the input to the network no longer looks like a step
voltage and the result is that some of the amplitude of the signal is lost.
This is referred to as ballistic deficit (see Section 2.3) and can be avoided only by
keeping the time constants long compared to the charge collection time of the detector.

Typical values for τ range from a few tenths of a µs for small semiconductor detectors through
to a few µs for `large' Ge detectors to tens of µs for some types of proportional counter.

2.6 GAUSSIAN SHAPING

When a single CR differentiator is followed by several stages of RC integration (four are

sufficient), the output pulse shape approximates that of a mathematical Gaussian.
In theory, a Gaussian pulse shape has some potential advantages for improvement in signal-
to-noise ratio over simple CR-RC shaping. Gaussian or semi-Gaussian shaping is very
commonly found in linear amplifiers used in nuclear pulse processing applications. A

L2-12
NI PULSE SHAPING

disadvantage, however, is that because of the greater pulse width, pile-up effects are more severe
at high rates.

2.7 DOUBLE DIFFERENTIATION OR CR-RC-CR SHAPING

Double differentiation can be used to generate a bipolar pulse shape. Another differentiation
stage is added with the result illustrated in Figure 2.11.
ƒ The most common choice is to make all three time constants equal.
ƒ The bipolar pulse shape makes baseline shifts (see later) much less severe.
ƒ This type of signal is most useful at high count rates. However, at lower rates, its
signal-to-noise characteristics are usually not as good as single-stage RC shaping because
the amplitude is lower.

C1 R2 C3

C2 Vo
Vi R1 R3

first integration second

differentiation differentiation

Vo
Vi

0.25

1 3 4 5 6
2
time
τ

Figure 2.11 Creation of a bipolar pulse using a CR-RC-CR shaping network.

L2-13
NI PULSE SHAPING

2.8 SINGLE AND DOUBLE DELAY-LINE PULSE SHAPING

Some properties of coaxial cables can be used to shape pulses. Recall that a cable shorted at
one end will produce a reflected pulse moving back toward the sending end of the cable with an
equal amplitude but opposite polarity.

2.8.1 Single delay-line shaping

A single delay line can be used to generate a

square pulse from a step voltage input as shown in Zo

Figure 2.12
transmission line
The transmission line is assumed to be long shorted at bottom end

enough so that the propagation time is long

compare to the rise time of the step. The op amp input step

provides impedance isolation and, since its output

impedance is zero, the resistor Z0 terminates the
cable at its sending end. The width of the shaped 2τ
reflected step

pulse is just twice the propagation time through the

output pulse
length of the cable. For many applications, this is a
microsecond or so. Special delay lines, with Figure 2.12 Single delay-line shaping.
reduced pulse propagation velocities, are often
used to avoid excessively long cables.
Figure 2.13 illustrates problems, which can arise if the delay time T is (a) too short or (b) too
long.
At one extreme, T must not be shorter than the pulse rise time, otherwise ballistic-deficit
effects will appear [see Figure 2.13(a)]. On the other hand, T must not be so long that it is
comparable with the decay time of the input pulse, or there will be an undershoot, as illustrated
in Figure 2.13(b). However, if it is not possible to achieve this condition, the undershoot can be
alleviated by attenuating the reflected pulse, as illustrated in the lower two waveforms in Figure
2.13(b).

L2-14
NI PULSE SHAPING

DELAY TIME

(a) too short (b) too long

Input Input

Reflected pulse Reflected pulse

Output pulse
Undesirable undershoot

Reflected and attenuated

Modified output pulse

Figure 2.13 Single delay-line shaping where the delay time is (a) less than the pulse rise time and (b) is comparable
with the input pulse decay time.

2.8.2 Double delay-line shaping

Bipolar pulses can be produced if two delay

Zo
lines are used, as shown in Figure 2.14. If both
delay lines have equal length, the resulting pulse transmission lines
shorted at bottom ends
will have equal positive and negative lobes and
an average dc level of zero can be maintained. input

This will virtually eliminate any baseline shift in

output of first stage
subsequent circuits. While DDL shaping has
excellent high counting rate capabilities, it does
reflected pulse from

not apply any high-frequency filtering to the second delay line

signal and therefore, has poorer signal-to-noise

characteristics than RC shaping networks. bipolear output from
double delay line
Therefore, it is seldom used in high resolution
Figure 2.14 Double delay-line shaping and the
systems (i.e. Ge detectors). production of bipolar pulses.

L2-15
NI PULSE SHAPING

2.9 DELAY-LINE PLUS RC SHAPING

Yet another option is to send the output of a delay-line network to an integrator or low-pass
filter. This eliminates much of the high-frequency noise from the signal with a corresponding
improvement in the response of the network.

ƒ If the output of a DDL network is integrated, a triangular pulse can be produced

as shown in Figure 2.15.
ƒ As the time constants are increased, the approximation to a triangular shape gets
better but the amplitude of the pulse is reduced.
ƒ The signal-to-noise characteristics of triangular pulses are quite good and such
pulses are sometimes used in high-resolution systems when the count rate is not
excessively high.

τ = T/ 10
E out
(a)

τ = 4T Time

τ=T

τ = T/ 10
E out
(b)

τ=T
τ = 4T

0 2T 4T Time
Figure 2.15 (a) Production of triangular shaped pulses using a DDL network through an integrating RC network.
(b) Similarly for an SDL network.

L2-16
NI PULSE SHAPING

2.10 POLE-ZERO AND BASELINE RESTORATION.

2.10.1 Pole-zero

The simple CR-RC circuit described above produces a significant undershoot as the amplifier
pulse attempts to return to zero. This is due to the long exponential decay of the preamplifier
pulse. In high count rate situations it is possible (likely) that another pulse will arrive during this
time and `ride' on the undershoot of the original pulse. Obviously, in this case, the apparent
amplitude of the second pulse will be somewhat reduced, resulting in an undesirable broadening
of the peaks in the energy spectrum.
The problem can be alleviated by use of a pole-zero cancellation network, such as shown in
Figure 2.16. In Figure 2.16(a) the preamplifier signal is applied to the input of a normal CR
differentiator circuit. The output shows the typical undershoot.

Preamp Output
Input to differentiator

R pz
Input to integrator

a)
Uncorrected output

b)
Corrected Output

Figure 2.16 The benefit of pole-zero correction.

The following equation holds:

undershoot amplitude differentiation time
=
pulse amplitude preamp pulse decay time

L2-17
NI PULSE SHAPING

Therefore, for a longer preamp decay time, a longer shaping time in the amplifier leads to
larger undershoots.
In Figure 2.16(b), the resistor Rpz placed in parallel with the capacitor can be adjusted to
cancel the undershoot. The result is an output pulse with a simple exponential decay to zero.
Virtually all spectroscopy amplifiers incorporate this pole-zero circuit (the term pole-zero comes
from the mathematical representation of the circuit, the resistor `cancels' a pole in the
expression).

2.10.2 Baseline restoration

To ensure good energy resolution and peak position stability, high-performance spectroscopy
amplifiers are entirely dc coupled (except for the CR differentiator network located close to the
input which is ac coupled almost by definition). Therefore, any DC offsets in the early stage of
the system will be greatly amplified to cause a large unstable dc offset at the output. Since the
amplitude of the pulses in nuclear applications carries much of the desired information such an
offset is extremely undesirable.
The basic principle of the baseline restorer is illustrated in Figure 2.17. We consider two
modes of operation:

Simple baseline restoration.

For a simple time-invariant baseline restorer, the switch S1 is always closed and the restorer
behaves like a CR differentiator. The baseline between pulses is restored to zero by the resistor
Rblr. The time constant CblrRblr must be at least 50 times the shaping constant in order not to harm
the signal-to-noise ratio.
Such a time invariant system does not adequately maintain the baseline at high counting rates.
Since this simple circuit is essentially a CR differentiator, the average signal area above ground
potential must equal that below (a capacitor cannot pass a dc level). For low count rates, the
spacing between pulses is very long and so the baseline is restored to very close to ground
potential. However, as the rate increases, the baseline must shift down, the amount of shift
increasing with count rate.

L2-18
NI PULSE SHAPING

C blr
Vi V0 Baseline
Input restored
output
Low output High input
impedance R blr impedance
S1

Low rate High rate

Vi Input

V0 Simple
restorer

V0 Gated
restorer
Switch S1 O C O C O C

Figure 2.17 Operation of a base-line restorer in simple mode and in gated mode.

Gated baseline restoration:

The gated baseline restorer virtually eliminates the problem caused by changing count rates.
In this case, the switch S1 in Figure 2.17 is opened (O) only during the duration of each amplifier
pulse and is closed (C) otherwise. Therefore, the CR differentiator is active only on the baseline
in between pulses and the effect of signal pulse is virtually eliminated. The gated baseline
restorer `thinks' that it is operating at zero counting rate!

2.11 GATED INTEGRATOR

Performance at high count rate can be improved using a gated integrator at the output stage of
the amplifier shaping network. The principle is illustrated in Figure 2.18.
The unipolar output of a shaping pre-filter is integrated on a capacitance C, which is part of an
active circuit, for a time that encompasses the duration of the pulse. At the end of this interval,
the capacitance is discharged by closing a switch. The amplitude of the pulse from the gated
integrator is now proportional to the area of the unipolar pulse from the shaping circuit rather
than its peak value. This is much less sensitive to the ballistic deficit caused by variable charge

L2-19
NI PULSE SHAPING

collection times in the detector.

As an illustrative example, consider the two delay-line shaped pulses shown in Figure 2.18.
The top waveform shows a pulse with a fast rise time. The area of the pulse is VT, where T is the
delay time and V is the pulse amplitude. The central waveform shows a pulse with a ballistic
deficit, because it has a rise time Tr that exceeds T. The pulse reaches an amplitude mT, where m
is the voltage gradient of the leading edge, and there is a ballistic deficit because Tr > T. The
pulse falls to zero at time (V/m)+T after its initial rise. The area of the pulse is the area of the
trapezoid ABCD, which is VT, i.e. it is independent of Tr. The output of the gated integrator is
shown in the bottom waveform. Before the start of the pulse, switch S1 is open and S2 is closed.
During integration, from t1 to t2, S1 is closed and S2 is open and at the end of the pulse, S1 is
open and S2 is closed, awaiting the arrival of the next pulse. Provided that (t2 - t1) exceeds the
total pulse duration, the output of the gated integrator is independent of ballistic-deficit effects of
using a shaping time comparable with Tr .

Vi V

Vi V0
t
S1
T
Shaping C Vi
pre filter V
mt D C

A B
S2 t
Gated integrator V/m T
V0

t
S1 open S1 closed S1 open
S2 closed S2 open S2 closed
t1 t2

Figure 2.18 Principles of operation of a gated integrator.

Figure 2.19 shows the timing and output pulse from a gated integrator acting on a unipolar
pulse from the output of a Gaussian shaping network The integration continues for a time that is
about 8 to 10 times the shaping time used in the Gaussian network (or 2 to 3 times the peaking
time). With a gated integrator, much shorter shaping times can be used than if a gated integrator
is not used, which will minimize the effect of pile up on the resolution. The waveform shown in

L2-20
NI PULSE SHAPING

Figure 2.19 uses a Gaussian shaping time of 0.25 µs resulting in a total time of 5 µs above the
baseline for the output of the gated integrator. Using a conventional Gaussian shaping amplifier,
would need a much longer shaping time of 3 µs (20 µs above the baseline) to achieve the same
resolution. Thus, use of a gated integrator enables higher-rate operation before the effects of pile
up become a limiting factor. Systems using gated integrators can operate with germanium
detectors up to a count rate of 106 pulses/sec with good energy resolution.

TW
TW − TP
TP

UNIPOLAR

GI

TW

Figure 2.19 Output pulses from (upper) semi-Gaussian and (lower) gated-
integrator amplifier.

TP = time to peak; TW = total pulse width.

TP is, typically, 2×amplifier shaping time.

L2-21