Ministry of Higher Education

University College of Wisdom

Computer technology engineering

Transfer Functions

Student report

Mahmoud Mohamed Khodeir

The supervisor DR. Najam Abdullah

2019 \ 2020
Transfer Functions

The typical regulator system can frequently be described, in essentials, by

differential equations of no more than perhaps the second, third or fourth
order. . . . In contrast, the order of the set of differential equations describing
the typical negative feedback amplifier used in telephony is likely to be very
much greater. As a matter of idle curiosity, I once counted to find out what
the order of the set of equations in an amplifier I had just designed would
have been, if I had worked with the differential equations directly. It turned
out to be 55.
Henrik Bode, 1960 [Bod60].

This chapter introduces the concept of the transfer function, which is a

compact description of the input/output relation for a linear system. Com-
bining transfer functions with block diagrams gives a powerful method for
dealing with complex linear systems. The relationship between transfer
functions and other system descriptions of dynamics is also discussed.

8.1 Frequency Domain Analysis

Figure 8.1 shows a block diagram for a typical control system, consisting of a
process to be controlled and a (dynamic) compensator, connected in a feed-
back loop. We saw in the previous two chapters how to analyze and design
such systems using state space descriptions of the blocks. As was mentioned
in Chapter 2, an alternative approach is to focus on the input/output char-
acteristics of the system. Since it is the inputs and outputs that are used to
connect the systems, one could expect that this point of view would allow
an understanding of the overall behavior of the system. Transfer functions
are the main tool in implementing this point of view for linear systems.
 CHAPTER 8. TRANSFER FUNCTIONS

reference feedback d process n

shaping controller dynamics
e u y
r F Σ C Σ P Σ

−1

Figure 8.1: A block diagram for a feedback control system.

The basic idea of the transfer function comes from looking at the fre-
quency response of a system. Suppose that we have an input signal that is
periodic. Then we can always decompose this signal into the sum of a set
of sines and cosines,
∞
X
u(t) = ak sin(kωt) + bk cos(kωt),
k=1

where ω is the fundamental frequency of the periodic input. Each of the

terms in this input generates a corresponding sinusoidal output (in steady
state), with possibly shifted magnitude and phase. The magnitude gain and
phase at each frequency is determined by the frequency response, given in
equation (5.21):
G(s) = C(sI − A)−1 B + D, (8.1)
where we set s = j(kω) for each k = 1, . . . , ∞. If we know the steady
state frequency response G(s), we can thus compute the response to any
(periodic) signal using superposition.
The transfer function generalizes this notion to allow a broader class of
input signals besides periodic ones. As we shall see in the next section, the
transfer function represents the response of the system to an “exponential
input,” u = est . It turns out that the form of the transfer function is
precisely the same as equation (8.1). This should not be surprising since we
derived equation (8.1) by writing sinusoids as sums of complex exponentials.
Formally, the transfer function corresponds to the Laplace transform of the
steady state response of a system, although one does not have to understand
the details of Laplace transforms in order to make use of transfer functions.
The power of transfer functions is that they allow a particularly conve-
nient form for manipulating and analyzing complex feedback systems. As we
shall see, there are many graphical representations of transfer functions that
DERIVATION OF THE TRANSFER FUNCTION

capture interesting properties of dynamics. Transfer functions also make it

possible to express the changes in a system because of modeling error, which
is essential when discussing sensitivity to process variations of the sort dis-
cussed in Chapter 12. In particular, using transfer functions it is possible
to analyze what happens when dynamic models are approximated by static
models or when high order models are approximated by low order models.
One consequence is that we can introduce concepts that express the degree
of stability of a system.
The main limitation of transfer functions is that they can only be used
for linear systems. While many of the concepts for state space modeling
and analysis extend to nonlinear systems, there is no such analog for trans-
fer functions and there are only limited extensions of many of the ideas
to nonlinear systems. Hence for the remainder of the text we shall limit
ourselves to linear models. However, it should be pointed out that despite
this limitation, transfer functions still remain a valuable tool for designing
controllers for nonlinear systems, chiefly through constructing their linear
approximations around an equilibrium point of interest.

Derivation of the Transfer Function

As we have seen in previous chapters, the input/output dynamics of a linear
system has two components: the initial condition response and the forced
response. In addition, we can speak of the transient properties of the system
and its steady state response to an input. The transfer function focuses on
the steady state response due to a given input, and provides a mapping
between inputs and their corresponding outputs. In this section, we will
derive the transfer function in terms of the “exponential response” of a
linear system.

Transmission of Exponential Signals

To formally compute the transfer function of a system, we will make use of
a special type of signal, called an exponential signal, of the form est where
s = σ + jω is a complex number. Exponential signals play an important role
in linear systems. They appear in the solution of differential equations and in
the impulse response of linear systems, and many signals can be represented
as exponentials or sums of exponentials. For example, a constant signal is
simply eαt with α = 0. Damped sine and cosine signals can be represented
by
e(σ+jω)t = eσt ejωt = eσt (cos ωt + i sin ωt),
 CHAPTER 8. TRANSFER FUNCTIONS

s=0 s=−1 s=1

3
1 1
2

y
0.5 0.5
1

0 0 0
0 0.5 1 0 2 4 0 0.5 1
x x x
z=0 z=0.2 z=−0.2
1 20
1
0.5 0.5 10

0 0 0
y

y
−0.5 −0.5 −10
−1
−1 −20
0 5 10 15 0 5 10 15 0 5 10 15
x x x

Figure 8.2: Examples of exponential signals.

where σ < 0 determines the decay rate. Many other signals can be rep-
resented by linear combinations of exponentials. Figure 8.2 give examples
of signals that can be represented by complex exponentials. As in the case
of sinusoidal signals, we will allow complex valued signals in the derivation
that follows, although in practice we always add together combinations of
signals that result in real-valued functions.
To investigate how a linear system responds to the exponential input
u(t) = est we consider the state space system

ẋ = Ax + Bu
(8.2)
y = Cx + Du.

Let the input signal be u(t) = est and assume that s 6= λi (A), i = 1, . . . , n,
where λi (A) is the ith eigenvalue of A. The state is then given by
Z t Z t
x(t) = eAt x(0) + eA(t−τ ) Besτ dτ = eAt x(0) + eAt e(sI−A)τ B dτ.
0 0

If s 6= λ(A) the integral can be evaluated and we get

t
At At −1 (sI−A)τ 
x(t) = e x(0) + e (sI − A) e  B
τ =0
 
At At −1 (sI−A)t
= e x(0) + e (sI − A) e −I B
 
= eAt x(0) − (sI − A)−1 B + (sI − A)−1 Best .
8.2. DERIVATION OF THE TRANSFER FUNCTION

The output of equation (8.2) is thus

y(t) = Cx(t) + Du(t)

   
= CeAt x(0) − (sI − A)−1 B + C(sI − A)−1 B + D est , (8.3)

a linear combination of the exponential functions est and eAt . The first term
in equation (8.3) is the transient response of the system. Recall that eAT
can be written in terms of the eigenvalues of A (using the Jordan form) and
hence the transient response is a linear combinations of terms of the form
eλi t , where λi are eigenvalues of A. If the system is stable then eAT → 0 as
t → ∞ and this term dies away.
The second term of the output (8.3) is proportional to the input u(t) =
est . This term is called the pure exponential response. If the initial state is
chosen as
x(0) = (sI − A)−1 B,
then the output only consists of the pure exponential response and both the
state and the output are proportional to the input:

x(t) = (sI − A)−1 Best = (sI − A)−1 Bu(t)

y(t) = C(sI − A)−1 B + D est = C(sI − A)−1 B + D u(t).

The map from the input to output,

Gyu (s) = C(sI − A)−1 B + D, (8.4)

is the transfer function of the system (8.2); the function

Gxu (s) = (sI − A)−1 B

is the transfer function from input to state. Note that this latter transfer
function is actually a vector of n transfer functions (one for each state).
Using transfer functions the response of the system (8.2) to an exponential
input is thus
 
y(t) = CeAt x(0) − (sI − A)−1 B + Gyu (s)est . (8.5)

An important point in the derivation of the transfer function is the fact

that we have restricted s so that s 6= λi (A), i = 1, . . . , n, where λi (A).
At those values of s, we see that the response of the system is singular
(since sI − A will fail to be invertible). These correspond to “modes” of
 CHAPTER 8. TRANSFER FUNCTIONS

the system and are particularly problematic when Re s ≥ 0, since this can
result in bounded inputs creating unbounded outputs. This situation can
only happen when the system has eigenvalues with either positive or zero
real part, and hence it relates to the stability of the system. In particular,
if a linear system is asymptotically stable, then bounded inputs will always
produce bounded outputs.

Coordinate Changes
The matrices A, B and C in equation (8.2) depend on the choice of coor-
dinate system for the states. Since the transfer function relates input to
outputs, it should be invariant to coordinate changes in the state space. To
show this, consider the model (8.2) and introduce new coordinates z by the
transformation z = T x, where T is a nonsingular matrix. The system is
then described by

ż = T (Ax + Bu) = T AT −1 z + T Bu = Ãz + B̃u

y = Cx + DU = CT −1 z + Du = C̃z + Du

This system has the same form as equation (8.2) but the matrices A, B and
C are different:

à = T AT −1 B̃ = T B C̃ = CT −1 D̃ = D. (8.6)

Computing the transfer function of the transformed model we get

G̃(s) = C̃(sI − Ã)−1 B̃ + D

= CT −1 T (sI − A)−1 T −1 T B + D
= CT −1 (sI − T AT −1 )−1 T B + D

−1
= C T −1 (sI − T AT −1 )T B+D
= C(sI − A)−1 B + D = G(s),

which is identical to the transfer function (8.4) computed from the system
description (8.2). The transfer function is thus invariant to changes of the
coordinates in the state space.
 Another property of the transfer function is that it corresponds to the por-
tion of the state space dynamics that are both reachable and observable. In
particular, if we make use of the Kalman decomposition (Section 7.5), then
the transfer function only depends on the dynamics on the reachable and
observable subspace, Sro (Exercise 2).
8.2. DERIVATION OF THE TRANSFER FUNCTION

Transfer Functions for Linear Differential Equations

Consider a linear input/output system described by the differential equation

dn y dn−1 y dm u dm−1 u
+ a1 + · · · + an y = b 0 + b 1 + · · · + bm u, (8.7)
dtn dtn−1 dtm dtm−1

where u is the input and y is the output. This type of description arises in
many applications, as described briefly in Section 2.2. Note that here we
have generalized our previous system description to allow both the input
and its derivatives to appear.
To determine the transfer function of the system (8.7), let the input be
u(t) = est . Since the system is linear, there is an output of the system
that is also an exponential function y(t) = y0 est . Inserting the signals in
equation (8.7) we find

(sn + a1 sn−1 + · · · + an )y0 est = (b0 sm + b1 sm−1 · · · + bm )e−st

and the response of the system can be completely described by two polyno-
mials
a(s) = sn + a1 sn−1 + · · · + an−1 s + an
(8.8)
b(s) = b0 sm + b1 sm−1 + · · · + bm−1 s + bm .

The polynomial a(s) is the characteristic polynomial of the ordinary

differential equation. If a(s) 6= 0 it follows that

b(s) st
y(t) = y0 est = e = G(s)u(t). (8.9)
a(s)

The transfer function of the system (8.7) is thus the rational function

b(s)
G(s) = , (8.10)
a(s)

where the polynomials a(s) and b(s) are given by equation (8.8). Notice that
the transfer function for the system (8.7) can be obtained by inspection,
since the coefficients of a(s) and b(s) are precisely the coefficients of the
derivatives of u and y.
Equations (8.7)–(8.10) can be used to compute the transfer functions of
many simple ODEs. The following table gives some of the more common
forms:
 CHAPTER 8. TRANSFER FUNCTIONS

Type ODE Transfer Function

1
Integrator ẏ = u
s
Differentiator y = u̇ s
1
First order system ẏ + ay = u
s+a
1
Double Integrator ÿ = u
s2
1
Damped oscillator ÿ + 2ζωn ẏ + ωn2 = u
s2 + 2ζωn s + ωn2
R ki
PID controller y = kp u + kd u̇ + ki u kp + kd s +
s
Time delay y(t) = u(t − τ ) e−τ s

The first five of these follow directly from the analysis above. For the PID
controller, we let the input be u(t) = est and search for a solution y(t) = est .
It follows that
ki
y(t) = kp est + kd sest + est ,
s
giving the indicated transfer function.
Time delays appear in many systems: typical examples are delays in
nerve propagation, communication and mass transport. A system with a
time delay has the input/output relation

y(t) = u(t − T ). (8.11)

As before the input be u(t) = est . Assuming that there is an output of the
form y(t) = y0 est and inserting into equation (8.11) we get

y(t) = y0 est = es(t−T ) = e−sT est = e−sT u(t).

The transfer function of a time delay is thus G(s) = e−sT which is not a
rational function, but is analytic except at infinity.

Example 8.1 (Operational amplifiers). To further illustrate the use of ex-

ponential signals, we consider the operational amplifier circuit introduced
in Section 3.3 and reproduced in Figure 8.3. The model introduced in Sec-
tion 3.3 is a simplification because the linear behavior of the amplifier was
modeled as a constant gain. In reality there is significant dynamics in the