ANALOG CONTROL SYSTEMS - A REVIEW

ABSTRACT
The aim of this chapter is to review briefly the main concepts of analog control systems. The
presentation us such that it will permit an easy transition to digital control systems at a later
stage.
The subject matter handled relates to the description of continuous-time models in the time and
frequency domains, the properties of closed-loop systems, and the synthesis of PI and PID
controllers.

1.1 CONTINUOUS-TIME MODELS

The behavior of the systems to be controlled around an operation point is described in most
cases by a linear dynamic model. The controllers also are generally characterized by a linear
dynamic model.
The linear dynamic models of the systems to be controlled or of the controllers are described
in the time domain by linear differential equations, and in the frequency domain, by transfer
functions.
1.1.1 Time Domain
Equation (1.1.1) gives n example of a differential equation describing a simple dynamic
system:

𝑑𝑦 1 𝐺
= − 𝑦(𝑡) + 𝑢(𝑡)
𝑑𝑡 𝑇 𝑇
(1.1.1)

In Equation (1.1.1) u represents the input (or the control) of the system and y
represents the output. This equation may be simulated by analog means, as illustrated in
Figure 1.1.1.
The step response illustrated in Figure 1.1.1 reveals the speed of the output variations,
characterized by the time constant T, and the final value, characterized by the static gain G.
Using the differential operator p = d/dt, Equation (1.1.1) is written as :
1 𝐺 𝑑
(𝑝 + ) 𝑦(𝑡) = 𝑢(𝑡); 𝑝=
𝑇 𝑇 𝑑𝑡
(1.1.2)

1.1.2 The frequency Domain

The characteristics of the models in the form of Equation (1.1.1) can be studied either in the
time domain or in the frequency domain. In the latter, the aim is to study the behavior of the
system when the input u is a sinusoidal or a consinusoidal input that spans a certain spectrum
of frequencies/
Remember that :
𝑒 𝑗𝜔𝑡 = cos 𝜔𝑡 + 𝑗 sin 𝜔𝑡
(1.1.3)
And consequently it can be considered that the study of the dynamic system described by an
equation such as Equation (1.1.1) in the frequency domain corresponds to the study of the
system output for inputs of the type u(t) = 𝑒 𝑗𝜔𝑡 .
Since the system is linear, the output will be a signal containing only the frequency 𝜔,
the input being amplified or attenuated (with the possible appearance of a phase deviation)
according to 𝜔; 1. e., the output will be of the form:
𝑦(𝑡) = 𝐻(𝑗𝜔)𝑒 𝑗𝜔𝑡
(1.1.4)
Figure 1.12 illustrates the behavior of a system for an input u(t) = 𝑒 𝑗𝜔𝑡 .
However, there is nothing to stop us from assuming that the input is formed by damped
or undamped sinusoids and cosinusoids, which in this case are written as :
𝑢(𝑡) = 𝑒 𝜎𝑡 𝑒 𝑗𝜔𝑡 = 𝑒 𝑠𝑡 ; 𝑠 = 𝜎 + 𝑗𝜔
(1.1.5)
s is interpreted as a complex frequency. As a result of the linearity of the system, the
output will reproduce the input signal, amplified (or attenuated), with or without a phase
deviation, depending on the values of s; i.e., the output will be of the form.
𝑦(𝑡) = 𝐻(𝑠)𝑒 𝑠𝑡
(1.1.6)
LIST OF PRINCIPAL NOTATIONS

fs Sampling frequency
Ts Sampling period
t Continuous-time or normalized discrete-time (with respect to
the sampling period)
k Normalized discrete-time (t/Ts)
u(t), y(t) Plant input and output
e(t) Discrete-time Gaussian white noise
q-1 Backward shift operator [𝑞 −1 𝑦(𝑡 + 1) = 𝑦(𝑡)]
s,z Complex variables (𝑧 = 𝑒 𝑠𝑇𝑠 )
A(q-1), B(q-1), C(q-1) Polynomials in the variable q-1
d Delay of the discrete-time system (integer number of sampling
period)
𝐴̂(𝑡, 𝑞 −1 ), 𝐵̂ (𝑡, 𝑞 −1 ), 𝐶̂ (𝑡, 𝑞 −1 ) Estimations of the polynomials A(q-1), B(q-1), C(q-1) at the
instant t
𝑎̂; (𝑡), 𝑏̂𝑖 (𝑡), 𝑐̂𝑖 (𝑡) Estimations of the coefficients of the polynomials A(q-1), B(q-
1
), C(q-1) (they are coefficients of the polynomials
𝐴̂(𝑡, 𝑞 −1 ), 𝐵̂ (𝑡, 𝑞 −1 ), 𝐶̂ (𝑡, 𝑞 −1 )
𝜃 Parameter vector
𝜃̂(𝑡) Estimated parameter vector
𝜑(𝑡) Measurement or observation vector
F,F(t) Adaptation gain
𝜖 𝑜 (𝑡), 𝜖(𝑡) A priori and a posteriori prediction error
𝑣 𝑜 (𝑡), 𝑣(𝑡) A priori and a posteriori adaptation error
A, M, F Matrices
F>0 Positive definite matrix
tR Rising time
tS Settling time
𝑀 Maximum overshoot
𝜔𝑜 , 𝜁(𝓌𝑛 , 𝓍𝑖 ) Natural frequency and damping coefficient of a second-order
system
E{.} Expectation
ARMAX Auto-regressive moving average with exogenous input
MV Mean value
Var. Variance
R(i) Auto correlation or cross-correlation
RN(i) Normalized autocorrelation or cross-correlation
𝒯 Time delay (continuous time systems)
PRBS Pseudo-random binary sequence

and it must satisfy Equation (1.1.1) for 𝑢(𝑡) = 𝑒 𝑠𝑡 . From Equation (1.1.6) one gets :
𝑑𝑦(𝑡)
= −𝑠𝐻(𝑠)𝑒 𝑠𝑡
𝑑𝑡
(1.1.7)
And by introducing Equation (1.1.7) in Equation (1.1.1) while bearing in mind that 𝑢(𝑡) =
𝑒 𝑠𝑡 , one obtains:
1 𝐺
(𝑠 + ) 𝐻(𝑠)𝑒 𝑠𝑡 = 𝑒 𝑠𝑡
𝑇 𝑇
(1.1.8)
H(s), which gives the gain and phase deviation introduced by the system shown in Equation
(1.1.31) at different complex frequencies, is known as the transfer function. From Equation
(1.1.8) it turns out that for the system described by Equation (1.1.1) the transfer function is :
𝐺
𝐻(𝑠) =
1 + 𝑠𝑇
(1.1.9)
Note that the transfer function H(s) can also be obtained by two other techniques:
- Replacing p with s in Equation (1.1.2) and algebraic computation of the y/u ratio
- Using the Laplace transform
The use of the representation of dynamic models in the form of transfer functions presents a
certain number of advantages for the analysis and synthesis of closed-loop control systems. In
particular:
- The roots of the denominator polynomial (the poles) and the roots of the numerator
polynomial (the zeros) allow the behavior of the system to be characterized in both
frequency and time domains.
- The concatenation of dynamic models described by transfer functions is extremely easy.

1.1.3 Stability
The stability of a dynamic system means the behavior of the system from an initial condition
for an identically input. Take as an example the first-order system describe by the differential
equation (1.1.1) or by the transfer function (1.1.9).
Consider the free response of the system (1.1.1) for u ≡ 0. From an initial condition y(0) =
y0:
𝑑𝑦 1
+ 𝑇 𝑦 = 0; 𝑦(0) =y0 (1.1.10)
𝑑𝑡

Section 1.1 Continuous-Time Models

A solution for y will be of the form:
𝑦(𝑡) = 𝐾𝑒 𝑠𝑡 (1.1.11)
in which K and s are to be determined.1
From Equation (1.1.11) :
𝑑𝑦
= 𝑠𝐾𝑒 𝑠𝑡 (1.1.12)
𝑑𝑡

is obtained, and Equation (1.1.10) becomes:

1
𝐾𝑒 𝑠𝑡 (𝑠 + ) = 0
𝑇
(1.1.13)
From which one obtains:
1
𝑠 = − ; 𝐾 = 𝑦0
𝑇
(1.1.14)
and, respectively;
𝑦(𝑡) = 𝑦0 𝑒 −𝑡/𝑇
(1.1.15)

1
The structure of solution (1.1.11) results from the theory of linear differential equations.
The response for T > 0 and T < 0 is illustrated in figure 1.1.3. For T > 0, we have s < 0 and the
output will tend toward zero (asymptotic stability). For T < 0, we have s > 0 and the output
will diverge (instability).
It is sign of the real part of the roots of the transfer function denominator that
determines the stability or instability of the system.
In the example of the first-order system, this root is s = -1/T.
For the system to be asymptotically stable, all the roots of the transfer function
denominator must be characterized by Re s < 0. If one or several roots of the transfer function
denominator are characterized by Re s < 0, then the system is unstable.
For Re s = 0, we have a limit case of stability [e.g., pure integrator, dy/dt = u(t)], since
the initial condition neither increase or decrease. Figure 1.1.4 gives the stability and instability
domains in the plane of the complex variable s.

1.1.4 Time Response

The response of a dynamic system for a step input is studied and characterized for a step
input. The response of a stable system shown in Figure 1.1.5 allows the parameters generally
used to characterize the step response to be introduced.

These parameters are:

tR, rise time, generally defined as the time needed to attain 90% of the final value (or as the
time needed for the output to pass from 10 to 90% of the final value).
Ts, settling time, defined as the time needed for the output to reach and remain within a
tolerance zone around the final value (±10%, ±5%, ±2%).
FV, final value, a fixed output value obtained for t →∞.
M, maximum overshoot, expressed as a percentage of the final value.
For example, consider the first order system:
𝐺
𝐻(𝑠) =
1 + 𝑠𝑇
Since the input is a unitary step:
FV = G(static gain); tR = 2.2T
ts =2.2T (for ±10% FV); M=0
are obtained and the response of such a system is represented in Figure 1.1.6. Note that for t =
T, the output reaches 63% of the final value.

1.1.5 Frequency Responses

The frequency response of a dynamic system is studied and characterized for periodic inputs
of variable frequency but of constant magnitude. The output characteristics are represented on
a double logarithmic scale for the gain-frequency characteristic and on a single logarithmic
scale for the phase-deviation frequency characteristic.
The gain G(𝜔) = |𝐻(𝑗𝜔)| is expressed in dB (20 log |𝐻(𝑗𝜔)|) and represented on the
vertical axis, and the frequency 𝜔 in rad/s (𝜔 = 2𝜋𝑓) is represented on the horizontal axis.
Figure 1.1.7 gives some typical frequency response curves for first and second-order systems.
The characteristic elements of the frequency response are:
𝑓𝛽 , bandwidth, thre frequency from which the zero-frequency (steady state) gain (0)
is attenuated by more than 3 dB, i.e., for 𝜔𝐵 = 2𝜋𝑓𝐵 , G(𝜔𝐵 )=G(0) – 3 dB.