Elements of feedback control systems

Chapter 2
Elements of feedback control systems
Signals & Systems
Linearity

Laplace transform

Convolution systems & LTI systems

Impulse response & stability
Step response & DC gain
Summary

21

22

Elements of feedback control systems

Signals, Systems, Control

signals convey information and can have many forms

greeting & talk: hello, how are your?, where are you from? have you eaten?

facial expressions: agree, disagree, puzzle

communications: email, phone, telegraph, video

a system usually involves two signals: input and output

drive a car

sleep in a room

buy a soda at vending machine

physical systems have actuators and sensors to respond to inputs

feedback and control are used to improve the performance of the system
a car has cruise control

a room has temperature control

a vending machine has programmable logic control

1

D. S. Bernstein, Introducing signals, systems, and control in grades K through 12, IEEE Control Systems Magazine, April 2003, pp. 10-12.

23

Elements of feedback control systems

Signals
a signal is a function of time, e.g.,
f is the force on some mass

vout is the output voltage of some circuit

p is the acoustic pressure at some point

notation used in the book:

f , vout, p
refer to the whole signal or function
f (t), vout(1.2), p(t + 2)
refer to the value of the signals at times t, 1.2, and t + 2, resp.
for times we usually use symbols like t, , t1, . . .

24

Elements of feedback control systems

Domain of a signal
domain of a signal domain of a signal: ts for which it is defined
some common domains:
all t, i.e., R

nonnegative t: t > 0
(here t = 0 just means some starting time of interest)
t in some interval: a t b

t at uniformly sampled points: t = kh + t0, k = 0, 1, 2, . . .

discrete-time signals are defined for integer t, i.e., t = 0, 1, 2, . . .

(here t means sample time or epoch, not real time in seconds)

well usually study signals defined on all reals, or for nonnegative reals

25

Elements of feedback control systems

Units & dimension of signal

physical units of a signal, e.g., V, mA, m/sec

sometimes the physical units are 1 (i.e., unitless) or unspecified

dimension or type of a signal u, e.g.,

real-valued or scalar signal: u(t) is a real number (scalar)

vector signal: u(t) is a vector of some dimension

Boolean signal: u(t) is either 0 or 1

well usually encounter scalar signals

26

Elements of feedback control systems

Common signals with names

a constant (or static or DC) signal: r(t) = a, where a is some constant

the unit impulse, (t), or Diracs delta function

the unit step signal,

0, t < 0
1, t 0

0, t < 0
t, t 0

r(t) =

(sometimes denoted 1(t) or u(t))

the unit ramp signal,
a rectangular pulse signal,

a sinusoidal signal:

r(t) =

u(t) =

1, a t b
0, otherwise

u(t) = a cos(t + )
a, , are called signal parameters

27

Elements of feedback control systems

Measuring the size of a signal

size is measured in many ways, e.g.,
integral square or total energy :

u(t)2 dt

|u(t)| dt

square root of total energy

integral-absolute value:

peak or maximum absolute value of a signal:

max |u(t)|
t

root-mean-square (RMS) value:

lim
T T

average-absolute value:

T
0

1/2

u(t)2 dt

1 ZT
|u(t)| dt
lim
T T 0
for some signals these measures can be infinite, or undefined

Elements of feedback control systems

Qualitative properties of signals

u decays if u(t) 0 as t

u converges if u(t) a as t (a is some constant)

u is bounded if its peak is finite

u is unbounded or blows up if its peak is infinite

in practice we are interested in more specific quantitative questions, e.g.,

how fast does u decay or converge?
how large is the peak of u?

28

29

Elements of feedback control systems

Systems
definition:
a system transforms input signals into output signals

a system is a function mapping input signals into output signals

we concentrate on systems with one input and one output signal, i.e., single-input,
single-output (SISO) systems

notation used in the book:

y = Gu means the system G acts on input signal u to produce output signal y

y = Gu does not (in general) mean multiplication!

2 10

Elements of feedback control systems

Block diagrams
systems often denoted by block diagram:
u

G(s)

lines with arrows denote signals (not wires)

boxes denote systems;

arrows show inputs & outputs

special symbols for some systems

2 11

Elements of feedback control systems

Examples with single input

scaling system: y(t) = au(t)

called an amplifier if |a| > 1

called an attenuator if |a| < 1
called inverting if a < 0

a is called the gain or scale factor

sometimes denoted by triangle or circle in block diagram:
u

differentiator: y(t) = u(t) = du/dt

u

d
dt

2 12

Elements of feedback control systems

integrator: y(t) =

t
a u( ) d

(a is often 0 or -)

time shift system: y(t) = u(t T )

called a delay system if T > 0
called a predictor system if T < 0

sign detector or 1-bit limiter system:

y(t) = sgn(u(t)) =
u

1, u(t) 0
1, u(t) < 0
y

2 13

Elements of feedback control systems

Examples with multiple inputs

with inputs u1, u2, and output y
summing system: y(t) = u1(t) + u2(t)
y

u1

u2
difference system: y(t) = u1(t) u2(t)
u1

u2

2 14

Elements of feedback control systems

multiplier system: y(t) = u1(t)u2(t)

u1
u2
comparator system:

y(t) =

1, u1(t) u2(t)
1, u1(t) < u2(t)

u1

u2

2 15

Elements of feedback control systems

Electronic realizations
systems described previously can be realized as electronic circuits with op-amps
scaling: y(t) = (1 + R2/R1 )u(t)
+
u

R1

y
R2

difference: y(t) = u1(t) u2(t)

R
+

R
+
u1

R
+u
2

y
R

2 16

Elements of feedback control systems

integrator: y(t) = 1/(RC)

t
0 u( ) d

R
+
u

these are circuit schematics, not block diagrams

signals are represented by voltages (which is common but not universal)

2 17

Elements of feedback control systems

Linearity
a system F is linear if the following properties hold.
homogeneity: if u is any signal and a is any scalar,
F (au) = aF (u)
superposition: if u and u are any two signals,
F (u + u) = F u + F u
(watch out, just a few symbols here express a very complex meaning)
in words, linearity means scaling before or after the system is the same
y
y
u
u
a
a
F
F
summing before or after the system is the same
u1
u1
y
u2
F
u2

F
F

Elements of feedback control systems

examples of linear systems

scaling system
differentiator
integrator

running average

time shift (delay)

summer

difference
examples of nonlinear systems
sign detector
multiplier

comparator

2 18

2 19

Elements of feedback control systems

Interconnections of systems
we can interconnect systems to form new systems, e.g.,
cascade or series: y = G(F u) = GF u
u

(note that block diagrams and algebra are reversed )

sum or parallel: y = F u + Gu

F
G

2 20

Elements of feedback control systems

feedback: y = F (u Gy)
u

G
the minus sign is just a tradition, and often isnt there
in general, block diagrams are just a symbolic way to describe a connection of systems

Elements of feedback control systems

2 21

Systems by differential equations

many systems are described by a linear constant coefficient ordinary differential
equation (LCCODE):
any (n) + + a1y + a0y = bmu(m) + + b1u + b0u
with given initial conditions
y (n1)(0), y (n2)(0), . . . , y (0), y(0)
and given u
n is called the order of the system
b0, . . . , bm, a0, . . . , an are the coefficients of the system

when initial conditions are all zero, LCCODE systems are linear
an LCCODE gives an implicit description of a system; soon well be able to explicitly
express y in terms of u
example: first- and second-order system
dy
+ 2y = u
dt
dy
d2 y
+2 +y = u
dt2
dt

2 22

Elements of feedback control systems

Mass-spring-damper system
it represents vehicle suspension system, building during earthquake, . . .
simple model of vehicle suspension system
y
m
k

b
u

input u is road height (along vehicle path), output y is vehicle height

vehicle dynamics:

my + by + ky = bu + ku

2 23

Elements of feedback control systems

Laplace Transform
Idea: the Laplace transform converts integral and differential equations into algebraic
equations

this is like phasors, but

applies to general signals, not just sinusoids
handles non-steady-state conditions

the Laplace transform allows us to analyze

LCCODEs

complicated linear circuits with sources, Ls, Rs, and Cs

complicated linear systems with integrators and differentiators, gains, etc.

2 24

Elements of feedback control systems

Definition of Laplace Transform

well be interested in signals defined for
t0

the Laplace transform of a signal f is the function F = L(f ) defined by

Z

F (s) = 0 f (t)est dt
for those s C for which the integral makes sense

F is a complex-valued function of complex numbers

t is called the time variable (in sec)

s is called the complex variable or frequency variable with units sec1, i.e.,
frequency
st is unitless

common notation: by convention,

lower case letter denotes signal

capital letter denotes its Laplace transform

example: U denotes L(u), Vin denotes L(vin), etc.

2 25

Elements of feedback control systems

Laplace transform of et
lets find Laplace transform of f (t) = et
F (s) =
=

t st
dt
0 ee
Z
(1s)t
dt
0 e


1 (1s)t

=
e


1s
0
1
=
s1
provided we can say e(1s)t 0 as t , which is true for Re s > 1 (why?)
that is to say,
the integral defining F makes sense for s C with Re s > 1

but the resulting formula for F makes sense for all s C except s = 1

in practice, these important details are not mentioned, and we simply say that the
Laplace transform of et is 1/(s 1), or
1
t
(e
)
=
L
s1

2 26

Elements of feedback control systems

More examples
impulse function: f (t) = (t)
F (s) =

st
(t)e
dt

unit step: f (t) = 1 (for t 0)

=1

1 st 1
F (s) =
= e  =
s
s
0
integral makes sense for Re s > 0; formula makes sense for s 6= 0
st
dt
0 e

sinusoid function
first express f (t) = cos(t) as
1 jt 1 jt
f (t) = e + e
2
2

2 27

Elements of feedback control systems

now we can find F (s) as

1
1
jt
[e
]
+
L
L[ejt]
2
2

1 1 1 1
=
+

2 s j
2 s + j
s
= 2
s + 2
F (s) is valid for Re s > 0; but formula makes sense for s 6= j
F (s) =

similarly, Laplace transform of f (t) = sin(t) is

F (s) = 2
s + 2

2 28

Elements of feedback control systems

Properties of Laplace transform

the Laplace transform is linear
the definition of linearity
if f and g are any signals, and a is any scalar, we have
L(af ) = aF,

L(f + g) = F + G

i.e., homogeneity & superposition hold for the Laplace transform

example: consider the signal for t > 0,
2
3s 5
t
=
L(3(t) 2e ) = 3
s1
s1
the Laplace transform is one-to-one, i.e.,
if L(f ) = L(g) then f = g
F determines f

the inverse Laplace transform L1 is well defined

t
example: the only function f such that F (s) = 3s5
s1 is f (t) = 3(t) 2e
in other words,

3s

5
t

= 3(t) 2e
L1
s1

2 29

Elements of feedback control systems

Inverse Laplace transform

in principle, we can recover f (t) from F (s) via
1 Z +j
st
f (t) =
F
(s)e
ds
j
2j
where is large enough that F (s) is defined for Re(s) 0
Im + j

Re

note that the integration of F (s)est is calculated along the dash line in the splane
surprisingly, this formula is not widely used, except when we do numerical inverse of
Laplace transform

2 30

Elements of feedback control systems

Time scaling
let f be a signal, define signal g by
g(t) = f (at)
where a > 0, then
G(s) = (1/a)F (s/a)
this expression makes sense because times are scaled by a, frequencies scaled by 1/a

example: L(et) = 1/(s 1) so

at
L(e ) = (1/a)

1
1
=
(s/a) 1 s a

time-scaling formula requires a > 0, but this is true for all values of a

2 31

Elements of feedback control systems

Derivative
if signal f (t) is differentiable for t 0, then

L(f ) = sF (s) f (0)

time-domain differentiation becomes multiplication by frequency variable s (like

that appears in phasors)
plus a term that includes initial condition, i.e., f (0)
example: f (t) = et, so f (t) = et and
1
1

1
=
(f
)
=
s
L
s1
s1

example: sin(t) = dtd cos(t), so

1
s

1 = 2
L(sin t) = s 2
s +1
s +1

2 32

Elements of feedback control systems

Higher-order derivatives: applying derivative formula twice yields

L(f ) = s L(f ) f (0)

= s(sF (s) f (0)) f (0)
= s2F (s) sf (0) f (0)

last two terms are due to initial conditions

similar formulas hold for L(f (k))

2 33

Elements of feedback control systems

Integral
let g be the running integral of a signal f , i.e.,
g(t) =

t
0 f ( ) d

then

1
G(s) = F (s)
s
i.e., time-domain integral becomes division by frequency variable s
example: we know that the Laplace transform of the unit delta function
L((t)) = 1

and the integral of the unit delta function is the unit step function, u(t), hence,
L(u(t)) =

1
s

example: the integral of the unit step function is the unit ramp function, t,
1
(t)
=
L
s2

2 34

Elements of feedback control systems

Time delay
let f be a signal and T > 0
define the signal g as

g(t) =

g is f , delayed by T seconds

0,
0t<T
f (t T ), t T

G(s) = 0 estg(t) dt
Z

= T estf (t T ) dt
Z

= 0 es( +T )f ( ) d
= esT F (s)
Z

where = t T
f

sT

2 35

Elements of feedback control systems

example: lets find the Laplace transform of f (t)

f (t) =

1, a t b
0, otherwise

where 0 < a < b (f is rectangular pulse signal)

u(t a)
a

b
u(t b)

note that f (t) = u(t a) u(t b), hence

as

F (s) = e

bs

(1/s) e

eas ebs
(1/s) =
s

2 36

Elements of feedback control systems

Multiplication by t
let f be a signal and define g = tf (t), then we have
G(s) = F (s)
i.e., multiplication by time-domain variable corresponds to differentiation in the
frequency domain
c.f., L(f ) = sF (s) f (0)
example: L(et) = 1/(s + 1)
L(tet) =

1
d 1
=
ds s + 1 (s + 1)2

more generally, we have

hence,

k
k (k)
L(t f (t)) = (1) F (s)
k t
L(t e ) =

(k 1)!
(s + 1)k+1

2 37

Elements of feedback control systems

Exponential scaling
let f be a signal and a a scalar, and define
g(t) = eatf (t),
then

st at
L(g(t)) = 0 e e f (t) dt
Z

= 0 e(sa)tf (t) dt
= F (s a)
Z

i.e., exponential scaling in the time domain corresponds to a shift (by a) in the
frequency domain
example: L(cos t) = s/(s2 + 1),

(s + 1)
s+1
=
L(e cos t) =
(s + 1)2 + 1 s2 + 2s + 2
t

2 38

Elements of feedback control systems

Convolution
the convolution of signals f and g, denoted h = f g, is the signal
h(t) =
same as h(t) =

t
0 f (t

t
0 f ( )g(t

) d

)g( ) d

symmetric: f g = g f

its very important and soon will become clear

L(h(t)) = L(f g)
= L(f ) L(g)
that is, H(s) = F (s)G(s) or equivalently,
L1(F (s)G(s)) = f g
frequency-domain multiplication corresponds to time-domain convolution
Laplace transform turns convolution into multiplication

2 39

Elements of feedback control systems

Finding Laplace transform

we should know the Laplace transform of some basic signals, e.g.,
unit delta (F (s) = 1), unit step (F (s) = 1/s)

exponential fct, f (t) = eat (F (s) = 1/(s a))

sinusoids,

L(cos(t)) =

s
,
2
2
s +

L(sin(t)) =

s2 + 2

these basic pairs combined with

a table of Laplace transforms

linearity, scaling, and other properties given above

will help us derive other signals

of course, we can always integrate using the defining formula.

2 40

Elements of feedback control systems

Patterns
while the details of the Laplace transform pairs differ, we can see some interesting
symmetric patterns between
the time domain (i.e., signals) and

the frequency domain (i.e., their Laplace transforms)

example: let L(f (t)) = F (s)

differentiation in one domain corresponds to multiplication by the variable in the

other

L(f ) = sF (s) f (0),

L(tf (t)) = F (s)

multiplication by an exponential in one domain corresponds to a shift (or delay) in

the other
sT
at
L(f (t T )) = e F (s), L(e f (t)) = F (s a)

well see these patterns and others throughout the course

2 41

Elements of feedback control systems

Table of Laplace transforms

Time Function f (t) Laplace Transform F (s)
(t)
1
1 (unit step)
t (unit ramp)
eat
sin t
cos t

1/s
1/s2
1/(s + a)
/(s2 + 2)
s/(s2 + 2)

example: given 0 < < 1, consider the Laplace transform

2
F (s) =
s(s2 + 2s + 2)
determine the inverse Laplace transform
of F (s)

solution: let = and d = 1 2.

1
t
f (t) = 1
e
sin(dt + )
2
1
where = tan1

2 42

Elements of feedback control systems

Convolution systems
consider convolution system with input u, output y satisfying
y = hu
Z
t
= 0 h( )u(t ) d
Z
t
= 0 h(t )u( ) d
in the frequency domain:
Y (s) = H(s)U (s)
H(s) is called the transfer function (TF) of the system

h(t) is called the impulse response or convolution kernel of the system

also called transfer function system or filter

block diagram notation:

u

y
h

convolution systems are linear

y
H

2 43

Elements of feedback control systems

Properties
composition of convolution systems corresponds to
multiplication of transfer functions
convolution of impulse responses

u
A

u
BA

other properties:
we can manipulate block diagrams with transfer functions as if they were simple
gains
convolution systems commute with each other

2 44

Elements of feedback control systems

Examples
first order LCCODE:
y + y = u, y(0) = 0
take Laplace transform to get
sY (s) y(0) + Y (s) = U (s)
after simple algebra, we obtain
1
U (s)
s+1
transfer function H(s) = 1/(s + 1); impulse response h(t) = et
delay: with T > 0,

0,
0t<T
y(t) =
u(t T ), t T
Y (s) =

transfer function H(s) = esT ; impulse response h(t) = (t T )

integrator:
Z
t
y(t) = 0 u( ) d

transfer function H(s) = 1/s; impulse response h(t) = u(t) (unit step function)

2 45

Elements of feedback control systems

Linear time-invariant systems

a system F is time-invariant if it commutes with delays, i.e., if y(t) = F (u(t)), then
y(t T ) = F (u(t T ))
for any T
y

u
esT

F
y

u
F

esT

a system F is linear time-invariant (LTI) if it is

linear (F satisfies homogeneity & superposition)
time-invariant (F commutes with delays)

example: any convolution system is LTI

the converse is also true: any LTI system can be represented by a convolution system

2 46

Elements of feedback control systems

Feedback connection
what is transfer function H from u to y?
u

take Laplace transform:

Y = G(U Y )

solve for Y to get

G(s)
Y (s) =
U (s)
1 + G(s)
as if G were a simple scaling system
so we have
G(s)
H(s) =
1 + G(s)
in time domain, we have complicated integral equation
y(t) =

t
0 g(t

)(u( ) y( )) d

which is not easy to understand or solve . . .

2 47

Elements of feedback control systems

Mass-spring-damper system
it represents vehicle suspension system, building during earthquake, . . .
simple model of vehicle suspension system
y
m
k

b
u

input u is road height (along vehicle path), output y is vehicle height

vehicle dynamics:

my + by + ky = bu + ku

assuming y(0) = 0, y (0) = 0, and u(0) = 0, we obtain

(ms2 + bs + k)Y = (bs + k)U
Transfer function from road height u to vehicle height y is
bs + k
Y (s)
=
H(s) =
U (s) ms2 + bs + k

2 48

Elements of feedback control systems

Impulse response
if u = , we have
y(t) =

t
0

h(t )( ) d = h(t)

so h is the output response when u = (hence h has the name impulse response)
u

y
H

impulse response testing:

apply impulse input & record resulting output (h)

now we can predict output for any input signal

practical problem: linear model often fails for very large input signals

2 49

Elements of feedback control systems

Step response
the unit step response is the output when the input is a unit step, i.e., u(t) = 1
s(t) =

t
0 h( ) d

s(t) is the integral of impulse response

symbol s clashes with frequency variable (but usually it is quite clear because of its
context)
step response testing is widely used in control applications

in time domain, h(t) = s(t)

in frequency domain, S(s) = H(s)/s

2 50

Elements of feedback control systems

example: H(s) =

1
s+1

h(t) = et,

s(t) = 1 et
1
Step Response

Impulse Response

0.5

0
0

2
4
Time (sec)

0.5

0
0

2
4
Time (sec)

2 51

Elements of feedback control systems

DC gain
if a convolution system is stable and
H(0) =

0 h( ) d

is bounded, H(0) is called the DC (direct current) or static gain

if u is constant, then for large t,
y(t) = u

t
0 h( ) d

H(0)u

so H(0) gives the gain for static signals in terms of step response
Final value theorem:
given F (s) = L(f (t)), if limt f (t) exists, then

lim f (t) = lim sF (s)

s0

let h(t) be the impulse response of a convolution system, and H = L(h),

lim s(t) = lim sS(s) = H(0)

example: H(s) =

1
s+1

s0

DC gain is equal to 1

2 52

Elements of feedback control systems

Stability
we say that the convolution system is stable if impulse response h(t) 0 as t
if H(s) = L(h(t)) is rational, stability means poles are in the left-half s-plane (LHP)
example:

h(t) = e2t,

H(s) =

1
s+2

Impulse Response
1

0.9

0.8

0.7

Amplitude

0.6

0.5

0.4

0.3

0.2

0.1

0.5

1.5
Time (sec)

2.5

2 53

Elements of feedback control systems

Summary
signals convey information. systems transform input signals into output signals

a system F is linear if it satisfies homogeneity and superposition

the Laplace transform converts integral and differential equations into algebraic
equations
a system is linear time-invariant (LTI) if it is linear (it satisfies homogeneity &
superposition) and time-invariant (it commutes with delays)
the impulse response is the output of the system when input is a unit delta

the step response is the output of the system when the input is a unit step