Feedback Control Systems

Fall 2020

September 15, 2020

In This Lecture

Disturbances

r e u y
Controller System
−

ym

Measurements

Objective
I Fill
System Response

I In the previous lecture we saw examples of dynamic models of

systems - mostly in the form of differential equations
System Response

I In the previous lecture we saw examples of dynamic models of

systems - mostly in the form of differential equations
I Approximate linear models help us explain the way a system
behaves
System Response

I In the previous lecture we saw examples of dynamic models of

systems - mostly in the form of differential equations
I Approximate linear models help us explain the way a system
behaves
I Exact models are used to numerically simulate the response of a
dynamic system
System Response

I In the previous lecture we saw examples of dynamic models of

systems - mostly in the form of differential equations
I Approximate linear models help us explain the way a system
behaves
I Exact models are used to numerically simulate the response of a
dynamic system
I We will see linear analysis techniques
Laplace Transform

Linear time-invariant (LTI) system analysis rests on two properties:

I A linear system obeys the superposition
Laplace Transform

Linear time-invariant (LTI) system analysis rests on two properties:

I A linear system obeys the superposition
I The response of an LTI system is the convolution of the input
with the unit impulse response of the system
Response by Convolution
Superposition
I While at rest, a system produces output y1 to input u1
u1 (t) y1 (t)
System
Response by Convolution
Superposition
I While at rest, a system produces output y1 to input u1
u1 (t) y1 (t)
System

I The same system produces output y2 to input u2

u2 (t) y2 (t)
System
Response by Convolution
Superposition
I While at rest, a system produces output y1 to input u1
u1 (t) y1 (t)
System

I The same system produces output y2 to input u2

u2 (t) y2 (t)
System

I Then, the system produces output α1 y1 (t) + α2 y2 (t) to input

α1 u1 (t) + α2 u2 (t)
α1 u1 (t) + α2 u2 (t) α1 y1 (t) + α2 y2 (t)
System
Response by Convolution
Superposition
I While at rest, a system produces output y1 to input u1
u1 (t) y1 (t)
System

I The same system produces output y2 to input u2

u2 (t) y2 (t)
System

I Then, the system produces output α1 y1 (t) + α2 y2 (t) to input

α1 u1 (t) + α2 u2 (t)
α1 u1 (t) + α2 u2 (t) α1 y1 (t) + α2 y2 (t)
System

I If the input can be expressed as a sum of signals, then, the

output can be expressed as the sum of the individual responses to
these input signals
Superposition: Example

u y
ẏ + ky = u
Superposition: Example

u y
ẏ + ky = u

I Apply input: u = α1 u1 + α2 u2
Superposition: Example

u y
ẏ + ky = u

I Apply input: u = α1 u1 + α2 u2
I Expect output: y = α1 y1 + α2 y2 . Then, ẏ = α1 ẏ1 + α2 ẏ2
Superposition: Example

u y
ẏ + ky = u

I Apply input: u = α1 u1 + α2 u2
I Expect output: y = α1 y1 + α2 y2 . Then, ẏ = α1 ẏ1 + α2 ẏ2
I Superposition holds
Time Invariance
I In brief: The system response is independent of time
Time Invariance
I In brief: The system response is independent of time
I Consider the same system with k possibly a function of time:
u y
ẏ + ky = u
Time Invariance
I In brief: The system response is independent of time
I Consider the same system with k possibly a function of time:
u y
ẏ + ky = u

I Apply input at t1 :

ẏ1 (t1 ) + k(t1 )y1 (t1 ) = u1 (t1 )

Time Invariance
I In brief: The system response is independent of time
I Consider the same system with k possibly a function of time:
u y
ẏ + ky = u

I Apply input at t1 :

ẏ1 (t1 ) + k(t1 )y1 (t1 ) = u1 (t1 )

I If we had applied the same input before, e.g., at time t2 = t1 − τ

for an arbitrary τ > 0:

ẏ2 (t2 ) + k(t1 )y2 (t2 ) = u1 (t1 − τ )

Time Invariance
I In brief: The system response is independent of time
I Consider the same system with k possibly a function of time:
u y
ẏ + ky = u

I Apply input at t1 :

ẏ1 (t1 ) + k(t1 )y1 (t1 ) = u1 (t1 )

I If we had applied the same input before, e.g., at time t2 = t1 − τ

for an arbitrary τ > 0:

ẏ2 (t2 ) + k(t1 )y2 (t2 ) = u1 (t1 − τ )

I Assume y2 (t2 ) = y1 (t1 − τ ):

dy1 (t1 − τ )
+ k(t2 + τ )y1 (t1 − τ ) = u1 (t1 − τ )
dt
Time Invariance

u y
ẏ + ky = u

I System at t1 :

ẏ1 (t1 ) + k(t1 )y1 (t1 ) = u1 (t1 )

I Delayed system at t2 = t1 − τ :

dy1 (t1 − τ )
+ k(t2 + τ )y1 (t1 − τ ) = u1 (t1 − τ )
dt
Time Invariance

u y
ẏ + ky = u

I System at t1 :

ẏ1 (t1 ) + k(t1 )y1 (t1 ) = u1 (t1 )

I Delayed system at t2 = t1 − τ :

dy1 (t1 − τ )
+ k(t2 + τ )y1 (t1 − τ ) = u1 (t1 − τ )
dt
I These two are equal iff τ = 0 or k is constant. Then, the system
is said to be time-invariant.
Impulse Response
I By the superposition property we can solve for the response of a
system to given inputs
Impulse Response
I By the superposition property we can solve for the response of a
system to given inputs
I There are special classes of signals which ease the analysis:
Impulse and exponential
Impulse Response
I By the superposition property we can solve for the response of a
system to given inputs
I There are special classes of signals which ease the analysis:
Impulse and exponential
I Consider a unit-step input for a short duration:

1/∆, if 0 ≤ t ≤ ∆
p∆ (t) =
0, otherwise
Impulse Response
I By the superposition property we can solve for the response of a
system to given inputs
I There are special classes of signals which ease the analysis:
Impulse and exponential
I Consider a unit-step input for a short duration:

1/∆, if 0 ≤ t ≤ ∆
p∆ (t) =
0, otherwise

I If we apply p∆ (t), we get h∆ (t). Then:

αp∆ (t) αh∆ (t)
System
Impulse Response
I By the superposition property we can solve for the response of a
system to given inputs
I There are special classes of signals which ease the analysis:
Impulse and exponential
I Consider a unit-step input for a short duration:

1/∆, if 0 ≤ t ≤ ∆
p∆ (t) =
0, otherwise

I If we apply p∆ (t), we get h∆ (t). Then:

αp∆ (t) αh∆ (t)
System

I Similarly:
p∆ (t − ∆t) h∆ (t − ∆t)
System
Impulse Response
I Any input signal can be approximated as a series of pulses p∆ (t):
Impulse Response
I Any input signal can be approximated as a series of pulses p∆ (t):

I The response to the signal ∆u(k∆)p∆ (k∆) at time n∆ is:

∆u(k∆)h∆ (∆n − ∆k)

Impulse Response
I Any input signal can be approximated as a series of pulses p∆ (t):

I The response to the signal ∆u(k∆)p∆ (k∆) at time n∆ is:

∆u(k∆)h∆ (∆n − ∆k)

I The output at time k∆ is the sum of the responses to the inputs

applied at or before time k∆ (by superposition):
k=∞
X
y(t) = ∆u(k∆)h∆ (t − ∆k)
k=0
Impulse Response

I Let’s squeeze the duration:

lim p∆ (t) = δ(t)

∆→0
Impulse Response

I Let’s squeeze the duration:

lim p∆ (t) = δ(t)

∆→0

I The dirac delta function δ has the following properties:

Z ∞ = 0, t 6= 0
1. δ(t)
2. δ(t)dt = 1
−∞
Impulse Response

I Let’s squeeze the duration:

lim p∆ (t) = δ(t)

∆→0

I The dirac delta function δ has the following properties:

Z ∞ = 0, t 6= 0
1. δ(t)
2. δ(t)dt = 1
−∞
I In the limit, the summation becomes integral:
Z ∞
y(t) = u(τ )h(t − τ )dτ
0
Impulse Response
I Any continuous function can be represented as the integral of
impulses:
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞
Impulse Response
I Any continuous function can be represented as the integral of
impulses:
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞

I Thus, can write the input u as an integral of impulses:

Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞
Impulse Response
I Any continuous function can be represented as the integral of
impulses:
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞

I Thus, can write the input u as an integral of impulses:

Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞

I Remember the superposition property: If the input can be

expressed as a sum of signals, then the output can be expressed as
the sum of the individual responses to these input signals. Thus:
Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞
↓
Z ∞ Z ∞
y(t) = u(τ )h(t − τ )dτ = h(τ )u(t − τ )dτ
−∞ −∞
Impulse Response
I Any continuous function can be represented as the integral of
impulses:
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞

I Thus, can write the input u as an integral of impulses:

Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞

I Remember the superposition property: If the input can be

expressed as a sum of signals, then the output can be expressed as
the sum of the individual responses to these input signals. Thus:
Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞
↓
Z ∞ Z ∞
y(t) = u(τ )h(t − τ )dτ = h(τ )u(t − τ )dτ
−∞ −∞

I Convolution of u and y
Convolution
I Consider the system ẏ + ky = u = δ(t) with y(0) = 0.
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞
Convolution
I Consider the system ẏ + ky = u = δ(t) with y(0) = 0.
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞

I Thus, can write the input u as an integral of impulses:

Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞
Convolution
I Consider the system ẏ + ky = u = δ(t) with y(0) = 0.
Z ∞
f (t) = f (τ )δ(t − τ )dτ
−∞

I Thus, can write the input u as an integral of impulses:

Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞

I Remember the superposition property: If the input can be

expressed as a sum of signals, then the output can be expressed as
the sum of the individual responses to these input signals. Thus:
Z ∞
u(t) = u(τ )δ(t − τ )dτ
−∞
↓
Z ∞ Z ∞
y(t) = u(τ )h(t − τ )dτ = h(τ )u(t − τ )dτ
−∞ −∞
Convolution

I The response of a system can be expressed as the convolution of

u and y

y(t) = y ? h = h ? y
Z ∞ Z ∞
= u(τ )h(t − τ )dτ = h(τ )u(t − τ )dτ
−∞ −∞
Convolution

I The response of a system can be expressed as the convolution of

u and y

y(t) = y ? h = h ? y
Z ∞ Z ∞
= u(τ )h(t − τ )dτ = h(τ )u(t − τ )dτ
−∞ −∞

I Since the right-hand-side of the time axis is of interest, can write:

Z ∞
y(t) = u(τ )h(t − τ )dτ
0
Convolution

I The response of a system can be expressed as the convolution of

u and y

y(t) = y ? h = h ? y
Z ∞ Z ∞
= u(τ )h(t − τ )dτ = h(τ )u(t − τ )dτ
−∞ −∞

I Since the right-hand-side of the time axis is of interest, can write:

Z ∞
y(t) = u(τ )h(t − τ )dτ
0

I Can be difficult to calculate for an arbitrary input signal

Laplace Transform and Transfer Functions
I The two-sided Laplace Transform of y(t):
Z ∞
Y (s) = y(t)e−st dt
−∞
Laplace Transform and Transfer Functions
I The two-sided Laplace Transform of y(t):
Z ∞
Y (s) = y(t)e−st dt
−∞

I Let’s apply to the convolution:

Z ∞ Z ∞ 
Y (s) = h(τ )u(t − τ )dτ e−st dt
−∞ −∞
Laplace Transform and Transfer Functions
I The two-sided Laplace Transform of y(t):
Z ∞
Y (s) = y(t)e−st dt
−∞

I Let’s apply to the convolution:

Z ∞ Z ∞ 
Y (s) = h(τ )u(t − τ )dτ e−st dt
−∞ −∞

I Change the order of integrals

Z ∞ Z ∞ 
Y (s) = u(t − τ )e−st dt h(τ )dτ
−∞ −∞
Z ∞ Z ∞ 
−s(t̄+τ )
= u(t̄)e dt h(τ )dτ
−∞ −∞
Z ∞ Z ∞
= u(t̄)e−st̄ dt h(τ )e−sτ dτ
−∞ −∞
| {z }| {z }
U (s) H(s)
Laplace Transform and Transfer Functions
I The convolution operation is converted to a simple
transformation:

Y (s) = H(s)U (s)

I U (s) : Laplace transform of the input

I H(s) : Transfer function of the system
Laplace Transform and Transfer Functions
I The convolution operation is converted to a simple
transformation:

Y (s) = H(s)U (s)

I U (s) : Laplace transform of the input

I H(s) : Transfer function of the system
I s may be complex (s = σ + jω)
Laplace Transform and Transfer Functions
I The convolution operation is converted to a simple
transformation:

Y (s) = H(s)U (s)

I U (s) : Laplace transform of the input

I H(s) : Transfer function of the system
I s may be complex (s = σ + jω)
I Example: U (s) = est
Z ∞
y(t) = h(τ )u(t − τ )dτ
−∞
Z ∞
= h(τ )es(t−τ ) dτ
−∞
Z ∞
= h(τ )e−sτ dτ est
−∞
= H(s)est
Laplace Transform and Transfer Functions

I If the input is exponential, then the output is an exponential

signal, and the scaling factor is H(s)

y(t) = H(s)est
Laplace Transform and Transfer Functions

I If the input is exponential, then the output is an exponential

signal, and the scaling factor is H(s)

y(t) = H(s)est

I For causal systems, h(t) = 0 for t < 0. Therefore,

Z ∞
H(s) = h(τ )e−sτ dτ
0

and
Z ∞
y(t) = h(τ )u(t − τ )dτ
0
Laplace Transform and Transfer Functions

Example: Find the transfer function for the system

ẏ(t) + ky(t) = u(t) if u(t) = est for all time.
I Assume y(t) = H(s)est . Then,

ẏ(t) = sH(s)est
Laplace Transform and Transfer Functions

Example: Find the transfer function for the system

ẏ(t) + ky(t) = u(t) if u(t) = est for all time.
I Assume y(t) = H(s)est . Then,

ẏ(t) = sH(s)est

I The system equation becomes

sH(s)est + kH(s)est = est

Laplace Transform and Transfer Functions

Example: Find the transfer function for the system

ẏ(t) + ky(t) = u(t) if u(t) = est for all time.
I Assume y(t) = H(s)est . Then,

ẏ(t) = sH(s)est

I The system equation becomes

sH(s)est + kH(s)est = est

I Yields:
1
H(s) =
s+k
est
y(t) =
s+k
Laplace Transform and Transfer Functions
I The transfer function is the ratio of the Laplace transform of the
output to the Laplace transform of the input
Y (s)
H(s) =
U (s)
with the assumption that all initial conditions in the system are
zero.
Laplace Transform and Transfer Functions
I The transfer function is the ratio of the Laplace transform of the
output to the Laplace transform of the input
Y (s)
H(s) =
U (s)
with the assumption that all initial conditions in the system are
zero.
I It defines the relation between the input and the output
Laplace Transform and Transfer Functions
I The transfer function is the ratio of the Laplace transform of the
output to the Laplace transform of the input
Y (s)
H(s) =
U (s)
with the assumption that all initial conditions in the system are
zero.
I It defines the relation between the input and the output
I Example:
...
y + a1 ÿ + a2 ẏ + a3 y = b1 ü + b2 u̇ + b3 u

Assume all I.C. are zero...

(y(0) = ẏ(0) = ÿ(0) = y = u(0) = u̇(0) = ü(0) = 0)
Laplace Transform and Transfer Functions
I The transfer function is the ratio of the Laplace transform of the
output to the Laplace transform of the input
Y (s)
H(s) =
U (s)
with the assumption that all initial conditions in the system are
zero.
I It defines the relation between the input and the output
I Example:
...
y + a1 ÿ + a2 ẏ + a3 y = b1 ü + b2 u̇ + b3 u

Assume all I.C. are zero...

(y(0) = ẏ(0) = ÿ(0) = y = u(0) = u̇(0) = ü(0) = 0)
I

s3 Y (s) + a1 s2 Y (s) + a2 sY (s) + a3 Y (s) = b1 s2 U (s) + b2 sU (s) + b3 U (s)

Y (s) b1 s2 + b2 s + b3
H(s) = = 3
U (s) s + a1 s2 + a2 s + a3
Laplace Transform and Transfer Functions

Example: RC circuit
dy(t)
Ri(t) + y(t) = u(t), i(t) = C
dt
RC ẏ(t) + y(t) = u(t)
Laplace Transform and Transfer Functions

Example: RC circuit
dy(t)
Ri(t) + y(t) = u(t), i(t) = C
dt
RC ẏ(t) + y(t) = u(t)

I Take the Laplace transform of both sides for u(t) = δ(t)

RC(sY (s) − y(0)) + Y (s) = U (s) = 1
with the assumption that y(0) = 0
Laplace Transform and Transfer Functions

Example: RC circuit
dy(t)
Ri(t) + y(t) = u(t), i(t) = C
dt
RC ẏ(t) + y(t) = u(t)

I Take the Laplace transform of both sides for u(t) = δ(t)

RC(sY (s) − y(0)) + Y (s) = U (s) = 1
with the assumption that y(0) = 0
I Then,
Y (s) 1
H(s) = =
U (s) RCs + 1
Frequency Response

I Will analyze the response in the frequency domain by applying a

sinusoidal input to the system
A jωt
A cos(ωt) = (e + e−jωt )
2
Frequency Response

I Will analyze the response in the frequency domain by applying a

sinusoidal input to the system
A jωt
A cos(ωt) = (e + e−jωt )
2
I Consider s = jω:

u(t) = ejωt → y(t) = H(jω)ejωt

u(t) = e−jωt → y(t) = H(−jω)e−jωt
Frequency Response

I Will analyze the response in the frequency domain by applying a

sinusoidal input to the system
A jωt
A cos(ωt) = (e + e−jωt )
2
I Consider s = jω:

u(t) = ejωt → y(t) = H(jω)ejωt

u(t) = e−jωt → y(t) = H(−jω)e−jωt

I By superposition:
A
y(t) = (H(jω)ejωt + H(−jω)e−jωt )
2
Frequency Response

I Can represent any complex number in the polar form:

H(jω) = |H(jω)| H(jω)

= M (ω)ejϕ(ω)
Frequency Response

I Can represent any complex number in the polar form:

H(jω) = |H(jω)| H(jω)

= M (ω)ejϕ(ω)

I Rewrite the response:

A
y(t) = (H(jω)ejωt + H(−jω)e−jωt )
2
= M A cos(ωt + ϕ)
Frequency Response

I Can represent any complex number in the polar form:

H(jω) = |H(jω)| H(jω)

= M (ω)ejϕ(ω)

I Rewrite the response:

A
y(t) = (H(jω)ejωt + H(−jω)e−jωt )
2
= M A cos(ωt + ϕ)

I If a system with transfer function H(s) is excited by a sinusoidal

input with magnitude A, then the magnitude of the response will
be M A, and the phase of the input signal will be shifted by the
angle ϕ.
Inverse Laplace Transform

I Why do we study the frequency response?

Inverse Laplace Transform

I Why do we study the frequency response?

I One of the ultimate goals of feedback control analysis is to find
the response of a system to any type of input signal
Inverse Laplace Transform

I Why do we study the frequency response?

I One of the ultimate goals of feedback control analysis is to find
the response of a system to any type of input signal
Z ∞
I Consider the Laplace transform F (s) = f (t)e−st dt
−∞
Inverse Laplace Transform

I Why do we study the frequency response?

I One of the ultimate goals of feedback control analysis is to find
the response of a system to any type of input signal
Z ∞
I Consider the Laplace transform F (s) = f (t)e−st dt
−∞
I We saw that Y (s) = H(s)U (s). Thus, if we know the transfer
functions of the input signal and the system, we can find Y (s).
How can we convert it back into time domain?
Inverse Laplace Transform

I Why do we study the frequency response?

I One of the ultimate goals of feedback control analysis is to find
the response of a system to any type of input signal
Z ∞
I Consider the Laplace transform F (s) = f (t)e−st dt
−∞
I We saw that Y (s) = H(s)U (s). Thus, if we know the transfer
functions of the input signal and the system, we can find Y (s).
How can we convert it back into time domain?
I Inverse Laplace Transform y(t) = L−∞ Y(∫ )
Procedure for Finding the Response of a System

I Find the transfer function H(s) = L {impulse response} in two

steps:
I Take the Laplace transform of the equations of motion
I Find the resulting equations by the aid of block diagrams
Procedure for Finding the Response of a System

I Find the transfer function H(s) = L {impulse response} in two

steps:
I Take the Laplace transform of the equations of motion
I Find the resulting equations by the aid of block diagrams
I Find U (s) = L {u(t)}
Procedure for Finding the Response of a System

I Find the transfer function H(s) = L {impulse response} in two

steps:
I Take the Laplace transform of the equations of motion
I Find the resulting equations by the aid of block diagrams
I Find U (s) = L {u(t)}
I Compute Y (s) = H(s)U (s)
Procedure for Finding the Response of a System

I Find the transfer function H(s) = L {impulse response} in two

steps:
I Take the Laplace transform of the equations of motion
I Find the resulting equations by the aid of block diagrams
I Find U (s) = L {u(t)}
I Compute Y (s) = H(s)U (s)
I Partition Y (s) by partial-fraction expansion and get n separate
terms
Procedure for Finding the Response of a System

I Find the transfer function H(s) = L {impulse response} in two

steps:
I Take the Laplace transform of the equations of motion
I Find the resulting equations by the aid of block diagrams
I Find U (s) = L {u(t)}
I Compute Y (s) = H(s)U (s)
I Partition Y (s) by partial-fraction expansion and get n separate
terms
I Take the inverse Laplace transform of the n terms by the aid of
the table and solve for y(t) = L−1 {Y (s)}
Example: Response of a System
u y
ẏ + y = u

Find the response to u(t) = sin(10t)1(t), where


1, if t ≥ 0
1(t) =
0, otherwise
Example: Response of a System
u y
ẏ + y = u

Find the response to u(t) = sin(10t)1(t), where


1, if t ≥ 0
1(t) =
0, otherwise

I Find the transfer function of the system

1
H(s) =
s+1
Example: Response of a System
u y
ẏ + y = u

Find the response to u(t) = sin(10t)1(t), where


1, if t ≥ 0
1(t) =
0, otherwise

I Find the transfer function of the system

1
H(s) =
s+1
I Find U (s) = L {u(t)} (from the table):
10
U (s) = L {sin(10t)} =
s2 + 100
Example: Response of a System
u y
ẏ + y = u

Find the response to u(t) = sin(10t)1(t), where


1, if t ≥ 0
1(t) =
0, otherwise

I Find the transfer function of the system

1
H(s) =
s+1
I Find U (s) = L {u(t)} (from the table):
10
U (s) = L {sin(10t)} =
s2 + 100
I Compute Y (s)
1 10
Y (s) = H(s)U (s) =
s + 1 s2 + 100
Example: Response of a System
u y
ẏ + y = u

1 10
Y (s) = H(s)U (s) =
s + 1 s2 + 100
Example: Response of a System
u y
ẏ + y = u

1 10
Y (s) = H(s)U (s) =
s + 1 s2 + 100

I Partition Y (s) by partial-fraction expansion:

α1 α0 α0∗
Y (s) = + +
s + 1 s + j10 s − j10
10 j −j
101 2(1−j10) 2(1+j10)
= + +
s+1 s + j10 s − j10
Example: Response of a System
u y
ẏ + y = u

1 10
Y (s) = H(s)U (s) =
s + 1 s2 + 100

I Partition Y (s) by partial-fraction expansion:

α1 α0 α0∗
Y (s) = + +
s + 1 s + j10 s − j10
10 j −j
101 2(1−j10) 2(1+j10)
= + +
s+1 s + j10 s − j10
I Take the inverse Laplace transform to solve for y(t):
10 −t 1
y(t) = e +√ sin(10t + ϕ),
101 101
ϕ = tan−1 (−10) = −84.2◦
Example: Response of a System

u y
ẏ + y = u

10 −t 1
u(t) = sin(10t)1(t) → y(t) = e +√ sin(10t + ϕ)
101 101
Example: Response of a System

u y
ẏ + y = u

10 −t 1
u(t) = sin(10t)1(t) → y(t) = e +√ sin(10t + ϕ)
101 101

I The first term is the transient response: It decays to zero as time

goes to infinity
Example: Response of a System

u y
ẏ + y = u

10 −t 1
u(t) = sin(10t)1(t) → y(t) = e +√ sin(10t + ϕ)
101 101

I The first term is the transient response: It decays to zero as time

goes to infinity
I The second term is the steady-state response: It stands for the
response as t → ∞
Example: Response of a System

Draw in MATLAB
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
−∞
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
−∞

1. Superposition
L {α1 f1 (t) + α2 f2 (t)} = α1 F1 (s) + α2 F2 (s)
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
−∞

1. Superposition
L {α1 f1 (t) + α2 f2 (t)} = α1 F1 (s) + α2 F2 (s)
2. Time Delay: Due to linearity, if a function f (t) is delayed by
t1 > 0, its Laplace transform is given by
Z ∞
L {f1 (t − t1 )} = f (t − t1 )e−st dt
0
= e−st1 F (s)
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
−∞

1. Superposition
L {α1 f1 (t) + α2 f2 (t)} = α1 F1 (s) + α2 F2 (s)
2. Time Delay: Due to linearity, if a function f (t) is delayed by
t1 > 0, its Laplace transform is given by
Z ∞
L {f1 (t − t1 )} = f (t − t1 )e−st dt
0
= e−st1 F (s)
3. Multiplication by Time: f1 (t) = tf (t), its Laplace transform is
given by
d
F1 (s) = L {tf (t)} = − F (s)
ds
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0

4. Time Scaling: In some systems, it is desirable to scale time (e.g.,

divide by 1000) to improve precision. In this case,
Z ∞
L {f (at)} = f (at)e−st dt
0
1 s
= F( )
|a| |a|
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0

4. Time Scaling: In some systems, it is desirable to scale time (e.g.,

divide by 1000) to improve precision. In this case,
Z ∞
L {f (at)} = f (at)e−st dt
0
1 s
= F( )
|a| |a|
5. Shift in Frequency: Due to linearity, if a function f (t) is delayed
by t1 > 0, its Laplace transform is given by
Z ∞
L eat f (t) = eat f (t)e−st dt

0
= F (s + a)
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0

6. Differentiation:
  Z ∞
df (t) df (t) −st
L = e dt
dt 0− dt
= −f (0− ) + sF (s)

Generalization:

L {f m (t)} = sm F (s) − sm−1 f (0− ) − . . . − sf (m−2) (0− ) − f (m−1) (0− )

Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0

6. Differentiation:
  Z ∞
df (t) df (t) −st
L = e dt
dt 0− dt
= −f (0− ) + sF (s)

Generalization:

L {f m (t)} = sm F (s) − sm−1 f (0− ) − . . . − sf (m−2) (0− ) − f (m−1) (0− )

7. Integration:
Z t 
1
L f (x)dx = F (s)
0 s
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0
Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0

8. Convolution: Convolution in the time domain corresponds to

multiplication in the frequency domain. Assume
F1 (s) = L {f1 (t)} , F2 (s) = L {f2 (t)}. Then,

L {f1 (t) ∗ f2 (t)} . = F1 (s)F2 (s)

Inverse Laplace:

L−1 {F1 (s)F2 (s)} = f1 (t) ∗ f2 (t)

(Order of multiplication does not matter)

Properties of the Laplace Transform
Z ∞
F (s) = L {f (t)} = f (t)e−st dt
0

8. Convolution: Convolution in the time domain corresponds to

multiplication in the frequency domain. Assume
F1 (s) = L {f1 (t)} , F2 (s) = L {f2 (t)}. Then,

L {f1 (t) ∗ f2 (t)} . = F1 (s)F2 (s)

Inverse Laplace:

L−1 {F1 (s)F2 (s)} = f1 (t) ∗ f2 (t)

(Order of multiplication does not matter)

9. Product:
1
L {f1 (t)f2 (t)} = F1 (s) ∗ F2 (s)
2πj
Inverse Laplace Transform

I Once the transfer function of the output signal for a certain

input, we can convert it into time domain
Inverse Laplace Transform

I Once the transfer function of the output signal for a certain

input, we can convert it into time domain
I Assume the transfer function in the following form

b1 sm + b2 sm−1 + . . . + bm+1
F (s) =
a0 sm + a1 sm−1 + . . . + an
Inverse Laplace Transform

I Once the transfer function of the output signal for a certain

input, we can convert it into time domain
I Assume the transfer function in the following form

b1 sm + b2 sm−1 + . . . + bm+1
F (s) =
a0 sm + a1 sm−1 + . . . + an
I We assume m ≤ n. Let’s rewrite it in a more intuitive form:
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )
Inverse Laplace Transform

I Once the transfer function of the output signal for a certain

input, we can convert it into time domain
I Assume the transfer function in the following form

b1 sm + b2 sm−1 + . . . + bm+1
F (s) =
a0 sm + a1 sm−1 + . . . + an
I We assume m ≤ n. Let’s rewrite it in a more intuitive form:
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )
I This form is called “pole-zero” form.
Inverse Laplace Transform

I Once the transfer function of the output signal for a certain

input, we can convert it into time domain
I Assume the transfer function in the following form

b1 sm + b2 sm−1 + . . . + bm+1
F (s) =
a0 sm + a1 sm−1 + . . . + an
I We assume m ≤ n. Let’s rewrite it in a more intuitive form:
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )
I This form is called “pole-zero” form.
I The term s = zi is called a “zero”, s = pi is called a “pole” of the
system (can be complex). Assume that the poles {pi } are
distinct.
Inverse Laplace Transform
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )

I Rewrite F (s) as the partial fractions:

C1 C2 Cn
F (s) = + + ··· +
s − p1 s − p2 s − pn
Inverse Laplace Transform
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )

I Rewrite F (s) as the partial fractions:

C1 C2 Cn
F (s) = + + ··· +
s − p1 s − p2 s − pn
I Multiply both sides by s − pi :
C1 Ci−1 Ci+1 Cn
(s − pi )F (s) = Ci + + ··· + + + ··· +
s − p1 s − pi−1 s − pi+1 s − pn
Inverse Laplace Transform
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )

I Rewrite F (s) as the partial fractions:

C1 C2 Cn
F (s) = + + ··· +
s − p1 s − p2 s − pn
I Multiply both sides by s − pi :
C1 Ci−1 Ci+1 Cn
(s − pi )F (s) = Ci + + ··· + + + ··· +
s − p1 s − pi−1 s − pi+1 s − pn
I Solve for Ci = (s − pi )F (s)|s=pi
Inverse Laplace Transform
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )

I Rewrite F (s) as the partial fractions:

C1 C2 Cn
F (s) = + + ··· +
s − p1 s − p2 s − pn
I Multiply both sides by s − pi :
C1 Ci−1 Ci+1 Cn
(s − pi )F (s) = Ci + + ··· + + + ··· +
s − p1 s − pi−1 s − pi+1 s − pn
I Solve for Ci = (s − pi )F (s)|s=pi
I Finally, we have the following transformation (Table):
1
Fi (s) = → f (t) = epi t 1(t)
s − pi
Inverse Laplace Transform
Qm
(s − zi )
F (s) = K Qni=1
i=1 (s − pi )

I Rewrite F (s) as the partial fractions:

C1 C2 Cn
F (s) = + + ··· +
s − p1 s − p2 s − pn
I Multiply both sides by s − pi :
C1 Ci−1 Ci+1 Cn
(s − pi )F (s) = Ci + + ··· + + + ··· +
s − p1 s − pi−1 s − pi+1 s − pn
I Solve for Ci = (s − pi )F (s)|s=pi
I Finally, we have the following transformation (Table):
1
Fi (s) = → f (t) = epi t 1(t)
s − pi
n
X
I Thus, y(t) = Ci epi t 1(t)
i=1
Example: Inverse the Laplace Transform
(s + 2)(s + 4)
F (s) =
s(s + 1)(s + 3)

I What are the poles and zeros?

Example: Inverse the Laplace Transform
(s + 2)(s + 4)
F (s) =
s(s + 1)(s + 3)

I What are the poles and zeros?

I Distinct roots. Rewrite F (s) as the partial fractions:
C1 C2 C3
F (s) = + +
s s+1 s+3
Example: Inverse the Laplace Transform
(s + 2)(s + 4)
F (s) =
s(s + 1)(s + 3)

I What are the poles and zeros?

I Distinct roots. Rewrite F (s) as the partial fractions:
C1 C2 C3
F (s) = + +
s s+1 s+3
I Find Ci :
(s + 2)(s + 4) 8
sF (s)|s=0 = |s=0 =
(s + 1)(s + 3) 3
(s + 2)(s + 4) −3
(s + 1)F (s)|s=−1 = |s=0 =
s(s + 3) 2
(s + 2)(s + 4) −1
(s + 3)F (s)|s=−3 = |s=0 =
s(s + 1) 6
Example: Inverse the Laplace Transform
(s + 2)(s + 4)
F (s) =
s(s + 1)(s + 3)

I What are the poles and zeros?

I Distinct roots. Rewrite F (s) as the partial fractions:
C1 C2 C3
F (s) = + +
s s+1 s+3
I Find Ci :
(s + 2)(s + 4) 8
sF (s)|s=0 = |s=0 =
(s + 1)(s + 3) 3
(s + 2)(s + 4) −3
(s + 1)F (s)|s=−1 = |s=0 =
s(s + 3) 2
(s + 2)(s + 4) −1
(s + 3)F (s)|s=−3 = |s=0 =
s(s + 1) 6
I Thus:
8 −3 −t −1 −3t
y(t) = 1(t) + e 1(t) + e 1(t)
3 2 6
Inverse the Laplace Transform

(s + 2)(s + 4)
F (s) =
s(s + 1)(s + 3)

MATLAB code
Final Value Theorem
Qm
(s − zi )
Y (s) = K Qni=1
i=1 (s − pi )
Three cases for the final value of y(t):
1. Unbounded. If any of the poles is in the RHP, i.e., pi > 0 for any
i, then y(t) will grow unbounded.
Final Value Theorem
Qm
(s − zi )
Y (s) = K Qni=1
i=1 (s − pi )
Three cases for the final value of y(t):
1. Unbounded. If any of the poles is in the RHP, i.e., pi > 0 for any
i, then y(t) will grow unbounded.
2. Undefined. If any of the poles is on the imaginary axis, i.e.,
pi = ±jω for any i, then y(t) will contain a sinusoid, and the
final value is undefined.
Final Value Theorem
Qm
(s − zi )
Y (s) = K Qni=1
i=1 (s − pi )
Three cases for the final value of y(t):
1. Unbounded. If any of the poles is in the RHP, i.e., pi > 0 for any
i, then y(t) will grow unbounded.
2. Undefined. If any of the poles is on the imaginary axis, i.e.,
pi = ±jω for any i, then y(t) will contain a sinusoid, and the
final value is undefined.
3. Constant. If all of the poles are on the LHP, i.e., pi < 0 for all i,
except one at the zero, i.e., pi = 0, then y(t) will decay to zero
except the effect of the pole at zero which excites a constant
output.
Final Value Theorem
Qm
(s − zi )
Y (s) = K Qni=1
i=1 (s − pi )
Three cases for the final value of y(t):
1. Unbounded. If any of the poles is in the RHP, i.e., pi > 0 for any
i, then y(t) will grow unbounded.
2. Undefined. If any of the poles is on the imaginary axis, i.e.,
pi = ±jω for any i, then y(t) will contain a sinusoid, and the
final value is undefined.
3. Constant. If all of the poles are on the LHP, i.e., pi < 0 for all i,
except one at the zero, i.e., pi = 0, then y(t) will decay to zero
except the effect of the pole at zero which excites a constant
output.
4. The constant value corresponds to the pole at s = 0.
Theorem: If all poles are on the LHP, then

lim y(t) = lim sY (s)

t→∞ s→∞
Final Value Theorem: Example

3(s + 2)
Y (s) =
s(s2+ 2s + 10)

Find the final value of y(t).

Final Value Theorem: Example

3(s + 2)
Y (s) =
s(s2+ 2s + 10)

Find the final value of y(t).

I

3(s + 2)
lim sY (s) = lim
s→∞ s→∞ (s2 + 2s + 10)
= 0.6
Solving Differential Equations with LT
Recall that for zero I.C.:
...
y + a1 ÿ + a2 ẏ + a3 y = b1 ü + b2 u̇ + b3 u
↓
s Y (s) + a1 s Y (s) + a2 sY (s) + a3 Y (s) = b1 s2 U (s) + b2 sU (s) + b3 U (s)
3 2

Can transform a differential equation into Laplace domain, solve it

with partial fraction expansion, and convert back to the time domain.
Example: Given

ÿ(t) + y(t) = 0, y(0) = α, ẏ(0) = β

Solving Differential Equations with LT
Recall that for zero I.C.:
...
y + a1 ÿ + a2 ẏ + a3 y = b1 ü + b2 u̇ + b3 u
↓
s Y (s) + a1 s Y (s) + a2 sY (s) + a3 Y (s) = b1 s2 U (s) + b2 sU (s) + b3 U (s)
3 2

Can transform a differential equation into Laplace domain, solve it

with partial fraction expansion, and convert back to the time domain.
Example: Given

ÿ(t) + y(t) = 0, y(0) = α, ẏ(0) = β

I Answer:

y(t) = [α cos(t) + β sin(t)] 1(t)

Solving Differential Equations: Exercise

Given

ÿ(t) + 5ẏ(t) + 4y(t) = 3, y(0) = α, ẏ(0) = β

Note that the RHS is not zero (not homogeneous)!

Solving Differential Equations: Exercise

Given

ÿ(t) + 5ẏ(t) + 4y(t) = 3, y(0) = α, ẏ(0) = β

Note that the RHS is not zero (not homogeneous)!

I LT of both sides:
3
s2 Y (s) − αs − β + 5 [sY (s) − α + 4Y (s)] =
s
s(αs − β + 5α) + 3
Y (s) =
s(s + 1)(s + 4)
Solving Differential Equations: Exercise

Given

ÿ(t) + 5ẏ(t) + 4y(t) = 3, y(0) = α, ẏ(0) = β

Note that the RHS is not zero (not homogeneous)!

I LT of both sides:
3
s2 Y (s) − αs − β + 5 [sY (s) − α + 4Y (s)] =
s
s(αs − β + 5α) + 3
Y (s) =
s(s + 1)(s + 4)
I Partial fraction expansion yields:
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
2. p2 = −1 : Stands for the term that decays as e−t
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
2. p2 = −1 : Stands for the term that decays as e−t
3. p3 = −4 : Stands for the term that decays as e−4t (faster)
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
2. p2 = −1 : Stands for the term that decays as e−t
3. p3 = −4 : Stands for the term that decays as e−4t (faster)
4. The effect of p1 , p3 will disappear after the first few seconds
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
2. p2 = −1 : Stands for the term that decays as e−t
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
2. p2 = −1 : Stands for the term that decays as e−t
3. p3 = −4 : Stands for the term that decays as e−4t (faster)
Solving Differential Equations: Exercise

A quick analysis
3 3−β−4α 3−4β−4α
4 3 12
Y (s) = − +
s s+1 s+4
↓
 
3 −3 + β + 4α −t 3 − 4α − 4β −4t
y(t) = + e + e 1(t)
4 3 12

Poles: {0, −1, −4}

1. p1 = 0 : Determines the constant steady-state value
2. p2 = −1 : Stands for the term that decays as e−t
3. p3 = −4 : Stands for the term that decays as e−4t (faster)
4. The effect of p1 , p3 will disappear after the first few seconds
Exercise

Give matlab exercise