P1.

Unit step function uH may be defined in three ways, which allows three interpretations for the delayed
unit step function uv due to a constant delay c :

u H (t )

for t

0, 0.5 or 1

for t

0 , uv ( t )

for t

u H (t c )

0, 0.5 or 1

for t c 0

for t c 0

for t c

uV ( c ) . Also replace 0 by 0 .

But here we do not need to specify uH ( 0)

Then the Laplace Transform U v (s ) of uv is by definition

U v ( s)

L uv

uv ( t ) e

st

dt

The function is presented separately for two successive time intervals.

0
Split therefore the integration interval into two subintervals:
c

U v ( s)

st

u v (t ) e

dt

uv ( t ) e

st

dt

Substitute next the appropriate presentations for both subintervals:

c

U v ( s)

0 e

st

dt

1 e

st

dt

c
0

0 dt

st

dt

st

dt

Find the integral (antiderivative) function of the integrand (the function to be integrated):
st

U v ( s)

if

st

or with another notation U v ( s )

if

if

if

c
c

0t

U v ( s)

0 (yielding e

e0

1 ) the result is unbounded, which means that no transform exists for

0 , i.e. in the origin of the complex plane. But for s 0 we obtain

For s

lim
t

exp( s t )
s

exp( s c )
s

e st for t
. According to the
0
Re s 0 , i.e. iff
James and lecture notes the limit exists and is bounded if and only if Re s
(if and only if) Re s
0 . For this condition the value of the limit is 0 (zero). Therefore the Laplace
The existence depends on the asymptotic behavior of exp( s t )

transform requested is

U v ( s)

exp( c s )
s

cs

1
s

for Re s

but does not exist otherwise.

For c 0 the time domain function reduces to the ordinary unit step function and the transform reduces to
1 / s , i.e. to that of the ordinary unit step function.
E05-1

P2. The unit pulse function with width W is defined as

u pf (t )

1/W
0

for 0 t W
for
t W

1/W
t
W

Here the area of the shaded rectangle between the horizontal axis and the unit pulse function is equal to
(1 / W ) W 1 (one, unity) for any positive width W of the rectangular pulse. In probability science and
statistics this function is the probability density function of an even distribution on the interval 0...W .
BTW, rand of Matlab uses W 1 . The integral of u pf over the whole range is 1 , the appropriate value for
the final value of a cumulation function of a distribution, the ultimate or maximum possible probability.
The Laplace transform requested is by definition, since we may again replace 0

U pf ( s)

L u pf

by 0 ,

u pf (t ) exp( s t ) dt

Splitting this according to the subintervals yields

U pf ( s )
U pf ( s )
For s
For s

W
0

(1 / W ) exp( s t ) dt

(1 / W )

W
0

0 exp( s t ) dt

exp( s t ) dt

(A)

0 the function exp( s t ) is equal to 1 , which gives U pf ( s )

0 we have:

U pf ( s )

1
W

exp( s t )
s

1
W

exp( s W )
s

(1 / W ) W

1 (the area!).

exp( 0 W )
s

U pf ( s )

1 exp( s W )
s W

With s
0 this transform derived for s 0 will approach the value 1 . In fact, this occurs from any
direction of the environment of the origin which is necessary for the limit to exist. The unity value was obtained
above also in the case with s 0 . These observations show that is U pf (s ) continuous even in the origin.
To summarize, the Laplace Transform U pf (s ) of the Unit Pulse Function is given by

U pf ( s )

1 exp( s W )
for s
s W

and

U pf (0) 1

Here the formula for s 0 will approach the value for s 0 as s

0 . This justifies the usual
practice to use the transform formula for s 0 even for s 0 without too many further comments.
Important. The case W
0 is interesting: the pulse area is still equal to 1 , and the transform will
approach the value 1 . This occurs since the integral in Equation (A) approaches the value W since

st

1 on the integration interval. In the case W

0 the function u pf is called the unit impulse

function, which seems to have a unity Laplace Transform.

Plot the function integrated for a real positive value of s and interprete the transform as an area
E05-2

P3. Poles of a Laplace transform (or any other function of a complex variable) are points (of the complex
plane) such that the absolute value of the transform grows without any finite bound when the free
complex variable approaches the point tightly enough. Here each transform is a rational function with a
finite order denominator polynomial and a numerator polynomial the degree of which does not exceed
that of the denominator polynomial. Therefore the pole condition is satisfied just iff (if and only if)
division with zero is approached when Assuming that cancellations can be excluded thanks to the
ability to select another row of the table when a cancellation threatens the poles will be precisely the
points for which the limit of the denominator polynomial is zero. So the only possible poles here are the
roots of the denominator polynomial. The denominator polynomial is also called the charpo or the
characteristic polynomial.
The answers are collected to the table below assuming that a and b are real numbers. In some cases a
condition of the form b 0 is assumed not to have a reduction the consequence of which may be dealt (as was
proposed) in another row of the table or is trivial. Two examples will be explained in more details. Note that at a
pole the transform expression is in fact not the transform of the time domain function because the Laplace
integral is not convergent for that point:

f (t )
1
t
t2
exp( a t )
sin( b t ) , b

cos(b t ) , b

exp( a t ) sin( b t ) , b
exp( a t ) cos(b t ) , b
exp( a t ) t

F ( s)
1
s
1
s2
2
s3
1
s a
b
2
s b2
s
2
s b2
b
0
( s a) 2 b 2
s a
0
( s a) 2 b 2
1
( s a )2

poles

Re parts Im parts

0,0

0,0

0,0

0,0,0

0,0,0

0,0,0

jb

jb

jb

jb

a,a

a,a

0,0

For example, a time domain function f defined with

f (t )

exp( at ) cos(bt ) , b

0, t

has a irreducible transform F (s ) the charpo of which is

( s)

(s a) 2

b2

For b 0 both f and F could be simplified so that another row of the table becomes actual. The roots of
are the solutions of the equation

( s a) 2

b2

This so called characteristic equation may be replaced by

( s a) 2

b2

( j) 2 b 2

( j b) 2
E05-3

s a

j b

Hence the so called characteristic roots are

j b

Assuming that a and b are real numbers we see, that the two roots of charpo, here the poles of the
transform, are complex conjugates to each other.
Another time domain function f defined with

f (t )

t ,t

0
2

seems to have the characteristic polynomial s . Its characteristic equation is

s2

s s

This is satisfied iff

or

(first s )

(second s )

So we have the poles 0 and 0 , i.e. two poles at s

0 or, as we often say, a double pole in the origin.

The figure given shows the plots for

f1 (t ) exp( at ) cos(bt )

(curve rows 1-2)

f 2 (t )

(curve row 3)

exp( at )

for a few parametric cases. The shape of a missing function represented by

f 3 (t ) exp( at ) sin( bt )
is that of f1 but with a suitable time shift so that the function starts with zero value at zero time.

Note that the real part of the pole multiplies the time inside exp function,
while the imaginary coordinate of the pole multiplies the time inside sin and/or cos function!

A second order twin-pole-function not illustrated

A unilateral function f 4 defined with

f 4 (t )

t exp( a t ) , t

has a transform which has the poles a and a , i.e. a double pole at s a . Assuming that the pole is real
we can illustrate the plots of f 4 as below where a special value of a is adopted vs. a normalized time is used:

E05-4

Impulse Response

Impulse Response

Impulse Response

0.4

1.4

35

0.35

1.2

30

25

0.8

20

0.3

0.2
0.15
0.1
0.05
0

10

0.6

15

0.4

10

0.2

15

Amplitude

Amplitude

Amplitude

0.25

0.5

0.5

1.5

Time (sec)

Time (sec)

a<0

a=0

a>0

lim f 4 (t )

lim f 4 (t )

lim f 4 (t )
t

2.5

Time (sec)

Find answers where you refer to features of poles:

a) When will the functions discussed converge to a finite value, perhaps even to zero?
b) When a function discussed is not oscillating?
c) When is an oscillating function discussed above
convergent
non-convergent (divergent) & bounded
non-convergent & unbounded ?
d) How is the exponential decay changed if the pole is moved leftwards?
e)

How is the temporary oscillation changed if the pole is moved leftwards?

f)

How is the temporary oscillation changed if the pole is moved upwards?

An excellent source for further Laplace studies (also available as an ebook in TUT library net service):
James: Advanced Modern Engineering Mathematics, Pearson.

P4.
b) u (t )

a cos(

t ) b sin(

t) ,

0 , with real parameters

Both sine and cosine have a Laplace Transform for Re s

U ( s)

L a cos(

t ) b sin(

t)

, Re s

0 . Assuming this lets go:

Sum Rule gives

U ( s)

L a cos(

t)

L b sin(

t)

Re s

,
E05-5

Gain Rule implies for constants a , b

U ( s)

a L cos(

t)

b L sin(

t)

Re s

Pick the transforms of sine and cosine from the appropriate table of Laplace Transforms,

U ( s) a

s
s

s2

Re s

Re s

Collect this into a standard ratio of two polynomials:

a s b
2
s2

U (s )

a s b
1 s2

Here
is assumed to be real. Here numerator degree does not exceed denominator degree. Hence each
pole must here belong to the set of denominator roots,

s2

s2

But a denominator root is a pole iff it is not also a numerator root. So lets study carefully:

0 the numerator has a root given by ( b

) / a . Then we have the poles

case where this root is also a den root which occurs iff
0 . This means

1) For a

U ( s)

a s
s2

except in the

1
s

with a single pole at the origin and making u the step function of amplitude a .

0 the numerator has no roots. Then we have the poles j except in the pathological case
where b
0 which occurs for both b 0 and
0 both of which actually imply U (s) 0
with no poles and making u to be the zero function.

2) For a

j on the Im axis while the other possibilities are associated with some
So for true oscillation the poles are
degenerate cases where the function is not really oscillating but has almost everywhere zero derivative. Now
lets have some repetition for improved learning and linking topics: Let the cos and sin functions be xcos and
xsin , respectively. Our function u can be generated using a function generator described below. Derive an
elementary block diagram as well as a state space model with matrix objects, and study observability of the
model to find something concerning reduction of the time domain function and its Laplace Transform:

u( t )

a xcos (t ) b xsin (t ) , xcos (t )

xsin (t ) , xsin (t )

xcos (t ) , xcos (0) 1 , xsin (0)

P2 Revisited. The time domain function studied in P2 may be rewritten in terms of Heaviside unit step
function u H as in Equation (1) below. Sum Rule and Gain Rule (or the Linearity Rule joining them) can then
be used to get the Laplace Transform. But this approach assumes Re s
0 . However, it is not a
necessary assumption in the derivation of P2. This illustrates that sometimes a procedure used introduces
unnecessary assumptions.

u pf (t )

1
1
u H (t )
u H (t W )
W
W

(1)

E05-6

P5. Let

F ( s ) U 2pf ( s ) . For W

lim F ( s)
W

lim U pf ( s) U pf ( s)

Hence

F (s)

0 this has the limit

lim U pf ( s) lim U pf ( s ) 1 1 1

f (t ) to be found may also serve as a Dirac impulse by letting W

1 exp( Ws )
Ws

( 1 exp( Ws ) ) 2
(Ws ) 2

F (s)

1 2 exp( Ws) exp( 2Ws)

W 2s 2

F (s)

1 1
W 2 s2

2
W

exp( Ws)

1
s2

12

0 . Lets derive f (t ) :

2 exp( Ws ) exp 2 ( Ws )
W 2s2

1
1
exp( 2Ws) 2
2
W
s

Using the Linearity Rule of Inverse Laplace Transform, the Laplace Transform of the Unit Ramp Function
reversely and the Delay Rule reversely (for the Unit Ramp Function) we obtain

f (t )

0 , t 0
t , t 0

1
W2

2
W2

0
t W

, t W
, t W

1
W2

0 0 0
2

f (t )

W
W

W t 0 0
t 2W 2 (t W ) 0

t 2W
0

f (t )

W
2W

(t W ) W
,

t
W

, t 2W
, t 2W

, 0 t W
, W t 2W

(t 2W ) ,

2W

t 0

, 0 t W
t , W t 2W
,

t 2W

We obtain a triangle symmetric with respect to t

Here

0
t 2W

W , with the height equal to 1 / W and with unity area

u pf can serve as the PDF (Probability Density Function) of a scalar random variable distributed evenly on

f could represent the PDF of x1 x 2 if x1 and x2 are mutually independent

random variables but have the same PDF u pf . Illustrate this with rand (which applies W 1 ), hist and +.
the range 0...W while

Solve

f also using syms, laplace, * or ^ , ilaplace, ezplot. Make the same also for U pf4 ( s) to see a curve

shape close to the famous Gaussian bell curve. In fact, each PDF with zero mean and extremely small positive
variance is Dirac.
E05-7

P6. Inertia J

0 , coefficient of viscous damping b

(t ) 0

(t ) b

L J

(t ) b

L J

(t )

J L (t )
s2 L

(J s2
L

equivalence of the functions

(t )

Lb

0.

L0

equivalence of the LTs

(t )

Sum Rule

b L (t )

Gain Rule

(t )

b s) L

(t )

(0 )

( 0)

(t )

b s L

( 0)

Derivative Rules

J (0) s ( J (0) b (0) )

J ( 0) s ( J ( 0) b ( 0 ) )
J s2 b s

(t )

LT of the natural response

Denominator roots 0 , ( J b) , here distinct. Use them to factor the denominator polynomial:

J ( 0) s ( J ( 0) b ( 0 ) )
J ( s 0) ( s ( J 1b) )

(t )

Partial Fraction Expansion due to Distinct Roots:

J ( 0) 0 ( J ( 0) b ( 0) )
J ( s 0) (0 ( J 1b) )

(t )

J (0) ( J 1b) ( J (0) b (0) )

J (( J 1b) 0) ( s ( J 1b) )

Represent it as a Linear Combination of a few elementary LTs available in Exam Tables:

J ( 0) b ( 0) 1
b
s

(t )

J ( 0)
1
b
s ( J 1b)

Exploit the LT formula of Step Function and exp function and keep the Linear Combination structure:

(t )

J (0) b (0)
J (0) (
1
e
b
b

J 1b ) t

, t

This has a finite final value,

lim (t )
t

J (0) b (0)
b

Check: Solve first

(t )

(t )

(t ) using LT, then use

(t )

(0)

t
0

( )d

E05-8