Professional Documents
Culture Documents
Laplace
Laplace
The definition of the Laplace Transform that we will use is called a "one-sided"
(or unilateral) Laplace Transform and is given by:
For our purposes the time variable, t, and time domain functions will
always be real-valued. The Laplace variable, s, and Laplace domain
functions are complex.
Since the integral goes from 0 to ∞, the time variable, t, must not occur
in the Laplace domain result (if it does, you made a mistake). Note that
none of the Laplace Transforms in the table have the time variable, t,
in them.
The lower limit on the integral is written as 0 -. This indicates that the
lower limit of the integral is from just before t=0 (t=0 - indicates an
infinitesimally small time before zero). This is a fine point, but you will
see that it is very important in two respects:
o It lets us deal with the impulse function, δ(t). If you don't know
anything about the impulse function yet, don't worry, we'll
discuss it in some detail later.
o It lets us consider the initial conditions of a system at t=0 -
. These are often much simpler to find than the initial conditions
at t=0+ (which are needed by some other techniques used to
solve differential equations).
Since the lower limit is zero, we will only be interested in the behavior
of functions (and systems) for t≥0.
You will sometimes see discussed the "two-sided" (or bilateral)
transform (with the lower limit written as -∞) or a one-sided transform
with the lower limit written as 0+. We will not use these forms and will
not discuss them further.
Since the upper limit of the integral is ∞, we must ask ourselves if the
Laplace Transform, F(s), even exists. It turns out that the transform
exists as long as f(t) doesn't grow faster than an exponential
function. This includes all functions of interest to us, so we will not
concern ourselves with existence.
Before we show how the Laplace Transform is useful, we need to lay some
groundwork. We start by finding the Laplace Transform of some functions and
from there move on to finding properties of the Laplace Transform. With tables of
the Laplace Transform of Functions and Properties of the Laplace Transform it
becomes possible to find the Laplace Transform of almost any function of interest
without resorting to the integral shown above. Applications of the Laplace
Transform are discussed next - mostly the use of the Laplace Transform to solve
differential equations.
Laplace Transform of Functions
Contents
but this quickly becomes tedious. Our goal is to avoid having to evaluate the
integral by finding the Laplace Transform of many useful functions and compiling
them in a table. Thereafter the Laplace Transform of functions can almost always
be looked by using the tables without any need to integrate. A table of Laplace
Transform of functions is available See Apedix
The Unit Step Function
The unit step function is defined as
Most references use u(t) instead of γ(t). However, u(t) has some other common uses, so we will
use γ(t) to avoid confusion (and because it's Laplace Transform Γ(s) looks a little like a step
input).
Some references change the definition of the step so that the bottom inequality becomes a strict
inequality (t>0) and either leave γ(0) undefined or say that γ(0)=�.
Although there is a discontinuity at t=0, we draw a vertical line to help guide the eye and to
indicate that the blue line is a single function.
so
Now we apply the sifting property of the impulse. Since the impulse is 0
everywhere but t=0, we can change the upper limit of the integral to 0 +.
Since e-st is continuous at t=0, that is the same as saying it is constant from t=0 - to
t=0+. So we can replace e-st by its value evaluated at t=0.
So the Laplace Transform of the unit impulse is just one. Therefore the impulse
function, which is difficult to handle in the time domain, becomes easy to handle in
the Laplace domain. It will turn out that the unit impulse will be important to much
of what we do.
The Exponential
Consider the causal (i.e., defined only for t>0) exponential:
The function is 0 for t<0, and then follows an exponential trajectory thereafter.
We can either define the function piecewise (the first definition), or as an
exponential multiplied by the unit step (the second definition). The second one is
more compact, so we will generally use that one. Because all of our functions are
equal to zero for t<0, the multiplication by γ(t) will often be implicit and we won't
actually show γ(t), except in cases where it is needed for clarity.
If a<0, the function increases without bound. If a>0 the function decays to zero -
decaying exponentials are much more common in the systems that we study.
Since γ(t) is equal to one for all positive t, we can remove it from the integral
(note)
Note that multiplication by the function γ(t) is implicit on the left hand side of the
last equation. This is how we will commonly write our functions.
Note that if a=0, we get a step function and Y(s)=1/s.
The Sine
so
But we've already done this integral (the exponential function, above)
The Cosine
The cosine can be found in much the same way, but using Euler's identity for
the cosine.
Note that if ω=0, we get a step function and Y(s)=1/s.
The rest of the derivation follows that of the sine function (i.e., put over a
common denominator, and solve)
We start as before
Let's see if we can use this information to find the Laplace Transform of the
rectangular pulse. We can form the original rectangular pulse function from the step
and the delayed step in two ways
Method 1 Method 2
The first method doesn't help us, because we have no property of the Laplace
Transform the lets us deal with multiplied functions in the time domain. However, the
second method can be used because it represents y(t) as the sum of functions, so we
can use the linearity property. So we get
The addition of the two functions is shown graphically below
and y(t)=2.5·z(t)
Example: Laplace Transform of a Triangular Pulse
Find the Laplace Transform of the function shown:
Solution:
We need to figure out how to represent the function as the sum of functions with which
we are familiar. For this function, we need only ramps and steps; we apply a ramp
function at each change in slope of y(t), and apply a step at each discontinuity.
Functionally we want:
so
Graphically this is shown as:
and this does, in fact, yield the correct function. However if we do this we end up with
the product of two functions that are in the table separately
However, these two functions have different time delays and we have no way to deal
with products of functions. That is why, when choosing the basic functions that make
up the composite function, only addition is allowed. (note)
Solution:
This function is more complicated than the last, but we can still create it as a sum of
ramps and steps; we apply a ramp function at each change in slope of y(t), and apply
a step at each discontinuity.
The constituent functions are shown in the plot below. If you add them up, you get
the original function (shown in black).
Solution:
This function is a little different than the previous in that it involves more than ramps
and steps. The most obvious way to represent this function is as a sine wave multiplied
by a rectangular pulse. But recall that we can't use generally use multiplication of
functions (we have no multiplication property), only addition (by use of the linearity
property). So the question becomes: what functions can we add to create the given
function?
After a little thought it becomes apparent that we can take a sine wave starting at
t=0, and subtract off a cosine beginning at t=2.5.
Appendix:
!.
ramp
parabola
tn
(n is
integer)
exponen
tial
power
time
multiplie
d
exponen
tial
Asympto
tic
exponen
tial
double
exponen
tial
asympto
tic
double
exponen
tial
asympto
tic
critically
damped
different
iated
critically
damped
sine
cosine
decaying
sine
decaying
cosine
generic
decaying
oscillator
y
generic
decaying
oscillator
y
(alternat
e)
(note)
Z-
domain
generic
decaying
oscillator
y
(note)
Prototype Second Order System (ζ<1, underdampded)
Prototyp
e
2nd order
lowpass
step
response
Prototyp
e
2nd order
lowpass
impulse
response
Prototyp
e
2nd order
bandpas
s
impulse
response
Since t=kT, simply replace k in the function definition by k=t/T. So, in this case,
Definition
Linearity
First Derivative
Second Derivative
nth Derivative
Integration
Multiplication by time
Time Shift
Complex Shift
Time Scaling
Convolution
('*' denotes convolution
of functions)
Initial Value Theorem
(if F(s) is a strictly proper
fraction)
Final Value Theorem
(if final value exists,
e.g., decaying exponentials )
Laplace Transform Properties
Contents
Linearity
Time Delay
First Derivative
Second Derivative
Nth Order Derivative
Integration
Convolution
Initial Value Theorem
Final Value Theorem
Other properties
Linearity
The linearity property of the Laplace Transform states:
Time Delay
The time delay property is not much harder to prove, but there are some
subtleties involved in understanding how to apply it. We'll start with the statement
of the property, followed by the proof, and then followed by some examples. The
time shift property states
We again prove by going back to the original definition of the Laplace Transform
Because
we can change the lower limit of the integral from 0- to a- and drop the step function
(because it is always equal to one)
The last integral is just the definition of the Laplace Transform, so we have the time
delay property
To properly apply the time delay property it is important that both the function
and the step that multiplies it are both shifted by the same amount. As an example,
consider the function f(t)=t·γ(t). If we delay by 2 seconds it we get (t-2)·γ(t-2), not
(t-2)t·γ(t) or t·γ(t-2). All four of these function are shown below.
The correct one is exactly like the original function but shifted.
Important: To apply the time delay property you must multiply a delayed
version of your function by a delayed step. If the original function is g(t)·γ(t),
then the shifted function is g(t-td)·γ(t-td) where td is the time delay.
First Derivative
The first derivative property of the Laplace Transform states
To prove this we start with the definition of the Laplace Transform and integrate by
parts
The first term in the brackets goes to zero (as long as f(t) doesn't grow faster than
an exponential which was a condition for existence of the transform). In the next
term, the exponential goes to one. The last term is simply the definition of the
Laplace Transform multiplied by s. So the theorem is proved.
We have taken a derivative in the time domain, and turned it into an algebraic
equation in the Laplace domain. This means that we can take differential
equations in time, and turn them into algebraic equations in the Laplace
domain. We can solve the algebraic equations, and then convert back into the
time domain (this is called the Inverse Laplace Transform, and is described later).
The initial conditions are taken at t=0-. This means that we only need to know the
initial conditions before our input starts. This is often much easier than finding
them at t=0+.
Second Derivative
Similarly for the second derivative we can show:
where
or
where
Integration
The integration theorem states that
Example: Find Laplace Transform of Step and Ramp using Integration Property
Given that the Laplace Transform of the impulse δ(t) is Δ(s)=1, find the Laplace
Transform of the step and ramp.
Solution:
We know that
so that
Likewise:
Convolution
The convolution theorem states (if you haven't studied convolution, you can skip
this theorem)
We then invoke the definition of the Laplace Transform, and split the integral into
two parts:
We take the limit as s→∞:
Several simplifications are in order. In the left hand expression, we can take the
second term out of the limit, since it doesn't depend on 's.' In the right hand
expression, we can take the first term out of the limit for the same reason, and if
we substitute infinity for 's' in the second term, the exponential term goes to zero:
The two f(0-) terms cancel each other, and we are left with the Initial Value
Theorem
This theorem only works if F(s) is a strictly proper fraction in which the numerator
polynomial is of lower order then the denominator polynomial. In other words is will
work for F(s)=1/(s+1) but not F(s)=s/(s+1).
However, we can only use the final value if the value exists (function like sine,
cosine and the ramp function don't have final values). To prove the final value
theorem, we start as we did for the initial value theorem, with the Laplace
Transform of the derivative,
We let s→0,
As s→0 the exponential term disappears from the integral. Also, we can take f(0-
) out of the limit (since it doesn't depend on s)
We can evaluate the integral
Neither term on the left depends on s, so we can remove the limit and simplify,
resulting in the final value theorem
Examples of functions for which this theorem can't be used are increasing
exponentials (like eat where a is a positive number) that go to infinity as t increases,
and oscillating functions like sine and cosine that don't have a final value..
Other properties
Some other properties that are important but not derived here are listed below.
Multiplication by time:
Complex Shift:
Time Scaling:
Laplace Transform Applied to Differential Equations and
Convolution
Contents
Differential Equations
Convolution
The Laplace Transform has many applications. Two of the most important are
the solution of differential equations and convolution. These are discussed below.
Differential Equations
The Laplace Transform can greatly simplify the solution of problems involving
differential equations.
Solving a first order differential equation
Consider the differential equation given by:
Find y(t).
Aside: Origin of the First Order Differential Equation
The differential equation with input f(t) and output y(t)
can represent many different systems. Two examples are given below, one for a
mechanical system and one for an electrical system.
Mechanical System
Input = force = φ(t). Output = velocity of mass. b=2, m=1.
Free Body Diagram
You can see that this is equivalent to the original equation (with output v(t) and
input φ(t)).
Electrical System
Input = current = ia(t). Output = voltage across elements. R=½, C=1.
This is equivalent to the original equation (with output eo(t) and input ia(t)).
Solution:
The solution is accomplished in four steps:
4. Get result from the Laplace Transform tables. (look up the terms
individually)
In this case all terms are in the table and we need to apply the time
delay property. If the terms are not in the table, we need to apply
the methods of the Inverse Laplace Transform. Note that the
function is implicitly zero for t<0. We could make this explicit by
multiplying by the unit step:
Find y(t).
Aside: Origin of the Second Order Differential Equation
Note: we could get such a second order differential equation by adding a spring
(k=10) between the mass and a fixed support, or an induction (L=1/10) in parallel to
the capacitor and inductor.
Solution:
Again, the solution can be accomplished in four steps.
Convolution
The convolution integral is very important in the study of systems. In short,
convolution can be used to calculate the zero state response (i.e., the response to
an input when the system has zero initial conditions) of a system to an arbitrary
input by using the impulse response of a system. Given a system impulse
response, h(t), and the input, f(t), the output, y(t) is the convolution of h(t) and f(t):
However, this integral can be quite hard to calculate in this form, but is quite easy
if using the Laplace Transform.
Example: Convolution in the Laplace Domain
Find y(t) given:
you can see that the Laplace domain solution is much easier.
Solution:
To evaluate the convolution integral we will use the convolution property of the
Laplace Transform:
[Math Processing Error]y(t)=f(t)∗h(t)⟷LY(s)=F(s)H(s)
We need the Laplace Transforms of f(t) and h(t), but we can look them up in
the tables:
[Math Processing Error]F(s)=1s−e−2s1s [Math Processing Error]H(s)=1s+1
So
[Math Processing Error]Y(s)=F(s)H(s)=(1s−e−2s1s)(1s+1)=1s(s+1)−e−2s1s(s+1)
We can look up both of these terms in the tables.
[Math Processing Error]1s(s+1)⟷L(1−e−t)γ(t)e−2s1s(s+1)⟷L(1−e−(t−2))γ(t−2)
We can now write y(t) (which is implicitly zero for t<0, so I will drop the γ(t) term)
[Math Processing
Error]y(t)=(1−e−t)−(1−e−(t−2))γ(t−2)