Circuit Analysis
Jason Creighton
May 13, 2011
Abstract
This paper presents an overview of the Laplace transform along with
its application to several well-known electrical circuits. There is a focus
on systems which other analytical methods have difficulty solving, such
as those containing discontinuous forcing terms. Each analytical solution
was tested empirically against the actual behavior of the circuit.
Introduction
Before we examine the Laplace transform, lets first discuss notation and then
a problem related to the Laplace transform.
1.1
Notation
instead of i to avoid confusion with current. Also, real and imaginary parts
of a complex number will be denoted as:
Re(a + bj) = a
Im(a + bj) = b
1.2
a0 = 7;
a1 = 26
In other words, for every an in the sequence the above equation must hold,
and the first two numbers in the series are 7 and 26. How could we go about
finding what the rest of the numbers in this sequence are? One way would be
to re-arrange the above equation to find an in terms of the previous two entries
in the sequence:
an = 7an1 10an2
Since we already know a0 and a1 , we can start calculating the sequence:
a2 = 7(26) 10(7)
= 112
a3 = 7(112) 10(26)
= 524
a4 = 7(524) 10(112)
= 2548
a5 = 7(2548) 10(524)
= 12596
an xn
n=0
This seems like a strange thing to define, but first convince yourself that this
function has meaning. If we add up every entry in the (infinite!) sequence an ,
multiplying each entry by xn , then for some values of x, the summation will
converge and A(x) will be defined.
Lets try to get our recurrence relation into a form resembling A(x). We can
multiply each term by xn+2 :
an+2 xn+2 7an+1 xn+2 + 10an xn+2 = 0
We can sum both sides and factor some powers of x where appropriate:
an+2 xn+2 7xan+1 xn+1 + 10x2 an xn = 0
n=0
Next we rewrite this as three summations and subtract terms outside the
summations to allow us to write the term being summed as an xn :
an x a1 x a0 7x
n=0
!
n
an x a0
n=0
Since A(x) =
n=0
+ 10x2
an xn = 0
n=0
7 23x
10x2 7x + 1
Using partial fractions decomposition, we find that:
A(x) =
A(x) =
3
4
+
1 2x 1 5x
rn =
n=0
1
1r
(2x)n + 4
(5x)n
n=0
n=0
Simplifying:
A(x) =
[3(2n ) + 4(5n )] xn
n=0
PWhat nis the point of all this? Remember we originally defined A(x) =
n=0 an x . Comparing that to the above equation, its clear that:
an = 3(2n ) + 4(5n )
By defining a function A(x) as a power series, we were able to find an explicit
solution for a recurrence relation. Notice that we never actually evaluated A(x)
for any value of x; we merely used its definition to reach our solution.
1.3
It might be useful to have a continuous version of the power series. The continuous analogue of summation is integration, and we will write f (t) instead of
an :
Z
F (x) =
f (t)xt dt
0
Now, this is an improper integral, and since we dont know what f (t) is,
we cant say for sure when well be able to evaluate it. But if x > 1, for most
functions, the integral will not converge. And we want to avoid x < 0 to keep
from having to deal with imaginary numbers. So in most cases well only be
able to evaluate this when 0 < x < 1.
When integrating, it is generally easier to have e as the base of your exponential functions. So lets substitute: x = eln(x) .
f (t)eln(x)t dt
F (x) =
0
And since our variable x is now somewhat awkward, lets introduce a substitution s = ln(x):
Z
F (s) =
f (t)est dt
0
= aL {f (t)} + bL {g(t)}
2.1
Lets start with a very simple example, f (t) = 1. By the definition of the
Laplace transform,
Z
L {1} = F (s) =
est (1) dt
0
est dt
F (s) = lim
i
1 h sT
e
es(0)
s
1
,s > 0
s
Notice that a constant function in the t-domain is not a constant function
in the s-domain.
L {1} =
2.2
Suppose we want the Laplace transform of f 0 (t), but we dont know f (t). Can
we find the transform in terms of L{f (t)}?
Z
0
L{f (t)} =
est f 0 (t) dt
0
2.3
Then:
1
F (s)
s
So with the Laplace transform, derivatives and integrals turn into algebraic
factors.
G(s) =
2.4
1
s
1
s2
n!
n+1
s
1
sa
a
2
s + a2
s
s2 + a2
2.5
A Simple Example
y(0) = 7;
y 0 (0) = 26
7s + 23
7s + 23
=
s2 7s + 10
(s 2)(s 5)
6
3
4
+
s2 s5
We recognize the form of the Laplace transform of eat , and taking advantage
of the linearity of the transform, we have:
y(t) = 3e2t + 4e5t
Notice the similarity between the solution of this problem and the solution
of the recurrence relation. In particular, if you were to substitute x = 1/s into
A(x), you would get sY (s).
2.6
In order to have discontinuous forcing terms, we can make use of the so-called
Heaviside function H(t):
0, t < 0
H(t) =
1, t 0
In particular, it is useful to note that:
L{H(t c)f (t c)} = ecs F (s)
And of course if follows that:
L1 ecs F (s) = H(t c)f (t c)
In essence, this says that shifting a function in the t-domain multiplies the
function some exponential factor in the s-domain.
2.7
(t) =
, t = 0
0, t 6= 0
The physical analogue of the delta function might be a sharp blow from
a hammer to a mass-spring system, which imparts a nearly instantaneous but
finite impulse.
Notice that the derivative of the step function is the delta function:
d
H(t) = (t)
dt
2.8
Periodic Functions
Lets say that we want to want to find the Laplace transform of a function that
repeats periodically. We say that the window of a function f with a period
of T is:
f (t),
0t<T
w(t) =
0,
T t < , t < 0
Notice that f (t) can be written as a sum of translated window functions:
f (t) = w(t) + w(t T ) + w(t 2T ) + =
w(t kT )
k=0
H(t kT )w(t kT )
k=0
k=0
eT s
k
k=0
esT
k
k=0
So finally:
1
1 eT s
L{f (t)}(s) =
W (s)
1 eT s
3.1
Convolutions
Circuits
4.1
1
q;
C
q(0) = 0
For our example, lets say that e(t) is a periodic square wave of period 2
and amplitude A. A square wave is exactly the kind of function that analytical
approaches tend to have difficulty with.
Taking the Laplace transform, we find:
E(s) = R[sQ(s) q(0)] +
1
Q(s)
C
1
Rs+
1
C
E(s)
We cant go much further without finding E(s). Recall that to find the
Laplace transform of e(t), we need to find the Laplace transform of its window,
which well call w(t):
w(t) = A[H(t) H(t 1)]
where A is the amplitude of the square wave. The Laplace transform is:
1 es
1 es
=A
W (s) = A
s
s
s
Now, we found earlier that if W (s) is the Laplace transform of a window of
some function with period T , then the Laplace transform of that function is:
E(s) =
W (s)
1 es
=
A
1 eT s
s(1 e2s )
(1 es )
1
=A
s
s
s(1 e )(1 + e )
s(1 + es )
1
Rs+
Q(s) = A
1
s(R s +
1
C
1
s(1 + es )
Rearranging terms:
1
C)
1
1 + es
1
C)
=C
10
1
1
1
s s + RC
X
X
1
1
s n
=
=
e
=
(1)n ens
1 + es
1 (es ) n=0
n=0
So:
1
1
Q(s) = AC
1
s s + RC
X
(1)n ens
n=0
(1)n ens
n=0
1
1
1
s s + RC
= 1 e RC
L
1
s s + RC
Bringing it all together, we have:
q(t) = AC
t
(1)n H(t n) 1 e RC
n=0
R = 954.15 k
C = 0.2258 F
A = 10 V
Figure 1 shows the result.
11
Expected
Actual
10
0.5
1.5
2.5
3.5
Time (s)
4.2
The Laplace transform can also be used to find the transfer function of a
circuit. The transfer function is the ratio between the output voltage and the
input voltage of a circuit. Using the RC circuit above as an example, let vi (t) =
e(t) and vo (t) = C1 q(t). Then:
T (s) =
Q(s) C1
Vo (s)
=
Vi (s)
Q(s) Rs +
1
C
=
1
RCs + 1
R = 551.4
C = 0.2262 F
Figures 2 and 3 show the amplitude and phase, respectively, of the circuit we
analyzed here. Figures 4 and 5 show the amplitude and phase of a similar circuit
in a highpass configuration, meaning that the output voltage is measured
across the resistor, not the capacitor.
4.3
Consider a series RLC circuit. We can use the same sort of analysis we used on
the RC circuit to determine its frequency response.
vi (t) = R i(t) + L i0 (t) +
1
q(t);
C
q(0) = 0;
i(0) = 0;
where q(t) is the charge on the capacitor. Wed like to have q(t) in terms
of i(t). Recall that the charge on a capacitor and the current through the
capacitor is related by q 0 (t) = i(t). By the fundamental theorem of calculus:
Z t
q(t) = q(0) +
i(u) du
0
13
Expected
Actual
-5
-10
dB
-15
-20
-25
-30
-35
-40
101
102
103
104
105
Freq (Hz)
Expected
Actual
Phase (deg)
-20
-40
-60
-80
-100
101
102
103
104
105
Freq (Hz)
14
0
-5
Expected
Actual
-10
dB
-15
-20
-25
-30
-35
-40
101
102
103
104
105
Freq (Hz)
Expected
Actual
Phase (deg)
80
60
40
20
0
101
102
103
104
105
106
Freq (Hz)
15
vo (t) = R i(t)
We take the Laplace transform of vi (t):
1
Vi (s) = R I(s) + L[sI(s) i(0)] +
I(s)
sC
1
= I(s) R + Ls +
sC
Taking the Laplace transform of vo (t) is trivial:
Vo (s) = R I(s)
The transfer function is:
T (s) =
Vo (s)
R I(s)
=
Vi (s)
I(s) R + Ls +
1
sC
=
RC s
LC s2 + RC s + 1
This was tested against an actual RLC circuit with resistance, inductance,
and capacitance of:
R = 118.55
L = 50.66 mH
C = 0.2262 F
The result in shown in Figures 6 and 7.
Conclusion
The Laplace transform is a useful tool for solving differential equations analytically, especially in cases with discontinuous forcing terms or a periodic,
non-sinusoidal forcing term.
In addition, analysis of circuits in the s-domain can yield insights into the
frequency response of the circuits.
References
[1] John Polking, Albert Boggess, David Arnold, Differential Equations with
Boundary Value Problems, 2nd Edition, Pearson Prentice Hall, 2006.
[2] William Tyrrell Thomson, Laplace Transformation, 2nd Edition, PrenticeHall, 1960.
16
Expected
Actual
-5
-10
dB
-15
-20
-25
-30
-35
-40
101
102
103
104
105
Freq (Hz)
Expected
Actual
Phase (deg)
50
-50
-100
101
102
103
104
Freq (Hz)
17
105