CHAPTER 1

Circuit Analysis using Laplace Transforms

Let us consider the circuit shown in Fig. 1. The voltage source is a generated by a electrical sensor that is
measuring something in the physical world. The source is connected to a read-out display that is represented
by the 8 Ω load resistor, R3 . That display shows the voltage over the load resistor. The circuit components
between the source and the load resistor model the circuit connecting the sensor to the read-out display. If we
know there is a given input voltage, vin (t), generated by the sensor, then the question is what is the voltage
over the read-out display, vout (t), and how does that voltage compare to the input voltage?

The practical importance of this problem should be apparent. There are many applications where electrical
sensors are embedded in the physical world, and one wants the voltage displayed on the readout sensor to
represent what the sensor measures. The question is whether the “circuitry” between the sensor and read-out
distorts vout so it does not represent a good measure of what the sensor actually saw.

1 2 3

+ +

-
-

F IGURE 1. Sensor Circuit

This type of problem is distinct from the earlier ones we considered in DC or AC circuit analysis. In DC
circuit analysis, we assumed that the input voltage was constant and that all of the circuitry was purely
resistive. in AC circuit analysis we assumed the input was a sinusoid and were only interested in the steady-
state output sinusoid. In this case, the input voltage vin (t) is a general time-varying waveform and the circuit
components are no longer purely resistive.

Applying the mesh analysis to this circuit, we see there are three meshes for which we can identify the mesh
currents i1 (t), i2 (t), and i3 (t). We can also define two nodal voltages v1 (t) and v2 (t) as shown in Fig. 1.
Applying KVL around the loop traversed by i1 (t) yields

vin (t) = R1 i1 (t) + v1 (t)

Applying KVL around the loop traversed by i2 (t) yields,

di2
v1 (t) = R2 i1 (t) + L2 + v2
dt
1
2 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

Finally applying KVL around the loop traversed by i3 (t) yields

di3
v2 (t) = L3 + R3 i3 (t)
dt
and we take vout (t) = R3 i3 (t)

Combining the three loop equations yields a system of three ordinary differential equations,

(1) vin (t)= R1 i1 (t) + v1 (t)

di2
(2) v1 (t) = R2 i2 (t) + L2 + v2
dt
di3
(3) v2 (t) = L3 + R3 i3 (t)
dt
We also note that
dv1
(4) i1 (t) − i2 (t) = C1
dt
and
dv2
(5) i2 (t) − i3 (t) = C2
dt

Rewrite equation (4) to get an expression for i1 (t) and substitute this back into equation (1) and equation (2).
We then adjoin (3) and (5) to the revised equations (1-2) to get the following set of 4 equations.
 
dv1
vin (t) = R1 C1 + i2 (t) + v1 (t)
dt
di2
v1 ((t) = R2 i2 + L2 + v2
dt
di3
v2 (t) = L3 + R3 i3 (t)
dt
dv2
0 = C2 − i2 (t) + i3 (t)
dt
Note that this set of equations can be rewritten as a set of four coupled linear ordinary differential equations
dv1 vin (t) i2 (t) v1 (t)
= − −
dt R1 C 1 C1 R1 C1
dv2 1 1
= i2 (t) − i3 (t)
dt C2 C2
di2 1 1 R2
= v1 − v2 − i2
dt L2 L2 L2
di3 1 R3
= v2 (t) − i3 (t)
dt L3 L3
We can rewrite this in matrix-vector form
      
v1 − 1 0 − C11 0 v1 (t) 1
   R1 C 1 1 1    R1 C 1

 v2  =  0 0
d    C1 C2
  v2 (t)   0 
 +  vin (t)
dt  i2  
   1
L2 − L12 −RL2
2
0   i (t)
 2
 
  0 

i3 0 1
L3 0 −R
L3
3
i3 (t) 0
 1. NUMERICALLY INTEGRATING THE DIFFERENTIAL EQUATION 3

We can write this as a matrix differential equation

d
x = Ax + Bvin (t)
dt
with
     
v1 (t) − R11C1 0 − C11 0 1
R1 C1

 v2 (t)
  1 1   
  0 0 C1 C2
  0 
x= , A= , B=  vin (t)
 i (t)
 2




1
L2 − L12 −RL2
2
0 


 0 

i3 (t) 0 1
L3 0 −R
L3
3
0

and we can write the output voltage vout (t) as

vout (t) = Cx

where
h i
C= 0 0 0 R3

So what have we done. We’ve shown that the voltages across the two capacitors, v1 and v2 , and the currents,
i2 and i3 , through the two inductors satisfy a set of first-order ordinary differential equations and we’ve shown
that the output voltage is a linear combination of these four variables. This means that if we can find a way
to easily solve this system of ordinary differential equations, then we can find the response, vout (t), of the
circuit to any arbitrary input voltage vin (t). We are no longer confined to DC voltages or sinusoidal AC
voltages.

1. Numerically Integrating the Differential Equation

For large-scale networks, it often is convenient to use the computer to solve this. This is done by first obtaining
the first order ODE’s for the circuit and then numerically integrating the differential equation. Let us first see
how this might be done and then use our preceding example to illustrate what happens.

Let us start from a differential equation of the form

dx
= ax(t) + bu(t)
dt
where a and b are constants and u(t) is the applied input. We can approximate the derivative of x as
dx(t) x(t + h) − x(t)
≈
dt h
where h > 0 is a small time increment. If we insert this back into the differential equation we get
x(t + h) − x(t)
= ax(t) + bu(t)
h
which we rewrite as

x(t + h) = x(t) + (ax(t) + bu(t)) h

4 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

This is a recursive equation that takes the value of the state at time t and predicts what it will be at time t + h
dx(t)
by simply extrapolating by the derivative of x. Note that dt = ax(t) + bu(t), so this is essentially the same
as writing
 
dx(t)
x(t + h) ≈ x(t) + h
dt
This is sometimes called the forward Euler method for integration and provided h is sufficiently small then it
can provide the basis for an algorithm that recursively solves the above equation to compute x as a function
of time. Of course to start the recursion we need to know an initial value for x say at time t = 0.

In our example, we showed we could write the circuit equations as

dx(t)
= Ax(t) + Bvin (t)
dt
So the preceding discussion suggests we could compute a trajectory for the voltages and currents. In particu-
lar, we can implement this in MATLAB. I used the following script.

R1 = 1;R2 = .01;R3 = 1;
C1 = 1;C2 = 1;%
L2 = .1;L3 = 1;

%define ODE matrices

A = [-1/(R1*C1) 0 -1/C1 0;
0 0 1/C1 -1/C2;
1/L2 -1/L2 -R2/L2 0;
0 1/L3 0 -R3/L3];
B = [1/(R1*C1); 0 ; 0; 0];
C = [0 0 0 R3];

tstart = 0;
h = 1.e-2;
tduration = 50;
time = tstart:h:tstart+tduration;

x = zeros(4,1);
vin = ones(size(time));
%vin = cos(0.1*time);
vin(time<10) = 0;

data = [];
for t = 1:1:length(time)
xdot = A*x + B*vin(t);
x = x + xdot*h;
vout = C*x;
data = [data ; time(t) x’ vin(t) vout];
end

figure(1);
clf(1);
subplot(3,1,1);
plot(data(:,1),data(:,2),data(:,1),data(:,3),’linewidth’,2);
title(’capacitor voltages’);
legend(’v1(t)’,’v2(t)’);
subplot(3,1,2);
plot(data(:,1),data(:,4),data(:,1),data(:,5),’linewidth’,2);
title(’inductor currents’);
legend(’i2(t)’,’i3(t)’);

subplot(3,1,3);
plot(data(:,1),data(:,6),data(:,1),data(:,7),’linewidth’,2);
 1. NUMERICALLY INTEGRATING THE DIFFERENTIAL EQUATION 5

title(’input and output voltages’);

legend(’vin(t)’,’vout(t)’);

capacitor voltages capacitor voltages

0.6 0.5
v1(t) v1(t)
v2(t) v2(t)
0.4
0
0.2

-0.5
0
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50

inductor currents inductor currents

1 0.5
i2(t) i2(t)
i3(t) i3(t)
0
0.5
-0.5

0 -1
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50

input and output voltages input and output voltages

1 1
vin(t) vin(t)
vout(t) vout(t)

0.5 0

0 -1
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50

F IGURE 2. (a) step response (b) AC response

Figure 2(a) shows the capacitor voltages, v1 (t) and v2 (t), inductor currents, i2 (t) and i3 (t), and the in-
put/output voltage waveforms, vin (t) and vout (t) computed by this script for time between 0 and 50 seconds
with step size h = 0.01. The input waveform, vin (t), is a step input, that is zero for t < 10 and is 1 V for
t > 10. What we see in the other plots is the same basic behavior. That prior to the step input, the voltages,
currents, and outputs are all zero. After the input, these waveforms begin to rise until they reach a steady state
value. In other words, the behavior of the circuit, both in terms of its internal variables (we often call these
states) and the output have two phases. There is a transient interval over which these values are changing and
as time gets large, the system settles down to a constant value. In our earlier work with DC analysis, we were
only concerned with the steady state behavior. But clearly the transient behavior may be of interest also. In
this example, we see the initial voltages/currents are highly oscillatory immediately after the step occurs. For
obvious reasons, one may be concerned with what happens during this transient region.

We can also look and see what might happen with a sinusoidal input, by simply redefining the input as

vin = cos(0.1*time);

This input voltage is a unit amplitude sinusoid with a frequency of 0.3 rad/sec and zero phase angle. The
simulation plots are shown in Fig. 2(b), which again shows a highly oscillatory transient interval, after which
the response settles down to a sinusoid whose frequency is equal to that of the input voltage. This is, of
course, the steady-state AC circuit analysis that we studied in the preceding part of these lectures.

The preceding program used to generate the response of a circuit is called a digital simulation of the circuit. It
is often used to determine how circuits might respond, one of the best known such digital simulations is spice
and your book uses something called multisim. For most of these simulations, the main thing a student learns
is the “interface” between the user and the simulation. What we showed you how to do here was how to write
your own simulation model so you have a better understanding of what these programs can and cannot do.
6 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

So these simulations are able to simulate any input that I put into it. But, there are a number of things it
cannot do and some care must be taken in using them. In the first place, recall that the simulation was based
on discrete approximation of the derivative. This means the results will be sensitive to the size of h. If you
choose it very small, the simulation takes very long to run. If you choose it too big it will not be a good
approximation of the actual circuit response.

In addition to this, the simulation provides very little insight into why the circuit behaves the way it does. We
see for instance, there is a highly oscillatory transient in the circuit’s response, but why that is there is not
immdiately apparent. For a slightly different choice of parameters, the oscillation may get better or worse;
but we have no way of predicting how we might change this. Moreover, if we are interested in designing
the circuit so that the output “tracks” the input in some specified manner, then the simulation provides no
guidance in how we can redesign the circuit to accomplish this goal.

For these reasons, we need to introduce a more analytical way of predicting the circuit’s response, that at least
helps us predict what a simulation might produce; so we can see whether or not the simulation is accurate or
not. Such analytical tools, also provide the framework in which to better understand how such “oscillations”
occur and may, in some cases, be used to help control or minimize the impact of such oscillatory transients.
The second part of this course focuses on one particular approach for the formal analysis of a circuits total
response. This method is based on an integral transformation known as the Laplace transform; a mathematical
tool that is often taught in beginning ODE classes (so you may have already been exposed to some aspects of
this method).

2. Circuit Elements in the s-Domain

Rather than using the nodal analysis to write out the nodal equations as we did above, one can start directly
in the s-domain. We start by considering the constitutive relations for each circuit element.

From Ohm’s law we know that

v(t) = Ri(t)

But if we take the Laplace transform of both sides we get

V (s) = RI(s) ⇒ I(s) = GV (s)

1
where G = R.

Now consider the current-voltage equation for an inductor,

di(t)
v(t) = L
dt
We take the Laplace transform of both sides to get

V (s) = L(sI(s) − i(0− )) = sLI(s) − Li(0− )

 3. SIMPLE EXAMPLES 7

We can invert this relation to solve for I(s)

V (s) i(0− )
I(s) = +
sL s

Now consider the current-voltage equation for a capacitor,

dv(t)
i(t) = C
dt
Applying the transform yields,

I(s) = C(sV (s) − v(0− )) = sCV (s) − Cv(0− )

As before if we put this in terms of V (s) we get

v(0− )
 
1
V (s) = I(s) +
sC s

The s-domain current/voltage relationships for each passive device (resistor, inductor, capacitor) are illus-
trated in Fig. 3. Essentially, they allow one to “redraw” and “relabel” the impedance diagram drawn in the
time-domain into an s-domain impedance diagram. Note that in the s-domain, each circuit component with
a zero initial condition can be written as

V (s) = Z(s)I(s), I(s) = Y (s)V (s)

where Z(s) is the s-domain impedance of the device in units of ohms. For the resistor, Z(s) = R, for the
1
inductor Z(s) = sL and for capacitor Z(s) = sC . We can also choose to write the relationship in terms of
1
admittances; in which case the resistor’s admittance becomes Y (s) = R, the inductor’s admittance becomes
1
Y (s) = sL and the capacitor’s admittance becomes Y (s) = sC. These impedance/admittance values are
used to relabel the time-domain impedance diagram. If there are also initial currents through the inductor
or initial voltages over the capacitor, the circuit element would need to be augmented with the voltage or
current source, respectively, to account for that initial condition. Once this is done we have an s-domain
impedance/admittance diagram for the circuit and from that one can readily use KCL/KVL or the circuit
reduction methods discussed before to identify a set of algebraic equations that can be solved for the Laplace
transforms of the current/voltages in the original circuit. The benefit of this transformation is that solving for
that set of Laplace transforms is an algebraic problem and is much easier to do than the numerical integration
we presented above. Moreover, since the circuit is linear, we know the resulting transforms will be rational
functions and this means we can readily invert the transforms using partial fraction methods to obtain a
complete closed-form time-domain solution of the circuit.

3. Simple Examples

Natural Response of RC Circuit: Let us first consider the RC circuit on the left hand side of Fig. 4. This
is an RC circuit with an initial voltage over the capacitor of V0 and the switch then closes at t = 0 allowing
current, i(t), from the capacitor to flow through the resistor. The resulting output voltage v(t) represents the
“natural” response of the RC circuit without forcing.
8 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

s (Frequency)
Time Domain Domain

+ +

R R

- -

+ +

L
- -

- +
-

+
+ +
+
C

+ -
- -
-
-

F IGURE 3. s-domain circuits

 3. SIMPLE EXAMPLES 9

t=0
+
+ + +
+
- - - - -

F IGURE 4. RC Natural Response

In the time domain, we get the following differential equation model,

dv(t) v(t)
0=C + , v(0− ) = V0
dt R
We can start by taking the Laplace transform to get
V (s)
sCV (s) − CV0 + =0
R
Solving for V (s) gives
RCV0 V0
V (s) = =
RCs + 1 s + 1/RC
Taking the inverse transform from the table gives

v(t) = V0 e−t/RC u(t)

We can get this same Laplace transform relation by redrawing the time-domain impedance diagram using the
transformations in Fig. 3. If model the initial voltage as a source, then we get the middle figure. Applying
KVL around the loop gives
V0 1
=( + R)I(s)
s sC
which gives
CV0 V0 /R
I(s) = =
RCs + 1 s + (1/RC)
Taking the inverse transform gives
V0 −t/(RC)
i(t) = e u(t)
R
and since v(t) = Ri(t) we can conclude

v(t) = V0 e−t/(RC) u(t)

which is the same thing we got above.

We can also model the initial voltage on the capacitor using a current source. KCL at the top node gives
V (s)
CV0 = sCV (s) +
R
10 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

and solving V (s) gives

V0
V (s) =
s + (1/RC)
which is identical to what we got with the time-domain analysis.

Step Response of Parallel RLC Circuit: Let us now consider the parallel RLC circuit shown in Fig. 5.
In this case the switch is closed prior to time t = 0 so there is no initial voltage over the capacitor (i.e.
vC (0− ) = 0) and no initial current through the inductor. Immediately after the switch is open the current
flows through the parallel elements.

We can apply KCL at the top node to get

Z t
v(t) dv(t) 1
Idc u(t) = +C + v(t)dt
R dt L 0

We can solve this by taking the Laplace transform

Idc V (s) V (s)
= + sCV (s) +
s R sL
and solving for V (s) gives
Idc /C
V (s) =
s2 + (1/RC)s + (1/LC)
We get this same equation if we apply KCL at the top node of the s-domain diagram on the right side of
Fig. 5.

25 nF 625 ! 25 mH +
24 mA
t=0
-

F IGURE 5. Parallel RLC

The current through the inductor is

V (s) Idc /LC
IL (s) = =
sL s(s2 + (1/RC)s + (1/LC)
and substituting the numerical values for Idc , R, L, and C we get
384 × 105
IL (s) =
s(s2 + 64, 000s + 16 × 108
We factor the terms in the quadratic part of the dominator to get
384 × 105
IL (s) =
s(s + 32, 000 − j24, 000)(s + 32, 000 + j24, 000)
 3. SIMPLE EXAMPLES 11

We can test the s-domain respresentation for IL by using the initial value theorem to make sure that iL (∞) =
24mA. In particular, we see
384 × 105
lim sIL (s) = = 24 mA
s→0 16 × 108
So this checks out.

We now proceed with a partial fraction expansion to find the inverse Laplace transform.
K1 K2 K2∗
IL (s) = + +
s s + 32000 − j24000 s + 32000 + j24000
and evaluating the coefficients gives
384 × 104
K1 = = 24 × 10−3
16 × 108
384 × 105
K2 = = 20 × 10−3 ∠126.87◦
(−32000 + j24000)(j48000)
Substituting numerical values for K1 and K2 and inverse transforming gives,

iL (t) = 24 + 40e−32,000t cos(24, 000t + 126.87◦ ) u(t),

 
mA

Another example of using the Laplace transform arises from replacing the DC current source in Fig. 5 with a
sinusoidual current source of the form

ig (t) = Im cos ωt

where Im = 24 mA and ω = 40, 000 rad/s. As before the initial stored energy in the inductor and capacitor
is zero.

The s-domain expression for the source current is

sIm
Ig (s) =
s2 + ω2
The voltage across the parallel elements is
(Ig (s)/C)s
V (s) =
s2 + (1/RC)s + (1/LC)
(Im /C)s2
=
(s2 + ω 2 )(s2 + (1/RC)s + (1/LC))
The current through the inductor is then
V (s) (Im /LC)s
IL (s) = = 2
sL (s + ω 2 )(s2 + (1/RC)s + (1/LC))
Substituting numerical values for the parameters gives
384 × 105 s
IL (s) =
(s2 + 16 × 108 )(s2 + 64, 000s + 168 )
which we factor as
384 × 105 s
IL (s) =
(s − jω)(s + jω)(s + α − jβ)(s + α + jβ)
12 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

where ω = 40, 000, α = 32, 000, and β = 24, 000.

WE expand this into a sum of partial fractions

K1 K1∗
IL (s) = +
s − j40, 000 s + j40, 000
K2 K2∗
+ +
s + 32000 − j24000 s + 32000 + j24000
Evaluating the numerical coefficients gives
384 × 105 (j40, 000)
K1 =
(j80, 000)(32, 000 + j16, 000)(32, 000 + j65, 000)
= 7.5 × 10−3 ∠ − 90◦
384 × 105 (−32, 000 + j24, 000)
K2 =
(−32, 000 − j16, 000)(−32, 000 + j64, 000)(j48, 000)
= 12.5 × 10−3 ∠90◦

Inverting this then gives

15 cos(40, 000t − 90◦ ) + 25e−32000t cos(24000t + 90◦ ) u(t)

iL (t) =

The steady state current is 15 sin(40000t) mA which can be verified using the phasor analysis we studied
earlier.

Step Response of Multi-Mesh Circuit: We now apply the Laplace transform method to determine the step
respose of a circuit with two meshes in it. The circuit is shown in Fig. 6.

8.4 H 10 H 8.4 H 10 H

t=0 t=0
42 ! 48 !
+ 42 ! 48 ! +

336V
- -

F IGURE 6. Example 13-15

Here we want to find the branch currents i1 (t) and i2 (t) when the 336 V DC voltage source is suddenly
applied across the circuit. We assume the initial current though both inductors is zesro. There are two mesh
current equations that in the s-domain are
336
= (42 + 8.4s)I1 (s) + 42I2 (s)
s
0 = −42I1 (s) + (90 + 10s)I2 (s)

We can write this in Matrix vector form as

" # " #" #
336
s (42 + 8.4s) −42 I 1 (s)
=
0 −42 (90 + 10s) I2 (s)
 3. SIMPLE EXAMPLES 13

" #
a b
Given a matrix A = , one can easily verify that
c d
" #
d −b
−c a
A−1 =
ad − bc
So we can readily show that
" # " #" #
336
I1 (s) 1 (90 + 10s) 42 s
=
I2 (s) (42 + 8.4s)(90 + 10s) − 422 42 (42 + 8.4s) 0
" 40(s+9)
# " 40(s+9)
#
s(s2 +14s+24) (s(s+2)(s+12)
= 168
= 168
s(s2 +14s+14) s(s+2)(s+12)

Expanding this out into partial fractions gives

15 14 1
I1 (s) = − −
s s + 2 s + 12
7 8.4 1.4
I2 (s) = − +
s s + 2 s + 12
and taking the inverse transform gives

i1 (t) = (15 − 14e−2t − e−12t )u(t)

i2 (t) = (7 − 8.4e−2t + 1.4e−12t )u(t)

Use of Thevenin Equivalent: Consider the s-domain circuit shown in Fig. 7. We assume the circuit is
initially at rest and that at t = 0 a 480V constant voltage is applied across the RLC combination. What we
want to do is determine the capacitor current, ic (t) and capacitor voltage , vc (t), waveforms. We will do this
by constructing a Thevenin equivalent circuit for the circuit shown in the yellow box.

t=0
a a a
+ +
+
+ + +

- - -
-
- -
b
b b

F IGURE 7. Example 13.17

The Thevenin equivalent circuit is found by first redrawing the impedance diagram in the s-domain and then
determining the open circuit voltage when the capacitor is not connected. This is obtained by a voltage divider
across the 20 ohm resistor and the 0.002s inductor,
480 0.002s 480 480
Vth (s) = = =
s 20 + 0.002s (20/.002) + s s + 10×4
The Thevenin equivalent impedance is obtained by first computing the short circuit current, Isc . Note that if
we short terminals a and b, then the impedance seen by the voltage source is
(60)(0.002s) 80(s + 7500)
Z(s) = 20 + =
60 + 0.002s s + 30000
14 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

The total current drawn out of the source is

480 60(s + 30000)
I(s) = =
sZ(s) s(s + 7500)
and the short circuit current is obtained by a current divider,
0.002s 6
Isc (s) = I(s) =
60 + 0.002s s + 7500
The Thevenin equivalent impedance is then
Vth (s) 80(s + 7500)
Zth (s) = =
I(s) s + 10000
which is shown in Fig. 7.

The current Ic (s) through the capacitor is then obtained as

480 2 × 105 1 6s
Ic (s) = =
s + 104 s 80(s+7500)
+ 2×105 (s + 5000)2
s+104 s

and the voltage drop is

1 2 × 104 6s 12 × 105
Vc (s) = Ic (s) = =
sC s (s + 5000)2 (s + 5000)2

We now compute the inverse transform for both the current and the voltage. The current’s inverse transform
is obtained from a partial fraction expansion,
6s −30000 6
Ic (s) = = +
(s + 5000)2 (s + 5000)2 s + 5000
which gives the inverse transform,

ic (t) = (−30, 000te−5000t + 6e−5000t )u(t)

and the voltage is simply

vc (t) = 12 × 105 te−5000t u(t)

Example with Mutual Inductance: The next example looks at the transient response of a circuit with a
mutual inductance. Fig. 8 shows the circuit. The switch shown in Fig. 8 switches the circuit topology at time
t = 0. We want to determine the current i2 (t).

t=0
3! 2!
t=0 t=0
9! 3! 2H 2! 3! 2!
10!
+
60V 2H 8H 10! 2H 10! -
-
+

F IGURE 8. Mutual Inductance Example

 3. SIMPLE EXAMPLES 15

The circuit model shown in the figure assumes that the voltage over primary and secondary coils are given as
di1 di2
v1 = L1 +M
dt dt
di1 di2
v2 = M + L2
dt dt
We want to develop an equivalent circuit model relating v1 and v2 that allows us to use KCL and KVL. In
other words, we want to replace the inductively coupled coils (no direct connection) with an equivalent circuit
over which KCL holds.

To obtain an equivalent (T-equivalent) circuit model, we regard the preceding equations as mesh-current
equations with i1 (t) and i2 (t) as the mesh variables. In this case, we can see the coupled coils being replaced
by a T -circuit in which the inductance on the primary side is L1 − M carrying i1 (t), the inductance on the
secondary sice is L2 − M carrying i2 (t), and the inductance in the T is M with both currents i1 (t) + i2 (t).
We use this transform to redraw our circuit in Fig. 8. We then transform this circuit into the s domain. We
note that prior to the switch the current
60
i1 (0− ) = =5
9+3
and the current in the secondary coil I2 (0− ) = 0.

We wil use mesh analysis in the s-domain, so we use the series equivalent circuit fo ran inductor carrying
an inditial current. This s-domain circuit is shown on right panel of Fig. 8. Note that there is only one
independent voltage source. This source appears in the vertical leg of the tee to account for the intiial value
of the curren tin the 2 H inductor, (L1 ), of i1 (0− ) + i2 (0− ) = 5 amps. The branch with the inductor L1 − M
has no source because L1 −M = 0. The branch with the inductor L2 −M has no source because i2 (0− ) = 0.

The two s-domain mesh equations are

(3 + 2s)I1 (s) + 2sI2 (s) = 10

2sI1 (s) + (12 + 8s)I2 (s) = 10

and solving for I2 yields,

2.5
I2 (s) =
(s + 1)(s + 3)
Expanding this as a sum of partial fractions generates
1.25 1.25
I2 (s) = −
s+1 s+3
and taking the inverse transform yields,

i2 (t) = (1.25e−t − 1.25e−3t )u(t)

If we plot this, we see that the current i2 increases from zero to a pea value of 481 mA in 550 ms after the
switch is moved to position b before decaying down to zero.

Use of Superposition: Figure 9 shows a circuit that has two sources. We will use the principle of superpo-
sition to determine this circuit’s transient response where we want to find the voltage V2 (s). The left side of
16 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

Fig. 9 shows the time-domain impedance diagram with an initial capacitor voltage γ V and an initial inductor
current of ρ A. The s-domain impedance diagram is shown on the right side of Fig. 9 where we opted to
shows the initial voltages and currents as current sources because we’ll be using a nodal voltage method.

+ -
+ +
+ +
- -
- -

F IGURE 9. Principle of Superposition

To find V2 (s) by superposition, we calculate the component of V2 resulting from each of the sources acting
alone, and then sum up the components. We begin with Vg (s) acting along. Opeing each of the three current
sources deactivates them. Fig. 10(a) shows the resulting circuit for Vg alone. The two equations that describe
this circuit are
 
1 1 (a) (a) Vg (s)
+ + sC V1 (s) − sCV2 (s) =
R1 sL R1
 
(a) 1 (a)
−sCV1 (s) + + sC V2 = 0
R2
For convenience, we represent the admittances in the above equation as
1 1
Y11 (s) = + + sC
R1 sL
Y12 (s) = −sC
1
Y22 (s) + + sC
R2
Substituting back into the equation yields
(a) (a) Vg (s)
Y11 (s)V1 (s) + Y12 (s)V2 (s) =
R1
(a) (a)
Y12 (s)V1 (s) + Y22 (s)V2 (s) = 0

We can solve these equations using any method you want to see that
(a) −Y12 (s)/R1
V2 (s) = 2 (s) Vg (s)
Y11 (s)Y22 (s) − Y12

We now look at the current source Ig (s) acting alone. The resulting circuit is obtained by shorting the voltage
source and opening the other current sources. The simplified circuit is shown in Fig. 10(b). In this case, the
two nodal equations describing the circuit are
(b) (b)
Y11 (s)V1 (s) + Y12 (s)V2 (s) = 0
(b) (b)
Y12 (s)V1 (s) + Y22 (s)V2 (s) = Ig (s)
 3. SIMPLE EXAMPLES 17

+ + +
+ +

-
- - - -

+ +
+ +
+

- -
- - -

F IGURE 10. Principle of Superposition

(b)
which we solve for V2 (s) to obtain

(b) Y11 (s)

V2 (s) = 2 (s) Ig (s)
Y11 (s)Y22 (s) − Y12

To find the component of V2 (s) resulting from the intiial energy stored in the indctuor, we solve the circuit
shown in Fig. 10(c). In this case the nodal equations are

(c) (c) ρ
Y11 (s)V1 (s) + Y12 (s)V2 (s) = −
s
(c) (c)
Y12 (s)V1 (s) + Y22 (s)V2 (s) = 0

which we solve to get

(c) Y12 (s)/s

V2 (s) = 2 (s) ρ
Y11 (s)Y22 (s) − Y12

Finally we determine the component of V2 resulting from the initial energy stored in the capacitor. The nodal
equations for this circuit in Fig. 10(d) are

(d) (d)
Y11 (s)V1 (s) + Y12 (s)V2 (s) = γC
(d) (d)
Y12 (s)V1 (s) + Y22 (s)V2 (s) = −γC

which gives

(d) −(Y11 (s) + Y12 (s))C

V2 (s) = 2 (s) γ
Y11 (s)Y22 (s) − Y12
18 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

So the total expression for V2 (s) is obtained by summing these four components together.
(a) (b) (c) (d)
V2 (s) = V2 (s) + V2 (s) + V2 (s) + V2 (s)
−(Y12 /R1 ) Y11
= 2 Vg + Y Y 2 Ig
Y11 Y22 − Y12 11 22 − Y12
Y12 /s −C(Y11 + Y12 )
+ 2 ρ+ Y Y 2 γ
Y11 Y22 − Y12 11 22 − Y12

4. Opening Circuit Example Revisited

In the opening of this chapter, we introduced a complex circuit in Fig. 1 that had four energy storing devices
and hence there were four circuit equations that we had to solve for. To summarize the resulting circuit
equations derived before were
      
v1 − R11C1 0 − C11 0 v1 1
R1 C1
   1 1    
 v2  = 0 0
d    C1 C2
  v2   0 
 i  + 
     vin (t)
dt  i2 




1
L2 − L12 −RL2
2
0  2   0 

i3 0 1
L3 0 −R
L3
3
i3 0
 
v1
h i 
 v2 
vout (t) = 0 0 0 R3 
 
i
 2 


i3
We write this more compactly as

ẋ = Ax + Bvin (t)
(6)
y = Cx
 
v1
 
 v2 
where x is a column vector whose components are the four variables we need to solve for x = 
  and
 i2 


i3
the A, B, and C matrices are defined in the preceding equation.

In that section, we did not try to solve for x(t) analytically, instead we simply used the computer to numeri-
cally integrate the differential equation. The reason for doing this was that, provided the numerical integration
was “good enough”, we could then use this method to predict the response of a circuit that has too many en-
ergy storing devices whose variables can be solved for “by hand”. The use of the computer, in this way to
simulate the response of a circuit is commonly done. In particular tools such as Multisim and Spice take the
circuit diagram, form the differential equations automatically, and then numerically integrate those equations
to predict the response.

Simulation based analysis of linear circuits is very effective and routinely used, but it provides little insight
into how one might design or modify an existing design to improve the circuit’s performance. The Laplace
 4. OPENING CIRCUIT EXAMPLE REVISITED 19

transform methods introduced above for circuit analysis do provide considerable insight into how one might
design a desired circuit. Those methods provide a basis for design that will be studied in future courses. I
will introduce some of the design concepts in the next chapter, but for now ask you to accept that the use
of Laplace transform methods have value beyond merely predicting system response and in this way they
provide something that circuit simulation is not really capable of addressing.

So if we are to use Laplace transform analysis methods to better understand how a complex circuit works,
does that mean one has to solve “large” circuit equations by hand in the way we did above? The answer is no.
The methods we showed above can, in fact, be implemented on a computer, essentially having the computer
do the “symbolic” analysis that you did by hand before. The purpose of this section is to show you how the
computer can be used in this manner, relying on the example we started with in Fig. 1.

Let us start by taking the “compact” matrix form of the circuit that we showed you in equation (6). Since the
Laplace transformation is “linear”, we can apply it directly to that equation to obtain

sX(s) − x(0+ ) = AX(s) + BVin (s)

Y (s) = CX(s)

where X(s) is the single sided Laplace transform of the vector x(t) and Y (s) is the single sided Laplace
transform of the y(t). Note that the above equations are “algebraic” equations rather than differential equa-
tions and so we can manipulate the first equation to get

(sI − A)X(s) = x(0+ ) + BVin (s)

where I is an 4 by 4 identity matrix (since x is a 4-vector). Note that the term in the parentheses on the
left side of the equation is a 4 by 4 matrix and provided the inverse matrix (sI − A)−1 exists, then we can
multiply both sides of the equation to get

X(s) = (sI − A)−1 x(0+ ) + (sI − A)−1 BVin (s)

And since Y (s) = CX(s), we can readilly conclude that the Laplace transform of the output y satisfies

(7) Y (s) = C(sI − A)−1 x(0+ ) + C(sI − A)−1 BVin (s)

where Vin (s) is the Laplace transform of the input voltage vin (t) and x(0+ ) is the initial energy on each of
the energy storing elements immediately after time t = 0.

Can we find the inverse (sI − A)−1 ? Of course, this is essentially what you did in solving the mesh circuit
example above. In that case, there was a well known “algebraic characterization” of the inverse of a 2
by 2 matrix. We can compute such inverses for larger matrices relying essentially on the same Gaussian
elimination methods that we talk about earlier. Because these methods are “algorithmic” in nature (i.e.
there is a well-defined sequence of steps to be taken to achieve the desired goal), one can write a program
that automates the computation of this inverse, the only difference being that rather than treating the s as a
number, we treat it as a “symbolic” variable that we just carry along in the computation. This is sometimes
called computer algebra and tools such as Mathematica, Maple, and SINGULAR are all computer programs
20 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

that do this. MATLAB even has a symbolic engine built into it, so in principle we can do the Laplace
transform analysis of our 4 element circuit using the computer, rather than doing a “hand” calculation.

The following script does this computation for the circuit in question.

%declare ’s’ as a symbolic variable

syms s

%Laplace transform of step input

step_input = 1/s;

%Laplace transform of inv(sI-A)

resolvent = simplify(inv(s*eye(4)-A));

%natural response, forced response, and total response

natural_response = C*resolvent*x0;
forced_response = C*resolvent*B*step_input;

total_response = simplify(forced_response+natural_response)

The output generated by this bit of script is

total_response =

100/(s*(10*sˆ4 + 21*sˆ3 + 222*sˆ2 + 312*s + 201))

So we know that
10
Y (s) =
s5 + 2.1s4 + 22.2s3 + 31.2s2 + 20.1s

To compute the response we first need to find the roots of the denominator polynomial and compute the
residues for each root. This procedure that you walked through before by hand, can be computed automati-
cally using the following MATLAB script

%extract numerator/denomator polynomial

% in total_response
[num,den] = numden(total_response);

%extract coefficients of num/den polynomial

% be sure to convert to double
num = double(coeffs(num,’All’));
 4. OPENING CIRCUIT EXAMPLE REVISITED 21

den = double(coeffs(den,’All’));

%compute residues of rational function num/den

[r,p,k] = residue(num,den)

In this script, we first extract the numerator and polynomial polynomials using the command numden. The
results are symbolic and need to be converted to a floating point array before it can be used by MATLAB
to find the residues. The command coeffs extracts the coefficients of the polynomials and we need to use
double to convert them from a symbolic type to a double floating point type. The MATLAB command
residue automates what you did by hand in finding the residues of the partial fraction expansion. This
command returns three floating point arrays, r is an array of residues, p is an array of poles, and k is
nonempty if the rational function is not strictly proper. In our case, the output generated by this script is

r =

0.0122 + 0.0017i
0.0122 - 0.0017i
-0.2610 + 0.2852i
-0.2610 - 0.2852i
0.4975 + 0.0000i

p =

-0.2966 + 4.4976i
-0.2966 - 4.4976i
-0.7534 + 0.6494i
-0.7534 - 0.6494i
0.0000 + 0.0000i

k =

[]
22 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

From this we can see that

0.0122 + 0.0017j 0.0122 − 0.0017j
Y (s) = +
s + 0.2966 − 4.4976j s + 0.2966 + 4.4976j
−.2610 + 0.2852j −.2610 − 0.2851j
+ +
s + 0.7534 − 0.6494j s + 0.7534 + 0.6494j
0.4975
+
s
K1 K1∗
= +
s + α1 − jβ1 s + α1 + jβ1
K2 K2∗
+ +
s + α2 − jβ2 s + α2 + jβ2
K3
+
s
These transforms are in our table we can immediately see the time domain response is

y(t) = 2|K1 |eα1 t cos(β1 t + θ1 )

+2|K2 |eα2 t cos(β2 t + θ2 )
+K3

in which

i |Ki | θi αi βi
◦
1 0.0123 8.0679 −0.2966 4.4976
◦
2 0.3866 132.4582 −0.7534 0.6494
◦
3 0.4975 0 0 0

The preceding table was computed directly from the output of the residue command and then used to
construct the response of each mode using the following script.

K1 = abs(r(1));
theta1 = angle(r(1));
alfa1 = real(p(1));
beta1 = imag(p(1));
yx(1,:) = 2*K1*exp(alfa1*time).*cos(beta1*time+theta1);

K2 = abs(r(3));
theta2 = angle(r(3));
alfa2 = real(p(3));
beta2 = imag(p(3));
yx(2,:) = 2*K2*exp(alfa2*time).*cos(beta2*time+theta2);

K3 = abs(r(5));
yx(3,:) = K3.*ones(size(time));
 4. OPENING CIRCUIT EXAMPLE REVISITED 23

figure(2);
subplot(2,1,1);
plot(data(:,1),data(:,7),’linewidth’,2);
title(’input and output voltages’);
axis([0 50 0 1]);
subplot(3,1,2);
plot(time,yx(1,:)+yx(2,:)+yx(3,:),’linewidth’,2);
title(’vout(t) s-domain predicted’);
axis([0 50 0 1]);
subplot(3,1,3);
plot(time,yx(1,:),time,yx(2,:),time,yx(3,:),’linewidth’,2);
legend(’mode 1’, ’mode 2’, ’mode 3’);
title(’response of each mode in PFE’);

The result of this is shown below in Fig. 11. The top plot shows the output vout (t) we computed earlier by
simulation. The middle plot shows the output vout (t) computed using Laplace transform analysis, in which
the computer was used to construct the Laplace transform of the output and was also used to compute the
residues of the transform’s partial fraction expansion. The bottom plot, plots each “mode” of the partial
fraction expansion where we see there are three modes. The third mode K3 /s was the step input. The second
mode shows a well defined response that is relatively slow (around 5 second rise time). The first mode (blue)
shows a much higher frequency behavior that accounts for the ripple in the actual step responses. It is in this
last plot, that we begin to get a deeper understanding of how the circuit actually behaves. In particular, we
see there are 3 fundamental responses of “modes” for the system’s response. Mode 3 simply captures the
steady state step input to the system. The other two modes (in blue and red) represent fundamental behaviors
associated with the system’s natural response. In particular, we see that the system has two types of natural
response. The first (mode 1) is a lightly dampled high frequency oscillation and the second is a slower more
heavily damped low frequency mode.
24 1. CIRCUIT ANALYSIS USING LAPLACE TRANSFORMS

input and output voltages

1

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35 40 45 50

vout(t) s-domain predicted

1

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35 40 45 50

response of each mode in PFE

0.5
mode 1
mode 2
mode 3

-0.5
0 5 10 15 20 25 30 35 40 45 50

F IGURE 11. Step Response for simulated and s-domain analysis