Sie sind auf Seite 1von 12

Sensors and Control Systems

Lecture Topic 5
The Dynamic Response of First- and Second-Order Systems
1 Introduction
In previous lectures, we have seen how to develop a transfer function to describe the dynamic behaviour
of a system. In these notes we will look at two of the most common (and widely studied) forms of transfer
function, first- and second-order, and examine how they respond to some commonly-used input signals.
These notes build upon discussion and data from the course text "Essentials of Control" by
J.Schwarzenbach, to which you referred for more details.
Throughout these notes this symbol has been used to mark points at which you should try to verify the
results given or answer the questions asked in your own time or during tutorials.

2 Examples of First-Order Systems


2.1

DC Motor with Inertia

Just to refresh our memory from the last lecture, we derived the following transfer function to relate the
input voltage for our DC motor system to the rotational speed of the final shaft.

G(s) =
=

W f (s)
Vi ( s )
AK t
B( JRa s + K b K t )

Note that this is a first-order system because it only includes the first power of s on the denominator of
the transfer function.
We can rearrange this expression as follows:

G ( s) =
=

AK t
B( JRa s + K b K t )
A

BK b

1
JRa
1+
s
Kb Kt

K
1 + s
JRa
and =
Kb Kt
=

where K =

A
BK b

K is known as the steady-state gain and (tau) is the time constant of the system. We will look at the
significance of these terms in more detail later in these notes.

Page 1 of 12

David Lee, Oct 2002

2.2

RC circuit

As another example, consider the simple RC network shown below. (If you recall, this circuit was
introduced as a filter for the output of the tacho-generator in your notes for Topic 4.)
R
i(t)

Vi(t)

Vo(t)
C

Note that

Vi (t ) Vo (t ) = i (t ) R
and

i (t ) = C

dVo (t )
dt

If we combine these expressions, we obtain

Vi (t ) = Vo (t ) + RC

dVo (t )
dt

and take Laplace transforms to give:

Vi ( s ) = Vo ( s )(1 + RCs )
So our transfer function can now be written as:

Vo ( s )
1
=
Vi ( s ) 1 + RCs
This now has the same form as our motor transfer function, with a steady state gain, K, of 1 and time
constant equal to RC.

Page 2 of 12

David Lee, Oct 2002

2.3

Water Tank

As a final example of a first-order system, consider the water tank shown below. It has a cross-sectional
area A and is filling with water at rate qi (t ) while draining through a narrow pipe at rate q o (t ) .

qi(t)

qo(t)

h(t)

A
If we let h(t ) represent the height of the water in the tank and assume that the rate of flow from the tank is
proportional to the height of the water, we can have:

qi (t ) q o (t ) = A

dh(t )
dt

and

q o (t ) = Ch(t )
We can then derive the transfer function from the input flow rate to the height of water as follows:

qi (t ) = Ch (t ) + A

dh (t )
dt

Qi ( s ) = H ( s )(C + As )
H (s)
1
=
Qi ( s ) C + As
1/ C
1 + ( A / C )s
K
=
1 + s
=

where K = 1 / C and = A / C .
Note the similarities between the transfer functions in all three of the above cases. Although the systems
are very different in their physical implementations, their transfer functions are all of the same form.
Since the transfer function is designed to embody all of the dynamic behaviour of the system, this implies
that they should behave similarly to similar types of input. In the following sections we will look at the
response of a first-order system to a range of inputs. For this discussion we will describe the behaviour of
the system in terms of its time constant and its gain K. If we wish to consider how one of our physical
systems would actually behave, we have to work out how, for example, changes as we change the
physical constants of our system.

Page 3 of 12

David Lee, Oct 2002

3 The Response of a First-Order System to Input Signals


3.1

Introduction

It is usual to examine the behaviour of a system when it is presented with a range of standard input signals
or "forcing functions". These standard signals are designed to represent, if in a rather idealized form,
some of the more common signals that would be encountered by a real system.

3.2

Step Input

A step input is one of the simplest inputs that we can consider for a system. The input suddenly changes
from zero to a constant value as if, for example, a switch had been turned on. A step input x(t ) is
considered to be zero for all t < 0 , and take a constant value for t 0 . For simplicity, we will consider a
unit step function so that x (t ) = 1 for t 0 . For step inputs of other magnitudes, we can simply scale the
output proportionally. ( Why?)
From a table of Laplace transforms, we can see that the Laplace transform of a unit step function is given
by

X ( s) =

1
s

So the Laplace transform of the output of out first-order system with a step input is given by

Y (s) =

K
s (1 + s )

By using the method of partial fractions and a table of inverse Laplace transforms, we can deduce ( )
that the output of our system is given by:
t
y (t ) = K 1 e

Note that, as t , y (t ) K . For large values of t the system settles into a steady state in which its
output is simply K times the input. Hence the name "steady-state gain" for K. But, before the system
settles into its steady state, it will exhibit short-term "transient" behaviour in response to the changes in
the input. The following image shows the output of a first-order system (with K set to 1) in response to a

unit step input.


Notice that the value of indicates how rapidly the output signal approaches its steady-state value,
reaching 63% of its steady-state value after one period of has elapsed. ( Why 63%? What fraction of
the steady-state value would be reached after 2, or 3?)

Page 4 of 12

David Lee, Oct 2002

3.2.1 Final Value Theorem.


We saw in the previous section that how the output of our system approached a constant value for large
values of t. The final value theorem shows that we could predict that final value simply by considering the
Laplace transform of the output. In other words, we wouldn't need to work out the inverse Laplace
transform if we were just looking for the final value.
The final value theorem states that:

lim f (t ) = lim sF ( s )
t

s0

So, to find the final value of a function of time, we take the value of s times its Laplace transform as s
approaches zero. Take, for example, the output of our first-order system with a unit step input. The
Laplace transform was:

Y (s) =

K
s (1 + s )

So, to find the final value of f(t), we calculate

sK
s 0 s (1 + s )
K
= lim
s 0 1 + s
=K

lim sY ( s ) = lim
s 0

which confirms the result we found by applying the inverse Laplace transform.
Notice, by the way, that the steady state gain for our DC motor system is given by

K=

A
BK b

which is the gain that we calculated for our system before we took the inertia of the system into account.
This makes sense because, by the time the system has reached its final state, the motor is no longer
accelerating and inertia is no longer having an effect.

3.3

Ramp Input

The ramp input corresponds to a gradual, linear increase in the input signal, as shown by the dotted line
on the graph below.

Mathematically, we say that x (t ) = t for t 0 . Again, by using a table of Laplace transforms, we find
that

X ( s) =
Page 5 of 12

1
s2

David Lee, Oct 2002

and so the Laplace transform of the output of our unit-gain first-order system is:

Y (s) =

1
s (1 + s )
2

Partial fractions and inverse transforms this time give us ( )

y (t ) = t + e

Notice that, once the transients have died down, the output settles into a steady state in which
y (t ) = t . This can indeed be viewed as a steady state although, unlike the previous case, it is not a
constant value.
( What happens if we apply the final value theorem in this case?)

3.4

Impulse Input

The third type of transient input that we usually consider is the unit impulse. This is a very short "pulse"
of input at time t=0. The Laplace transform of a unit impulse is X(s)=1 and the output of our unit gain
first-order system is therefore ( ):

y (t ) =
as shown on the following graph

Note that, as we might expect from a brief "pulse" of input, the steady-state response of the system is
zero. The output signal decays to about 2% of its initial value after a period of 4. ( Check this.)

3.5

Sinusoidal Input

Sinusoidal inputs are quite common in vibrations of mechanical systems and oscillations of electrical
systems. They are also useful because, by Fourier decomposition, we can form arbitrary periodic inputs
by summations of sinusoidal signals. Because we have limited the discussion to linear systems that obey
the principal of superposition, the response to our system will be the sum of the individual responses to
the component frequencies. In this way we can determine the response of the system to an arbitrary
periodic input signal, by considering the Fourier components of the input.

Page 6 of 12

David Lee, Oct 2002

If a first-order system receives a sinusoidal input of frequency , the steady-state output will also be
sinusoidal, with the same frequency, but typically with a different magnitude and a different phase from
the input, as shown below.

As we would expect, the transfer function allows us to determine this magnitude and phase for a given
frequency.
If we have a system with a transfer function G (s ) and we want to know how it will respond to a
sinusoidal input of frequency , we simply have to calculate the value of G ( j ) . This will tell us
the magnitude of the response (the "gain") and the phase difference between the input and output.
For example, with a sinusoidal input to our first-order system, we can calculate:

1
1 + j
1 j
=
1 + 2 2

1
=
j
2 2
1+
1 + 2 2

G ( j ) =

We can now express this complex number as a magnitude and a phase angle ( revise this if necessary),
to give:

G ( j ) =

1 + 2 2
G ( j ) = tan 1

Page 7 of 12

David Lee, Oct 2002

The response of a system to a sinusoidal input is often shown on a Bode plot like the one shown below,
with the magnitude in decibels, the phase in degrees, and the frequency on a logarithmic scale.

()
For very small , the magnitude is 1 and the phase lag is zero. For large , the magnitude tends to zero
(satisfying our earlier need for a low-pass filter) while the phase lag tends to a maximum of 90 degrees.
(Negative phase values correspond to a phase lag.) The dotted lines on these graphs show straight-line
approximations that can be used for these curves.
Expressing our transfer functions as magnitudes and phases makes it easier for us to combine transfer
functions for multiple blocks in series. To obtain the transfer function of the overall system, we can
simply multiply the individual magnitudes and add the phase shifts. (If the magnitudes are expressed in
decibels, we can also add the magnitudes.)

4 Second-Order Systems
4.1

Introduction

In practical control applications, it is common to find that our system can only be described using second
differentials (such as acceleration in a position control system). As with the first-order systems, we can
describe the transfer functions of second-order functions in a standard form and consider the response of
such systems to step, ramp, impulse and sinusoidal inputs.

4.2

RLC Circuit

As an example, take the RLC circuit shown below.


R

L
i(t)

Vi(t)

Vo(t)
C

Assume that the input impedance is zero, so that the input voltage vi(t) does not depend on the current
drawn and that the output impedance is infinite, so that there is no current flow to the right. Let v(t)
represent the voltage in between R and L.

Page 8 of 12

David Lee, Oct 2002

We then have:

vi (t ) v(t ) = i (t ) R
di (t )
dt
dv (t )
i (t ) = C 0
dt

v(t ) v o (t ) = L

The Laplace transforms of these equations give us:

Vi ( s ) V ( s ) = I ( s ) R
V ( s ) Vo ( s ) = LsI ( s )
I ( s ) = CsVo ( s )
which can be rearranged to give the following transfer function:

Vo ( s )
1
=
2
Vi ( s ) LCs + RCs + 1
=

1 / LC
s + ( R / L) s + 1 / LC
2

Notice that we do indeed have a second-order system, indicated by the s2 term in the denominator of the
transfer function.
When we looked a first-order systems, we found it useful to describe them in terms of a steady-state gain
K and a time constant . It was then easy to describe the behaviour of the system in terms of these
variables. For second-order systems, we can again express our transfer function in a standard form and
use a similar set of variables to describe the system.
The standard form of a second-order transfer function is:

K n2
Y (s)
= 2
X ( s ) s + 2 n s + n2
where K is again the steady-state gain, (zeta) is known as the damping ratio or damping factor and n is
known as the undamped natural frequency. As before, the reasons for these names will become apparent
as we consider the response of second-order systems to some standard inputs. Notice that in this standard
form the coefficient of the highest power of s is one.
If we compare the transfer function for our RLC circuit to the standard form, we can work out the values
of the standard variables in terms of L,R and C.
By looking at the constant term in the denominator, we can see that

n2 =
n =

1
LC
1
LC

Looking at the numerator then tells us that

K =1

Page 9 of 12

David Lee, Oct 2002

To calculate , look at the coefficient of s in the denominator

2 n = R / L

R
2 L n

R
LC
2L
R C
=
2 L
=

5 Response of Second-Order Systems


5.1

Introduction

We can adopt this same procedure for any second-order system: derive the transfer function and then
express it in standard form and determine the values of K, n and . We can then use the values of these
variables to describe the behaviour of the system with a variety of types of input. As before, we will look
at the response of the system to step, ramp, impulse and sinusoidal inputs.
For second-order systems, I will present the results without formal mathematical proof. If you are
interested in more detail, the course text can provide them.

5.2

Step Input

This graph shows the response of a unit-gain second-order system to a unit step input for a range of
values of .

Notice that values of less than 1 lead the system to overshoot and oscillate before settling on its steadystate value. Such systems are known as underdamped.
If is equal to 1, the system will not oscillate, but will reach its steady-state value without overshoot. This
is known as being critically-damped.
For values of larger than 1, the system doesnt overshoot but takes longer to reach its steady-state value
(similar to the effect of increasing for a first-order system). Such systems are known as overdamped.
Page 10 of 12

David Lee, Oct 2002

In practice, designers often choose to allow a small amount of overshoot in order to get a quick response
of the system. A value of close to 0.7 gives a reasonable compromise.
For underdamped systems, the value of also affects the frequency of oscillation. As approaches zero,
the oscillation frequency increases until approaches the undamped natural frequency n.

5.3

Ramp Input

For a ramp input, the behaviour of a second order system is influenced in a similar way by the damping
ratio . For equal to 1, the system settles into its steady time lag without overshoot. Smaller and larger
values of lead to overshoot and slower response respectively. Notice that larger values of also increase
the amount by which the output lags behind the input.

5.4

Impulse Input

As we would expect for an impulse input, the system settles to a steady state of zero after its initial
transient behaviour. The value of influences the damping as we would expect after looking at the step
and ramp inputs.

Page 11 of 12

David Lee, Oct 2002

5.5

Sinusoidal Input

To determine the response of the system to a sinusoidal input, we can again calculate G(j) from our
transfer function G(s) and express the result in terms of a magnitude and a phase angle. This Bode plot
shows the result for a range of values of . Notice that the frequency is normalized by expressing it as a
multiple of the undamped natural frequency n.

1
2

There are several points to note about this plot:

For highly underdamped systems ( much less than 1), the magnitude of the response shows a
marked peak when approaches the undamped natural frequency n.

For overdamped systems (>1), the magnitude of the response shows a rapid fall-off at higher
frequencies.

The phase lag starts close to zero for low frequencies and passes through 90 degrees at n,
approaching 180 degrees for higher frequencies. The transition through 90 degrees is more abrupt
for smaller values of .

Page 12 of 12

David Lee, Oct 2002

Das könnte Ihnen auch gefallen