Systems & Control: S C S H S R S KG S C

SYSTEMS & CONTROL
Basic Control Concepts
Closing the loop

In order to control a process we have to
1. Model the system dynamics of G(s)
2. Measure the output we want to control and try and regulate it in
some way.
A general control system has the following:
R(s)
E(s)
K
G(s)
C(s)
Cm(s)
H(s)
Here:
- G(s) is the plant or process that we are trying to control
- H(s) is the measurement system
- R(s) is the reference/regulated input
- C(s) is the controlled output
- K is the controller gain ( this may be within the plant G(s) )
We can show that:
C ( s ) = KG ( s ) E ( s )
E ( s ) = R( s ) C m ( s)
C m ( s) = H ( s )C ( s )
Combining these in terms of C(s) gives:
C ( s ) = KG ( s )[ R ( s ) H ( s )C ( s )]
Created By Dr Grant Covic for paper: Systems & Control 1998: Document 1 of 3
resulting in the closed loop transfer function:

C (s)
KG ( s )
=
R ( s ) 1 + KG ( s ) H ( s )
Nb: KG(s)H(s) must be dimensionless
For unity feedback systems H(s) = 1, thus:

C ( s)
KG ( s )
=
R( s ) 1 + KG ( s )
In many systems the dynamics of H(s) are much faster than that of
G(s) so that Cm(s) is a good measure of the output C(s). This is true
for most applications: ie H(s) may simply be a filter. In such cases
the system may be thought of as a unity feedback system as shown
below:
R(s)
E(s)
+
Cm(s)
G(s)H(s)
Why do we need closed loop control?

Open loop systems:
U(s)
G(s)
Y(s)
To determine the steady state output, we can use the final value
theorem for Laplace transforms:
lim f (t ) = lim sF ( s )
t
s 0
Ex:
G (s) =
b
s+a
y (t ) ss = lim s[ KG ( s )U ( s )]
s 0
If U(s) is a unit step input (1/s) then:

y (t ) ss =
Kb
a so that the final value of the output will grow with K.
A problem is that the plant dynamics may change from what we

expect due to uncertainty. For instance parameter changes due to
temperature (etc) may change the values of b and a. Under such
circumstances, increasing K simply amplifies the uncertainty in the
output, and will contribute to output error.
Closed loop systems:
Here the feedback is used to help remove this uncertainty in the
response of the output. Consequently we can use feedback control
to force the output to track a controlled input.
Ex: If we have a unity feedback system as K increases:
C (s)
KG ( s)
KG ( s)
=
1
Kincrea
sin
g
R ( s) 1 + KG ( s)
KG ( s)
Hence, changes in the plant G(s) have much less effect.

In practice we have made the system less sensitive to plant variations
by the factor (1+KG(s)).
We can define the closed loop system sensitivity (unity feedback) as:
S = (1 + KG ( s )) 1
What sort of control is necessary?

There are two approaches to controlling a plant (chap 7)
Servo Control: Here we are interested in the performance of the
system when it has to follow changes in the reference or demand
signal. Thus the control system is designed based on transient and
steady state performance objectives in the time and frequency
domain. Applications include torque and variable speed controlled
drives, vehicles, cranes etc.
Process Control: Here we are interested in the performance of the
loop as a regulator. The input is generally a set point and the
system is designed to try and avoid the effects of disturbances.
Applications include petrochemical, food, steel, glass, pulp&paper
and energy systems.
In both control systems the difference between the reference input
R(s) and the feedback signal Cm(s) is important. This signal is called
the error signal; E(s).
In a general system the error signal can be defined as:
E ( s ) = R ( s ) Cm ( s ) = R ( s ) C ( s ) H ( s )
E ( S ) = R ( s ) E ( s ) KG ( s ) H ( s )
resulting in:
E (s) =
R( s)
1 + KG ( s ) H ( s )
Ideally we want e(t) 0 under both transient and steady state

conditions.
Time Domain Performance Considerations

In a purely open loop system we can look at the performance and
stability of the system by looking at the position of the fixed poles in
the plant and apply our time domain performance specifications
Example:
G(s) =
1
s+a
For a step input we have: y (t ) =
K
1 e at
a
With variations in K we can only vary the steady state output, not the
time constant of the system, and as discussed previously we are
dependent on the plant dynamics and stability for good performance.
Figure 3.8 (Golten & Verwer)
G( s) =
1.2
s +1
and G ( s) =
6
( s + 1)( s + 5)
General Systems
Most complex systems have dominant features that typically can be
approximated by either a first or second order system response.
Thus an understanding of both first and second order system
responses is very useful.
Dominant First order systems
First order systems have monotonic step response. The response to a
unit step input can be expressed as:
y (t ) = 1 e
The response time is often redefined for first order systems as time to
rise to 63% of the final value (tr = )
A 5% settling time
results in: ts = ln(0.05) 3
Dominant Second order systems

The step response of a stable second order system (ie underdamped)
is defined by the pole locations in the s-plane resulting from a
complex pole pair.
Since the following is true in the s-domain:
Fig 4.3 Dazzo & Houpis (Linear Control System Analysis & Design)
k
( s + a + jb)(s + a jb)
k
= 2
s + 2as + (a 2 + b 2 )
G( s) =
K n2
= 2
s + 2 n s + n2
Where:
n: (undamped natural frequency) is the distance from the origin.
d: (damped natural frequency) is the distance along the imaginary axis.

: (the rate of decay) is the distance along the real axis
: (the damping factor) Cosine(angle from origin to the complex pole)
For such a system the transient response to a step input (Golten &
Verwer 3.7 & 3.8) can be written as:
y ( t ) = Re n t cos( d t + )
The time to first peak is half the oscillation period (ie the damped
natural frequency d) and the rise time is approximately half of this.
hence
tp =
1 2
=
2t r
2 d d
The decay ratio can be determined from the ratio of the times to two
positive/negative peaks mk+2/mk:
Decay ratio =
2 / 1 2
The absolute exponent of this is called the logarithmic decrement

The percentage overshoot is also given by the above expression for
the ratio of consecutive peaks. The first overshoot peaks m1 units
above the steady state value hence:
% overshoot = 100
m1
m 0 =100
/ 1 2
rearranging for damping factor:

=
ln(%O.S .)
2 + ln (%O.S .)
2 2
ex: if we want no more than 18% O.S.

if 10% overshoot is acceptable then
0.479 ( 0.5)
0.591 ( 0.6)
The settling time is mainly governed by the time for the exponential
envelope to decay to within a set tolerance band. This is directly
related to the exponential decay in the transient response:
=e
n t
Similar to the first order system the 5% settling time is:
3
3
=
ts 3 =
n
Design Regions on the s-plane using Time Domain Performance

Specifications
The step performance criteria are related to the position of the poles
on the s-plane.
Intuitively, we know that a best damping ratio will exist for any
second order system, since lightly damped systems will overshoot
(possibly many times) before settling and heavily damped systems
will be sluggish.
The optimum damping ratio for a given settling time is the minimum
damping ratio which can be used without the overshoot exceeding
the tolerance limit. eg the 5% settling time is optimised when the %
overshoot = 5%, giving a damping ratio of 0.69.
The percentage overshoot or decay ratio must generally be required
to be within some specified limits. For our second order system this
corresponds to poles located between two lines of constant damping
ratio (figure 3.13)
The settling time (as determined by , the real part of the poles) may
have a maximum desirable value in which case the system poles
must lie to the left of a vertical line on the s-plane (representing this
maximum - figure 3.13).
The rise time depends on the damped natural frequency d, thus if
this time is to be limited the upper half plane pole must lie above a
horizontal line (figure 3.13).
Performance Specs Applied to Closed Loop Systems

In a closed loop system the position of the new system poles now
change as K varies. This affects both the transient and steady state
response in the time and frequency domains, so that the system can
be made to respond much faster, but if were not careful we could
make it poorly damped or even go unstable.
Because of this the system performance and stability must be
investigated with changing K. Usually by looking at how the new
poles vary with K in the S-plane.
Nyquist/Bode analysis techniques are also used to obtain an overall
picture of system behaviour in the frequency domain, however we
cant measure a frequency response if the system is already unstable,
whereas we can determine this by looking to see if any poles exist in
the right half plane (RHP) of the S-plane.
The poles of the closed loop system are now defined by:
1 + KG(s)H(s) = 0
This is called the systems Characteristic Equation.
The poles of C(s)/R(s) are the roots of this equation.
Examples:
1. A First order system
Using the original plant with unity feedback and series gain K:
G ( s) =
1
s+a
K
C ( s)
K
CLTF =
= s+a =
R( s) 1 + K
s + ( K + a)
s+a
the root locus:
the time response:

j
K
1 e ( K + a )t
K +a
c (t ) =
K=0
K increasing
-a
The time response
Practical Systems
In a more practical system we have disturbances and measurement
inaccuracies (noise):
D(s)
R(s)
E(s)
+
U(s)
G(s)
C(s)
Cm(s)
+
M(s)
we can define the closed loop transfer function (complementary

sensitivity) as:
T=
KG ( s )
1 + KG ( s )
Thus if we want:
C (s) = R(s)
T 1 K must be Large
to keep the effect of M small: T 0 K must be small
to keep the effect of D small: S K must be large
As the control effort to the plant is:

U (s) =
K
R(s)
1 + KG ( s )
A large K results in large control effort to the plant. In practice this

may exceed the available drive energy, causing actuators to limit
(saturate), thereby possibility exciting non-linear modes of
behaviour.
2. A Second Order System:

A seperately excited DC servo drive with proportional control
If
Vf
Rf
Ra
Lf
La
m= d/dt
Va
Ia
J
B
As the time constant of the armature winding (La/Ra) is small, then:

Ia
Va
Ra
and using the torque equation with Te = K t I a

K t / Ra
Km
Km /
( s)
=
=
V (s) s( Js + B) s(s + 1) s( s + 1 / )
a
Va
d
d 2
Kt
=J 2 +B
dt
Ra
dt
If we now add a proportional controller then:

R(s) = ref
E(s)
+
U(s)
Km
s(s + 1 )
C(s) =
K=0
K=0
Nb: This second order system is unconditionally stable with K,

however damping becomes very poor. Thus K must be selected to
give a desirable damped response.
Higher Order Systems

In higher order closed loop systems, the possibility of instability
grows. Hence we must determine bounds on K.
eg:
R(s)
E(s)
+
1
s (1 + sT1 )(1 + sT2 )
C(s)
CLTF =
The characteristic equation is:
K
s (1 + sT1 )(1 + sT2 ) + K
T1T2 s 3 + (T1 + T2 ) s 2 + s + K = 0
For stability, a necessary but not sufficient condition is that all the
terms of the characteristic equation have the same sign and all
consecutive s terms be non-zero.
With varying K the complex plane gives a better picture of the
stability of the system.
Identifying the Closed Loop Systems Dominant Poles

(Chap 3.11 - Golten & Verwer)
We must eventually fix K (and therefore the position of the closed
loop poles). To do this we need to identify those poles that have the
dominant effect on the time domain response.
Fixed Poles
Assume K has already been fixed resulting in the following CLTF:
K g ( s + z1 )( s + z 2 )...
C (s)
= G( s ) =
( s + p1 )( s + p2 )( s + p3 )...
R(s)
z1, z2 ... are the zeros of the CLTF,

p1, p2 ... are the poles of the CLTF.
The positions of p1, p2, p3 ... in the complex plane are most important
not only because they determine the stability of the system, but also
because they determine the form of the time response of the output
c(t) to an input r(t).
eg. Suppose r(t) is a unit step function u(t) [ie R(s) =1/s]
using partial fractions we obtain:
C ( s) =
A
B
C
D
+
+
+
+ ...
s s + p1 s + p2 s + p3
and taking the inverse Laplace Transform results in:

c(t ) = Au (t ) + Be p1t + Ce p 2 t + ...
A,B,C can be determined for a known system, however we want to
control the values of the p1, p2, p3 ... in order to optimise the time
response of the closed loop system to give a well damped fast
response with transient terms which quickly die away.
Fixing K using the locus of the Dominant System Poles

The CLTF poles are functions of K, therefore we need to try and
identify the dominant root locus branches that will have the largest
influence on the time response.
The usual objective when designing a system is to try and fix our
dominant closed loop pole positions such that the transient time
response approximates that of a second order system, with a pair of
complex poles that are suitably damped. If we can do this then the
transient response to a step input will approximate:
C ( t ) D ( t ) + Ee t sin( t + )
If there is one complex pair locus that will be dominant for all
relevant values of K then the above objective is easily possible and
we can simply use the relationshiips between the second order time
domain specifications and the pole locations to fix these poles
somewhere on that locus that gives acceptable damping, response
and settling times (by fixing K).
The necessary conditions for the time response of a high order closed
loop system to be dominated by one pair of complex poles requires a
pole-zero pattern (with K fixed) that has the following
characteristics:
- All other poles and any zeros must be far to the left of the
dominant poles, so that the transients due to these are small in
amplitude and die out quickly
- Any pole which is not far to the left of the dominant complex
poles must be near a zero so that the magnitude of the transient
term due to that pole is small.
Ex:
As shown in the above figure, peak overshoot (Mp), risetime (tr),

settling time (ts), damping () etc, can all be approximated from the
dominant poles positions as for a second order system.
Multiple Complex Poles
In order to identify the effect of several complex poles, the damping
ratio should be taken into account. The distance from the imaginary
axis to a pole is called the pole attenuation. This attenuation can be
used to determine which poles have a lasting effect, however the
radius from the centre point is usually a better guide to the pole(s)
overall dominance. Poles which are at a radius of more than three
times that of the dominant pole(s) will have only a small
contribution.
When modelling higher order systems by the dominant features,
poles with negligible contribution can be neglected providing their
steady state gain contributions are retained.
Assume we have a higher order system that has the following fixed
set of poles:
G ( s ) =
100
( s 2 + s + 1)( s 2 + s + 9)( s + 3)
then the system poles are located at:
s = 0. 5 j 0. 886, 0 . 5 j 2. 96, and 3

Dominant poles have smallest radius.

- the first complex pair is dominant,
- all other poles have a radius > 3 times that of the dominant poles
- we can simplify the system model by approximating non-dominat
poles with their steady state gain only.
The simplified model is therefore:
G( s) =
100
9 3( s 2 + s + 1)
Mixed Dominant Systems (Golten & Verwer 3.11)

If there is not a single pole or pole pair which is dominant then the
response will be a mix of the individual pole contributions and will
always be slower than each poles attenuation suggests.
Pole-Pole Mix
A dominant mix of a complex pole pair with a single real pole causes the
system response to appear as a filtered second order response (Fig 3.18).
The extent of this filtering depends on how close the real pole is to the j
axis.

Pole-Zero Mixes
The effect of adding a dominant zero to a system (ie, a zero with a small
radius) is equivalent to modifying the input function as shown in Fig
3.20. Here a step appears to receive an additional "kick", as the zero
contributes an impulse with a weighting determined by its attenuation
(recall derivative action). If the zero is in the right half plane this "kick"
will be in the opposite direction to the input function and initially the
system response will tend to move in the opposite direction to its final
steady state value.

Control System Error in Unity Feedback Systems

Not only are we interested in the transient performance of practical
feedback systems, but we also want to know how good their system
response is after the transient has died away. ie their steady state
response to the reference input. Three standard inputs, which
approximate most command inputs, are:
- unit step (1/s)
- unit ramp (1/s2)
- unit acceleration. (1/s3)
We will consider the steady state error response of several common
stable system types.
Let the open loop transfer function take on the general form:
G ( s) =
K (1 + sT1 )(1 + sT2 ) K (1 + sTm )

s n (Ta s 2 + Tb s + 1)(1 + sTc ) K (1 + sT p )
where
sn = multiple pole at the origin of the complex plane

n = system type number = no of integrations in the transfer function.
In a unity feedback system:

E (s) =
R(s)
1 + G (s)
&
e(t ) ss = lim e(t )

t
Thus:
s n [(Ta s 2 + Tb s + 1)(1 + sTc ) K (1 + sT p )]R ( s )

e(t ) ss = lim s n
2
s0
s [(Ta s + Tb s + 1)(1 + sTc ) K (1 + sT p )] + K [(1 + sT1 )(1 + sT2 ) K (1 + sTm )]
The most common (practical) system types are type: 0, 1 and 2.

Unit Step Input Response

From our definition of e(t)ss then:
1
s (1 / s)
=
s 0 1 + KG ( s )
1 + lim KG ( s)
s0
e(t ) ss = lim
Thus if we define a step (position) co-efficient as:

K p = lim KG ( s )
s 0
then:
e(t ) ss =
1
1+ K p
The response of each system type to this input is summarised below:

System Type No
e(t)ss
Kp
Unit Ramp Input Response

1
s (1 / s 2 )
= lim
e(t ) ss = lim
s0 s ( KG ( s ))
s 0 1 + KG ( s )
if we define a ramp (velocity) co-efficient as:

K v = lim s ( KG ( s ))
s 0
thus:
e(t ) ss =
1
Kv
The system response follows as:

System Type No
e(t)ss
Kv
Unit Acceleration Input Response

1
s(1 / s 3 )
= lim 2
e(t ) ss = lim
s0 s ( KG ( s ))
s 0 1 + KG ( s )
We define a parabolic (acceleration) co-efficient as:

K a = lim s 2 ( KG ( s ))
s 0
thus:
e(t ) ss =
1
Ka
The system response follows is:

System Type No
Ka
e(t)ss
Example of error calculation for a Type 1 system:

(a) with step input

(b) with ramp input
(c) with acceleration input


Systems & Control: S C S H S R S KG S C

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Systems & Control: S C S H S R S KG S C

Uploaded by

Copyright:

Available Formats

SYSTEMS & CONTROL

Basic Control Concepts

Closing the loop

We can show that:

resulting in the closed loop transfer function:

For unity feedback systems H(s) = 1, thus:

Why do we need closed loop control?

If U(s) is a unit step input (1/s) then:

A problem is that the plant dynamics may change from what we

Hence, changes in the plant G(s) have much less effect.

What sort of control is necessary?

Ideally we want e(t) 0 under both transient and steady state

Time Domain Performance Considerations

For a step input we have: y (t ) =

Figure 3.8 (Golten & Verwer)

Dominant Second order systems

d: (damped natural frequency) is the distance along the imaginary axis.

The absolute exponent of this is called the logarithmic decrement

rearranging for damping factor:

ex: if we want no more than 18% O.S.

Similar to the first order system the 5% settling time is:

Design Regions on the s-plane using Time Domain Performance

Figure 3.13 (Golten & Verwer)

Performance Specs Applied to Closed Loop Systems

the root locus:

the time response:

The time response

we can define the closed loop transfer function (complementary

As the control effort to the plant is:

A large K results in large control effort to the plant. In practice this

2. A Second Order System:

As the time constant of the armature winding (La/Ra) is small, then:

and using the torque equation with Te = K t I a

If we now add a proportional controller then:

Nb: This second order system is unconditionally stable with K,

Higher Order Systems

The characteristic equation is:

Identifying the Closed Loop Systems Dominant Poles

z1, z2 ... are the zeros of the CLTF,

and taking the inverse Laplace Transform results in:

Fixing K using the locus of the Dominant System Poles

As shown in the above figure, peak overshoot (Mp), risetime (tr),

then the system poles are located at:

s = 0. 5 j 0. 886, 0 . 5 j 2. 96, and 3

Dominant poles have smallest radius.

Figure 3.19 (Golten & Verwer)

Mixed Dominant Systems (Golten & Verwer 3.11)

Figure 3.11 (Golten & Verwer)

Figure 3.17 (Golten & Verwer)

Figure 3.20 (Golten & Verwer)

Figure 3.21 (Golten & Verwer)

Control System Error in Unity Feedback Systems

K (1 + sT1 )(1 + sT2 ) K (1 + sTm )

sn = multiple pole at the origin of the complex plane

In a unity feedback system:

e(t ) ss = lim e(t )

s n [(Ta s 2 + Tb s + 1)(1 + sTc ) K (1 + sT p )]R ( s )

The most common (practical) system types are type: 0, 1 and 2.

Unit Step Input Response

Thus if we define a step (position) co-efficient as:

The response of each system type to this input is summarised below:

Unit Ramp Input Response

if we define a ramp (velocity) co-efficient as: