Performance Specifications
Dr. Kevin Craig
Professor of Mechanical Engineering
Rensselaer Polytechnic Institute
Performance Specifications
K. Craig
Performance Specifications
of Closed-Loop Systems
Basic Considerations
Time-Domain Performance Specifications
Frequency-Domain Performance
Specifications
Performance Specifications
K. Craig
Basic Considerations
Most of our discussion will involve rather specific
mathematical performance criteria whereas the ultimate
success of a controlled process generally rests on
economic considerations which are difficult to calculate.
This rather nebulous connection between the technical
criteria used for system design and the overall economic
performance of the manufacturing unit results in the
need for much exercise of judgment and experience in
decision making at the higher management levels.
Performance Specifications
K. Craig
Control system designers must be cognizant of these
higher-level considerations but they usually employ
rather specific and relatively simple performance
criteria when evaluating their designs.
Performance Specifications
K. Craig
Control System Objective
C follow desired value V and ignore disturbance U
Technical performance criteria must have to do with
how well these two objectives are attained
Performance depends both on system characteristics
and the nature of V and U.
U
N(D)
Basic Linear Feedback System
V
R
A(D)
+
B
G 1 (D)
_
H(D)
Performance Specifications
G 2 (D)
Z(D)
K. Craig
The practical difficulty is that precise mathematical
functions for V and U will not generally be known in
practice.
Therefore the random nature of many practical
commands and disturbances makes difficult the
development of performance criteria based on the
actual V and U experienced by real system.
It is thus much more common to base performance
evaluation on system response to simple "standard"
inputs such as steps, ramps, and sine waves.
Performance Specifications
K. Craig
This approach has been successful for several reasons:
In many areas, experience with the actual
performance of various classes of control systems
has established a good correlation between the
response of systems to standard inputs and the
capability of the systems to accomplish their required
tasks.
Design is much concerned with comparison of
competitive systems. This comparison can often be
made nearly as well in terms of standard inputs as for
real inputs.
Performance Specifications
K. Craig
Simplicity of form of standard inputs facilitates
mathematical analysis and experimental
verifications.
For linear systems with constant coefficients,
theory shows that the response to a standard
input of frequency content adequate to exercise all
significant system dynamics can then be used to
find mathematically the response to any form of
input.
Performance Specifications
K. Craig
Standard performance criteria may be classified as
falling into two categories:
Time-Domain Specifications: Response to steps,
ramps, and the like
Frequency-Domain Specifications: Concerned
with certain characteristics of the system
frequency response
Both time-domain and frequency-domain design
criteria generally are intended to specify one or the
other of:
speed of response
relative stability
steady-state errors
Performance Specifications
K. Craig
Both types of specifications are often applied to the
same system to ensure that certain behavior
characteristics will be obtained.
All performance specifications are meaningless
unless the system is absolutely stable. So we
assume absolute stability for the remainder of this
discussion.
Performance Specifications
K. Craig
10
Time-Domain Performance Specifications
For linear systems, the superposition principle allows
us to consider response to commands apart from
response to disturbances.
If both occur simultaneously, the total response is just
the superposition of the two individual responses.
In nonlinear systems, such treatment with
subsequent superposition is not valid.
Performance Specifications
K. Craig
11
Rise time, Tr, and peak time, Tp, are speed of
response criteria.
Percent overshoot, Op = (O/V) 100, is a relative
stability criterion, with 10% - 20% as an acceptable
value.
Closed-Loop
Response
of C to a Step of V
when U = 0
Performance Specifications
K. Craig
12
Settling time, Ts, the time it takes for the response to
get and stay within a specified percentage, e.g., 5%,
of V, combines stability and speed of response
aspects.
The decay ratio, the ratio of the second overshoot
divided by the first, is a relative stability criterion used
most often in the process control industry, with 1/4 a
common design value.
Which System is
Faster?
A or B?
Performance Specifications
K. Craig
13
Certain math models of systems will predict, for given
commands or disturbances, steady-state errors that
are precisely zero, but no real system can achieve
this perfection.
Nonzero errors are always present because of
nonlinearities, measurement uncertainties, etc.
To determine the steady-state error set up the
closed-loop system differential equation in which
error (V-C) is the unknown. Solution of this equation
gives a transient solution that always decays to zero
for an absolutely stable system.
Performance Specifications
K. Craig
14
The remaining solution is, by definition, the steadystate error, whether it is itself steady or time varying.
That is, steady-state error need not be a constant
value.
The steady-state error, Ess, depends on both the
system and the input command or disturbance that
causes the error.
There is a certain pattern of behavior as the input is
made more difficult from the steady-state viewpoint.
This type of pattern can be expected for both
commands and disturbances in all linear systems,
though details will vary.
Performance Specifications
K. Craig
15
Effect of
Command Severity
on Steady-State
Error
Performance Specifications
K. Craig
16
Steady-State Error
D(s)
R(s)
+
E(s)
B(s)
+
G c (s)
G(s)
C(s)
H(s)
Assume H(s) = 1 and D(s) = 0. The error is then
E(s) which equals R(s) C(s).
E(s)
1
R(s)
E(s) =
=
1 + G c (s)G(s)
R(s) 1 + G c (s)G(s)
R(s)
ess (t) = lim sE(s) = lim s
s 0
s 0 1 + G (s)G(s)
c
Performance Specifications
K. Craig
17
The final value theorem, assuming closed-loop
stability, tells us that:
ess ( t ) = limsE(s)
s 0
We are interested in the steady-state error for step,
ramp, and parabolic inputs, i.e.,
R(s) =
1
s
n +1
n = 0, 1, 2
Therefore
1
ess ( t ) = lim n n
s0 s + s G (s)G (s)
1
2
Performance Specifications
K. Craig
18
Step Input: R(s) = 1/s
1
1
1
s
= lim
=
ess (t) = lim s
s 0 1 + G (s)G(s)
s 0 1 + G (s)G(s)
1+ Kp
c
c
K p lim G c (s)G(s)
s 0
Ramp Input: R(s) = 1/s2
Static Error Constants:
Kp and Kv
1
2
1
s
= lim
ess (t) = lim s
s 0 1 + G (s)G(s)
s 0 s + sG (s)G(s)
c
c
1
1
= lim
=
s 0 sG (s)G(s)
Kv
c
Performance Specifications
K v lim sG c (s)G(s)
s 0
K. Craig
19
System type is the order of the input polynomial that
the closed-loop system can track with finite error.
For example, if G1(s)G2(s) has no poles at the origin,
the closed-loop is a Type 0 system and can track a
constant with finite steady-state error. A Type 1
system (one pole at the origin) can track a constant
with zero error and a ramp with finite error. A Type 2
system (two poles at the origin) can track both a
constant and ramp with zero error and a parabola
with finite error.
Performance Specifications
K. Craig
20
When system input is a disturbance U (V=0) some of
these criteria can still be applied, although others
cannot.
It is still possible to define a peak time Tp, however Tr,
Ts, and Op are all referenced to step size V, which is
now zero, thus they cannot be used.
One possibility is to use peak value Cp as a reference
value to define Tr and Ts.
To replace Op as a stability specification one could use
the decay ratio defined earlier or perhaps the number of
cycles to damp the amplitude to say, 10% of Cp. The
smaller the number of cycles, the better the stability.
Definition of steady-state error still applies and we
would again expect the same trend of worsening error
as U changed from step to ramp to parabola.
Performance Specifications
K. Craig
21
Time-Domain
Performance
Specifications
for a Disturbance
Input
Performance Specifications
K. Craig
22
Frequency Response Performance
Specifications
Let V be a sine wave (U = 0) and wait for transients
to die out.
Every signal will be a sine wave of the same
frequency. We can then speak of amplitude ratios
and phase angles between various pairs of signals.
C
AG1G 2 (i)
(i) =
V
1 + G1G 2 H(i)
Performance Specifications
K. Craig
23
The most important pair involves V and C. Ideally
(C/V)(iw) = 1.0 for all frequencies.
Amplitude ratio and phase angle will approximate the
ideal values of 1.0 and 0 degrees for some range of
low frequencies, but will deviate at higher
frequencies.
Performance Specifications
K. Craig
24
Typical Closed-Loop
Frequency Response
Curves
Performance Specifications
K. Craig
25
The frequency at which a resonant peak occurs, r, is
a speed of response criterion. The higher r, the
faster the system response.
The peak amplitude ratio, Mp, is a relative stability
criterion. The higher the peak, the poorer the relative
stability. If no specific requirements are pushing the
designer in one direction or the other, Mp = 1.3 is
often used as a compromise between speed and
stability.
For systems that exhibit no peak, the bandwidth is
used for a speed of response specification. The
bandwidth is the frequency at which the amplitude
ratio has dropped to 0.707 times its zero-frequency
value. It can of course be specified even if there is a
peak.
Performance Specifications
K. Craig
26
If we set V = 0 and let U be a sine wave, we can
measure or calculate (C/U)(i) which should ideally
be 0 for all frequencies. A real system cannot
achieve this perfection but will behave typically as
shown.
Closed-Loop Frequency Response to a Disturbance Input
Performance Specifications
K. Craig
27
Two open-loop performance criteria in common use
to specify relative stability are gain margin and phase
margin.
The open-loop frequency response is defined as
(B/E)(i). One could open the loop by removing the
summing junction at R, B, E and just input a sine
wave at E and measure the response at B. This is
valid since (B/E)(i) = G1G2H(i). Open-loop
experimental testing has the advantage that openloop systems are rarely absolutely unstable, thus
there is little danger of starting up an untried
apparatus and having destructive oscillations occur
before it can be safely shut down.
The utility of open-loop frequency-response rests on
the Nyquist stability criterion.
Performance Specifications
K. Craig
28
Gain margin (GM) and phase margin (PM) are in the
nature of safety factors such that (B/E)(i) stays far
enough away from 1 -180 on the stable side.
Gain margin is the multiplying factor by which the
steady state gain of (B/E)(i) could be increased
(nothing else in (B/E)(i) being changed) so as to put
the system on the edge of instability, i.e., (B/E)(i))
passes exactly through the -1 point. This is called
marginal stability.
Phase margin is the number of degrees of additional
phase lag (nothing else being changed) required to
create marginal stability.
Both a good gain margin and a good phase margin
are needed; neither is sufficient by itself.
Performance Specifications
K. Craig
29
Open-Loop Performance Criteria:
Gain Margin and Phase Margin
A system must have adequate stability margins.
Both a good gain margin and a good phase margin are needed.
Useful lower bounds:
GM > 2.5 PM > 30
Performance Specifications
K. Craig
30
Bode Plot View of
Gain Margin and Phase Margin
Performance Specifications
K. Craig
31
It is important to realize that, because of model
uncertainties, it is not merely sufficient for a system to
be stable, but rather it must have adequate stability
margins.
Stable systems with low stability margins work only
on paper; when implemented in real time, they are
frequently unstable.
The way uncertainty has been quantified in classical
control is to assume that either gain changes or
phase changes occur. Typically, systems are
destabilized when either gain exceeds certain limits
or if there is too much phase lag (i.e., negative phase
associated with unmodeled poles or time delays).
As we have seen these tolerances of gain or phase
uncertainty are the gain margin and phase margin.
Performance Specifications
K. Craig
32
Consider the following design problem: Given a plant
transfer function G2(s), find a compensator transfer
function G1(s) which yields the following:
stable closed-loop system
good command following
good disturbance rejection
insensitivity of command following to modeling
errors (performance robustness)
stability robustness with unmodeled dynamics
sensor noise rejection
Performance Specifications
K. Craig
33
Without closed-loop stability, a discussion of performance
is meaningless. It is critically important to realize that the
compensator G1(s) is actually designed to stabilize a
nominal open-loop plant G 2 (s) . Unfortunately, the true
plant is different from the nominal plant due to unavoidable
modeling errors, denoted by G2(s). Thus the true plant
may be represented by
G 2 (s) = G 2 (s) + G 2 (s)
Knowledge of G2(s) should influence the design of G1(s).
We assume here that the actual closed-loop system,
represented by the true closed-loop transfer function is
absolutely stable.
G1 (s) G 2 (s) + G 2 (s)
(unity feedback assumed)
Performance Specifications
1 + G1 (s) G 2 (s) + G 2 (s)
K. Craig
34
Desired Shape for Open-Loop
Transfer Function
Smooth transition from the low to highfrequency range, i.e., -20 dB/decade slope near
the gain crossover frequency
Gain below this
level at high
frequencies
Gain above
this level at
low
frequencies
Frequencies for good
command following,
disturbance reduction,
sensitivity reduction
Performance Specifications
Sensor noise,
unmodeled highfrequency dynamics
are significant here.
K. Craig
35
