Chapter 13

Frequency Response
Analysis

TABLE OF CONTENTS
13.1 Sinusoidal Forcing oí a Fírst-Order Proccss
13.2 Sinusoidal Forciog of an nth-Order Prccess
13.2.1 Shortcut Method for Fioding the Frequency Response
13.3 Bode Diagrams
13.3.1 Flrst-Order Process
13.3.2 lntcgrating Process
13.3.3 Second-Order Process
13.3.4 Process Zcro
13.3.S Time Dclay
13.4 Frequency Response Characferistics oí Feedback Controllers
13.5 Nyquist Dlagrams
Summary

In previous chapters, Laplace transform techniques were used to calculate transient responses from
transfer functions. This chapter focuses on an altemative way to analyze dynamic systerns, narnely, fre-
quency response analysis. We start with the response properties of a first-order process when forced by a
sinusoidal input and show how the output response characteristics depend on the frequency of the
input signa!. This is the origin of the term frequency response. Nex.t we introduce a simplified proce-
dure to ca1culate the frequency response characteristics from the transfer function of any linear
process. This procedure yields a powerful too! both for analyzing dynarnic systems and for designing
controllers. Computer methods for calculating and displaying frequencyresponse results are then dis-
cussed. These graphical representations form the basis for the frequency domain controller design
techniques of Chapter 14.
 13.1 Sinusoidal Forcing of a First-Order Process 335

13.1 SINUSOIDAL FORCING OF A FIRST-ORDER PROCESS

The responses for first- and second-order processes forced by a sinusoidal input were presented in
Chapter 5. Recall that these responses consisted of sine, cosine, and exponential terms. Specifically, for
a ñrst-order transfer function with gain K and time constant T, the response to a general sinusoidal
input, x(t) = A sin wt, is
KA . .
y(t) = wY + l (wu-tfr - WT cos wt + sm wt) (5-25)

Note that in (5-25) y is in deviation form.

If the sinusoidal input is continued for a long time, tlie exponential term becomes negligible. The re-
maining sine and cosine terms can be combined via a trigonometric identity to yield
Yt(t) = KA sin (wt + <ti) (13-1)
Jw'Lr2 + 1

where <!> = -tan-1(wT}. Toe long-time response ye(t) is called the frequency response of the first-order
system and has two distinctive features (see Fig. 13.1).
l. The output signa! is a sine wave that has the same frequency, but its phase is shifted relative to
the input sine wave by the angle <ti (referred to as the phase shift or the phase angley; the amount
of phase shift depends on the forcing frequency w. A

2. The output signal is a sine wave that has an amplitude A that also is a function of the forcing
'. '; frequency:

KA
A = �J;=2.r====+= (13-2)

Dividing both sides of (13-2) by tbe input signal amplitude A yields the amplitude ratio (AR)
AR=A= K (13-3a)
• A Jw2.r2 + 1

which can, in turn, be divided by the process gain to yield the normalized amplitude ratio (ARN)
ARN=AR = 1 (13-3b)
K jw'l.r2 +1

Because the process steady-state gain K is constant, the normalized amplitude ratio often is used for
frequency response analysis.
Next we examine the physical significance of the above equations, with specifi.c reference to the
blending process example discussed earlier. In Chapter 4 the transfer function model for the stirred-
tank blending system was derived as

X'(s)_ = TS�
1 Xí(s) + TS� 1 Wí(s) + Ts� 1 Wí(s) (4-69)

Figure 13.1 Attenuation and time shüt between input and

Time output sine waves (K = 1). Toe phase angle el> of the
shlft, 6t
output sigoal is given by el> = l!.tlP X 360º, where ót is the
Time, t time shift and Pis the period of oscillation.
:,:·.

336 Chapter 13 Frequency Response Analysis

Suppose flow rate w2 is varied sinusoidally about a constant value, while the other inlet conditions are
kept constant at their nominal values; that is, wL(t) = xí(t) = O. Because w2(t) is sinusoidal, the output
composition deviation x'(t) will eventually becorne sinusoidal according to Eq. 5-26. However, there
will be a phase shift in the output relative to the input as shown in Fig. 13.1, owing to the material
holdup of the tank. If the flow rate w2 oscillates very slowly relative to the residence timejr (w << lf'T),
the phase shift is very small, approaching Oº, while the normalized amplitude ratio (Al.KA) is very
nearly unity. For the case of a low-frequency input, the output is in phase with the input, tracking the
sinusoidal input as if the process model were G(s) = K.
On the other hand, suppose the flow rate is varied rapidly by increasing the input signa! frequency.
For eo >> lfr, Eq. 13-1 indicates that the phase shift approaches a value of -'TT/2 radians (-90º). The
presence of the negativo sign indicates that the output lags behind the input by 90º; in other words the
phase lag is 90°. The amplitude ratio approaches zero as the frequency becomes large, indicating that
the input signa! is almost completely attenuated, namely, the sinusoidal deviation in the output signal
is very small.
These results indicate that positive and negative deviations in wi are essentially canceled by the ca-
pacitance of the liquid in the blending system if the frequency is high enough. In this case, high fre-
quency implies w > > lfr. Most processes behave qualitatively like the stirred-tank blending system
when subjected to a sinusoidal input. For high-frequency input changes, the process output deviations
are so completely attenuated that the corresponding periodic variation in the output is difficult (per-
haps impossible) to detector measure.
Input-output phase shift and attenuation (or amplification) occur for any stable transfer func-
tion, regardless of its complexity. In ali cases, the phase shift and amplitude ratio are related to the
frequency eo of the sinusoidal input signal. In developments up to this point, the expressions for the
amplitude ratio and phase shift were derived using the process transfer function. However, the fre-
quency response of a process can also be obtained experirnentally. By performing a series of tests
in which a sinusoidal input is applied to the process, the resulting amplitude ratio and phase shift
can be measured for different frequencies. In this case, the frequency response is expressed as a
table of measured amplitude ratios and pbase shifts for selected values of w. However, the method
is very time-consuming because of tbe repeated experiments for different values of w. Thus, other
methods, such as pulse testing (Ogunnaike and Ray, 1994), are utilized because only a single test is
required.
In this chapter, the focus is on developing a powerful analytical method to calculate the frequency
response for any process transfer function, as sbown below. In Chapter 14 we show how this informa-
tion can be used to design controllers and analyze the properties of the controlled system responses.

13.2 SINUSOIDAL FORCING OF AN nTH-ORDER PROCESS

This section develops a general approach for deriving the frequency response of any stable transfer
function. We show that a rather simple procedure can be employed to find the sinusoidal response.
For a general transfer function G(s), rnultiplication by the Laplace transform of a sine wave input
with amplitude A and frequency w (see Table 3.1) gives
Aw
Y(s)= G(s)� (13-4)
s (J)

Suppose that the denominator of G(s) can be written as a product of n distinct factors (s + b1),
(s + b2), ... (s + bn) where b, can be real or imaginary and Re bi > O. Then Y(s) can be expressed as
ys _ N(s) Aw (13-5)
( ) - (s + b1)(s + b2) · · · (s + bn) �
where N(s) is the numerator polynomial. A partía! fraction expansion gives

Y(s) = � + � + ... + � + Cs +D (13-6)

o..l..J.. •-1-h� <+h. �
 13.2 Sinusoidal Forcing of an nth-Order Process 337

If we take the inverse Laplace transform of both sídes of (13-6), the first n terms become a sum of
exponential terms (possibly involving sorne damped sine and cosine expressions). Because Re bt > O,
all of the negative exponentials approach zero as t becomes large. On the other hand, the term
(Cs + D)/(s2 + w2) corresponds to C cos wt + (D!w) sin wt, which is unaffected by the length of time
that the sinusoidal forcing has been applied. In the sequel we focus on the long-time response, ignoring
the exponential terrns as in Eq. 13-1.
Toe Heaviside expansion (Chapter 3) can be employed to find the constants C and D; we do not
bother to calculate the a; because those terms approach zero and can be neglected for Jarge times.
Multiplying Y(s) by s2 + w2 and setting s = jw gives,
G(s)Awls=jw = ( Cs + D) ls=jw (13-7)

Then G(jw) = 1w (Cjw + D)

D .C (13-8)
=-+1-
wA A

If G(jw) is expressed as a complex number, G(jw) = R( w) + jl( w), we can equate the real and imagi-
nary parts of (13-7) and (13-8) with R + jl:
e =I
A (13-9a)
D
wA = R (13-9b)

Hence, one way to determine the coefficients C and D is to sets = jw in G(s) and then by algebraic
manipulation convert G(jw) into a complex number R + jl. Because the response for large values of
time is C cos wt + (D!w) sin wt, the coefficients of cos wt and sin wt are IA and RA, respectively.
Yt(t) = A(I cos wt + R sin wt) (13-10)

Using the trigonometric identity

sin (wt + <ji) = sin <f> cos wt + cos <ji sin wt

we can express the response in the alternative fonn
ye(t) = A sin (wt + <l>) (13-11)
where A and <f> are related to I(w) and R(w) by the following relations:
A =AJR2 +12 (13-12a)
<ji= tan-1(!/R) (13-llb)

Next we show that the preceding equations lead to a simple but elegant relation for the frequency
response. Toe polar form of the complex function, G(jw), is
G(jw) = IGleN = IGl(cos"1 + j sinl!i) (13-13)
In the complex plane IGI is the magnitude of G(jw) (also called the modulusy, and <ji is the angle of G(jw)
(or the argument). Equating the polar fonn to Eq. 13-12 shows that the amplitude ratio is given by

(13-14)
and the phase shift between the sinusoidal output and input is given by
<ji = LO= tan-l(JIR) (13-15)

Because R and I (and hence AR and <!>) can be found without calculating the complete transient re-
sponse y(t), these characteristics provide a shortcut method to determine the frequency response of
the first-order transfer function.
338 Chapter 13 Frequency Response Analysis

More importantly, Eqs. 13-14 and 13-15 provide a convenient technique for calculating the fre-
quency response characteristics of any stable G(s), including those with time-delay terms. These ex-
pressions are also valid and useful for unstable G(s). However, the physical interpretation of frequency
response is not valid for unstable systems because a sinusoidal input produces an unbounded output
response, instead of a sinusoidal response.

13.2.1 Shortcut Method for Finding the Frequency Response

Toe shortcut method consists of the following steps:

Step l. Sets= jw in G(s) to obtain G(jw).

Stcp 2. Rationalize G(jw): Express G(jw) as R + jl, where R and I are functions of w and possibly
model parameters, using complex conjugate multiplication: find the complex conjugate of
the denominator of G(jw) and multiply both nurnerator and denominator of G(jw) by this
quantity. A

Step 3. Toe output sine wave has amplitudeA = AJR2 + 12 and phase angle = tan-l(J/R). The am-
plitude ratio is AR = JR2 + ¡2 and is independent of the value of A.
Find the frequency response of a first-order system, with
,EXAMPLE 13.1
1
G(s)=- -
-ts + 1
(13-16)

SOLUTION First substítute s = jw in the transfer function

G(¡'w) = _l_ = _1_ (13-17)

TjW + 1 jWT + 1

Then multiply both numerator and denominator by the complex conjugate of the denorninator,
that is, -juYT + 1

G ·w = -jü>T + 1 = -i!!!7 + 1
(¡ ) (jWT + 1)(-jwT + 1) wlr2 + 1

1 . (-uYT) R 'I (13-18)

= + = +
w'lr2+T ¡� 1.
where

1
R = w2.r2 + 1 (13-19a)

and
-ü)T

I = w2-r2 +1 (13-19b)

From (13-12a)

Sirnplifying,
(l + w2r2) _ 1
AR= (13·20a)
(w2-r2 + 1)2 - ./w1.r2 + 1

<j> = LG(jw) = tan-1(-w-r) = -tan-l(WT) (13·20b)

 13.2 Sinusoidal Forcing of an nth-Order Process
339

lf the process gaín had been K instead of 1,

AR= K (13-21)
J,,,1,r1. + 1
and the phase angle would be unchanged (Eq. 13-20b). Both the amplitude ratio and
phase angle are identical to those values calculated in Section 13.l using Eq. 5-25.
•

From this example we conclude that direct analysis of the complex transfer function G(jw) is
com- putationally easier than solving for the actual long-time output response. The computational
advan- tages are even greater when dealing with more complicated processes, as shown below.
Start witb a general transfer function in factored form
G s _ Ga(s)Gb(s)Gc(s) · · · (13-22)
( ) - G1(s)G2(s)G3(s) · · ·
G(s) is cooverted to the complex form G(jw) by the substitution s = jw:
G(' ) = Ga(iw)Gb(jw)G,(jw) · · (13-23)
·
¡w G1(jw)G2(iw)G3(jw) · · ·
As a consequence ofEq. 13-13, we can express the magnitude and angle of G(jw) as
follows: (13-24a)
IG(jw)I = IGa(iw)l!Gb(jw)IIGc(jw)I·"
IG1(iw)JI Gi(jw)IIG3(jw)I · · ·
(13-24b)
LG(jw) = LG.(jw) + LGb(iw) + LGc(iw) + · · ·
- [LG1(iw) + LG2(jw) + LG3(jw) + · · ·]

Equations 13-24a and 13-24b greatly simplify the computation of IG(iw)I and LG(jw), and
consequently AR and 4>. These expressions eliminate much of the cornplex arithmetic associated
with the rationaliza- tion of complicated transfer functions. Hence, the factored forro (Eq. 13-22) is
preferred for frequency response analysis.

Calculare the amplitude ratio and phase angle for the overdamped second-order
transfer
function

Gs- K
() - (-r1s + 1)(-ris + 1)
SOLUTION Using Eq. 13-22,
Jet
Ga=K
G1 = TJS + 1
Gi=-rzs+l

Substitutíng s == jw
G0(jw) == K
G1(iw) = jwT¡ + 1
Gi(jw) = j(J)'f1. +1
The magnitudes and angles of each componen! of the complex transfer function
are:

LGa =O
IG1I = Jw2-r1 + 1 LG1 = tan-l((J)'f1)
I Gzl = Jw2-r1 + 1 LGi == tan-l(WT2)
340 Chapter 13 Frequency Response Analysis

Combining these expressions via Eqs. 13-24a and 13-24b yields

AR= IG(iw)I = 1<l<l!i2I

K (13-25a)
Jw2rf + 1 J1.J>'Lr1 + 1
<!> = LG(jw) = LGa - (LG1 + LG-z}
= -tan-1(1.J>T¡) - tan-1(úlT2) (13-25b)

•
13.3 BODE DIAGRAMS
A special graph, the Bode diagram or Bode plot, provides a convenient display of the frequency re-
sponse characteristics of a transfer function model in which AR and <!> are each plotted as a function of
w. Ordinarily, w is expressed in units of radians/time to simplify inverse tangent calculations (e.g., Eq.
13-15) where the arguments must be dimensionless, that is, in radians. Occasionally, a cyclic frequency, f
= 2-rr/w, with units of cycles/ time is used. Phase angle <!> is normally expressed in degrees rather than
radians. For reasons that will become apparent in the development below, the Bode diagram consists
of: (1) a log-log plot of AR versus w and (2) a semilog plot of <!> versus w. These plots are particularly
useful for rapid analysis of the response characteristics and stability of closed-loop systems.

13.3.1 First-Order Process

In the past, when frequency response plots had to be generated by hand, they were of lirnited utility.
A much more practical approach now utilizes spreadsheets or control-oriented software such as
MATLAB to simplify calculations and generate Bode plots. Table 13.1 illustrates the use of an Excel
spreadsheet program (Table 13.2) to calculate the frequency response of a first-order system. Al-

Table 13.1 Frequency Response of a First-Order Process Usiog Spreadsheet

Calculations: Numerical Results
G(s) = 5/(lOr + t)
Angle of Phase
Frequency Magnitude of !Os+ 1 Arnplitude Angle
(rad/nún) !Os+ 1 (rad) Ratio (deg)
0.001 1.00 0.01 5.00 -0.6
0.002 1.00 0.02 5.00 -1.l
0:004 1.00 0.04 5.00 -2.3
0.008 1.00 0.08 4.98 -4.6
0.01 1.00 0.10 4.98 -5.1
0.02 1.02 0.20 4.99 -11.3
0.04 1.08 0.38 4.64 -21.8
0.08 1.28 0.67 3.90 -38.7
0.1 1.41 0.79 3.54 -45.0
0.2 2.24 1.11 2.24 -63.4
0.4 4.12 1.33 1.21 -76.0
0.6 6.08 1.41 0.82 -80.5
0.8 8.06 1.45 0.62 -82.9
1 10.05 1.47 0.50 -84.3
2 20.02 1.52 0.25 -87.1
4 40.01 1.55 0.12 -88.6
8 80.01 1.56 0.06 -89.3
10 10!).00 1.56 0.05 -89.4
 13.3 Bode Diagrams 341

Table 13.2 Frequency Response oí a First-Order Process Using Spreadsheet Calculations: Cell Fonnulae
A B e D E
1
2
3 G(s) = 5/(lOs + 1}
4 Note log frequency scale
5
6
7 Gain 5
8 Time_Const 10
9
10 Magnitude of Angle of Amplilude Phase
11 Frequency lOs + 1 lOs + 1 Ratio Angle
12 (rad) (rad) (deg)
13
14 0.001 =SQRT(l+(Al4*Time_Const)"2) =ATAN(Al4*Time_Const) =Gain/814 =-C14*180/3.14159
15 0.002 =SQRT(l+(Al5*Time_Const)"2) =ATAN(A15*Time_Const) =Gain/B15 =-Cl5*180/3.14159
16 0.004 =SQRT(l+(Al6*Time_Const)"2) =ATAN(A16*Time_Const) =Gain/B16 = -C16*180/3.14159
17 0.008 =SQRT(l+(A17*Time_Const)A2) =ATAN(A17*Time_Const) =Gain/B17 =-C17*180/3.14159
18 0.01 =SQRT(l +(Al8*Time_Const)"2) =ATAN(A18*Time_Const) =Gain/818 =-C18*180/3.14159
19 0.02
20 0.04
21 0.08
22 0.1
23 0.2
24 0.4
25 0.6
26 0.8
27 1
28 2
29 4
30 8
31 10 =SQRT(l +(A31*Time_Const)"2) =ATAN(A31 •Time_Const) =Gain/B31 =-C31*180/3.14159
32

though spreadsheet software can be used to generate Bode plots, it is much more convenient to use
software designed specifically for control system analysis such as MATLAB. Thus, after describing the
qualitative features of a Bode plot of a first-order process (Fig. 13.2), we illustrate how the AR and <!>
components of such a plot are generated by a MATLAB program (Table 13.3).
Far a first-order process, Kl('rs + 1), Fig. 13.2 shows a log-log plot of the normalized amplitude ratio
versus orr, so that the figure applies for ali values of K and r. Also shown is a semiJog plot of <!> versus
342 Chapter 13 Frequency Response Analysis

"'b = lfr

Normalized
amplitude O. l 1:----f-----1------¡,.-----:1
ratio, ARN

úlT

-30
Phase angle

.¡, (dei,:l -60

-90
-]20'--'-LI..U..W.L-1.."-J.UWI--L..LJ.Ju.wJ.---L..LU.UW
0.01 0.1 10 100
Figure 13.2 Bode diagram for a first-order procesa.

orr. In Fig. 13.2 the abscíssa orr has units of radians. If K and r are known, ARN (or AR) and <I> can be
plotted as a function of w. Because of the variety of conventions, Bode plot labels should be carefully
used and observed.
Next we examine sorne properties of the Bode plot of the first-order systcm. At low frequencies,
(w << 111'), Eq. 13-3b reduces to: ·
ARN=l (AR= K) (13-26a)
<1>=0 (13-26b)

Table 13.3 MATLAB Program to Calculare and

Plot the Frequency Response of a First-Order
Process, G(s) = 5/(lOs + 1)
ww = logspace (-3, 2, 100); % Define
frequencies num = [5];
den= (101];
{mag, phase, ww] = bode (num, den, ww);
%
figure (1) subplot
(2, 1, l); loglog
(ww, mag);
axis ([0.001 10 0.1 lOJ);
tille ('Frequency Response for lst Order System')
ylabel ('AR')
%
subplot (2, 1, 2);
semilogx (ww,
phase); axis ((.00110
-90 OJ);
ylabel ('Phase Angle (degrees)')
xlabel ('Frequency (rad/s)')
%
 13.3 Bode Diagrams 343

Tbus, the amplitude ratio approaches the process gain, and tire phase shift becomes quite
small. At high frequencies, (w >> 1/.r), Eq.13-3b reduces to:
(13-
27a)
(13-
27b)
Here the amplitude ratio drops toan in.fi.n.ítesimal level, and the pitase lag (the phase shift expressed
as a positive value) approaches a máximum value of 90º. In Fig. 13.2 the low-frequency and
high- frequency asymptotes are:
Low frequencies: ARN =· 1 (13-28)
High frequencies: AAN = lfo>T (13-29)

Note that the asymptotes intersect at w = Wb = 1/.r, k.nown as the break frequency or comer
frequency.
From (13-21) and the definition ARN:
AR 1
ARN = - = -- = 0.707 for w = Wb (13-30)
K ji+1
An important feature of the log-log ARN plot for the first-order system is that the slope of the
high- frequency asymptote is .-1. This result follows from Eq. 3-27a.
log ARN = log 1 - log WT = -log (13-31)
orr
The phase angle always líes between O and -90º. The phase angle at
(13-32)
Wb is
<!>(w = Wb) = tan-1(-1) = -
45º

Sorne books and software define AR differently, in terms of decibels. The amplitude ratio in
decibels
ARdb is defi.ned as
ARdb = 20 log AR (13-
33) The use of decibels merely results in a rescaling of the Bode plot AR axis. The decibel
unit is ern-
ployed in electrical communication and acoustic theory and is seldom used in the process control
field. Notethat the MATLAB bode routine uses decibels as the default optíon; however, ít can be
modifi.ed to plot AR results as was done here. Also, if decibels are used, the AR asymptote slopes
discussed in this chapter and in later material need to be modified accordingly.
13.3.2 Integrating Process
The traosfer function for an integrating process was given in
Chapter 5:
w.\ K (5-34)
. ·'
: G(s)=�=-
U(s) s

Because of the single pote located at the origin, this transfer function represents a marginally
stable process. The shortcut method of determining frequency response outlined above was
developed for stable processes, that is, those that converge to a bounded oscillatory response.
Because the output of an integrating process is bounded when forced by a sinusoidal input, the
shortcut method does apply Ior this marginally stable process:

AR= IG(jw)I = 1 !< 1 = K (13-

34)
JW W

<!> = LG(jw) = LK - L(oo) = -90º

(13·35)
In a similar way, we can considera composite transfer function Gz consisting of a stable transfer
func- tion G1 and an integrator:
1
Gi(s) = - G1(s) (13·36)
s
344 Chapter 13 Frequency Response Analysis
When t is Iarge, the output response y(t) to a sinusoidal input, A sin wt, is:

y(t) = AIGi(jw)I sin (wt + <1>2) = A I G1(iw)I sin (wt + <1>1 - (13-37)
90º) w

1
Thus AR2 = -AR1 (13-38)
w
and <1>2 = <1>1 - 90º (13-39)

In general, the shortcut method of Section 13.2.1 can be used to evaluate the frequency response of any
marginally stable transfer function, that is, any stable transfer function containing a single integrator.
13.3.3 Second-Order Process
A general transfer function for a second-order system without numerator dynamics is

G(s) - K
+ nTS + 1
(13-40)
- T2S2

Substituting s = jw and rearranging into real and imaginary parts (see Example 13.1) yields

AR= K (13-41a)
J(l - wZ,.2)2 + (2tWT)2
� _ t
"' - an
-1 [ -2twT
1 - w2.r2
J (13-41b)

Note that in evaluating <f>, múltiple results are obtained because Eq. 13-41b has in.finitely many solu-
tions, each díffering by n180º, where nis a positive integer. The appropriate solution of (13-41b) for
the second-order system yields -180º < el> < O.
Figure 13.3 shows the Bode plots for overdamped (t > 1), critically damped (t = 1), and under-
damped (O < t < 1) processes as a function of orr, The low-frequency limits of the second-order

(
WT

el>
{deg) -90 (d:g) -90
-135t-----t---t---.CY"-k-----t -1351----lf----lM.''---t---i

0.1 10 100 0.1 10 100

WT

Figure 13.3 Bode diagrams far second-order processes. Right: underdamped. Left: overdamped and
critically damped.
 13.3 Bode Diagrams 345

o�������������������� Figure 13.4 Dependence of the nonnalized

�
o.o 0.2 0.4 0.6 0.8 amplitude ratio and resonant frequency on
t the damping coefficient t.

system are identical to those of the first-order system. However, the Iimits are different at high
frequencies
ARN = ll(W'T)2 (13-42a)
cp = -180° (13-42b)
The high-frequency AR asymptote has a slope of -2 in the log-log plot because
log ARN = log 1 - 2 log err
= -2log WT (13-43)

For overdamped systems, the normalized amplitude ratio is attenuated (A/KA < 1) for ali w. For un-
derdamped systems, the amplitude ratio plot exhibits a maximum (for values of O < t < Ji.12) at the
resonant frequency

w, = � (13·44)
'T

1
(ARN)max = ¡:,--;;; (13-45)
2t.;1 - t-
These expressíons can be derived by the interested reader. The resonant frequency w, is that frequency
for which the sinusoidal output response has the máximum amplitude for a given sinusoidal input. Fig-
ure 13.4 illustrates how w, and (ARN)mu depend on t. This behavior is used in designing organ pipes to
create sounds at specific frequencies. However, excessive resonance is undesirable, for example, in au-
tomobiles where a particular vibration is noticeable only at a certain speed. For industrial processes op-
erated without feedback control, resonance is seldom encountered, although some measurement
devices are designed to exhibit a lirnited amount of resonant behavior. On the other hand, feedback
controllers can be tuned to give the controlled process a slight amount of oscillatory or underdamped
behavior in order to speed up the controlled system response (see Chapter 12).
Find the frequency response and provide the Bode plot Ior the third-order transfer fuoction
. � 13.3
EXAMPLE. .. ·. below, where the time constanIs are in minutes

2
G(s) = (lOs + 1)(5s + l)(s + 1)
SOLUTION Application of Eq. 13-24 yields

AR= . K (13·46a)
J(10w)2
,
+ 1 J(5w)2 + 1 Jw2 + 1
I

(13·46b)
346 Chapter 13 Frequency Response Analysis

w (rad/min)

et,
(degl
-180����-t--����t---����

0.1 10
"' (rad/mln) Figure 13.5 Bode diagram for a third-order process.

Spreadsheet software can be used to evaluate the frequency response and provide appropriate plots. Al-
tematively, MATLAB can be used to generate a Bode plot that is similar to the one shown in Fig. 13.5.
Note in the plot that ARN approaches 1 at low frequencies. Toe slope of the asymptote for log AR
changes to - 1 at Wbt = 1/10, to -2 at Wb2 = 1/5, and to -3 at Wb3 = l. Note that as w - oo, the maximum
phase lag approaches 270º (<!> = -270"), ora - 90º contribution for each first-order transfer function. •

13.3.4 Process Zero

A term of the form -ts + 1 in the denominator of a transfer function is sometimes referred to as a
process lag because it causes the process output to lag the input (the phase angle contribution is nega-
tive). Similarly, process zero of the form TS + 1 (T > O) in the numerator (see Section 6.1) causes the si-
nusoidal output of the process to lead the input; hence, a left-half plane (LHP) zero often is referred to
as process /ead. Next we consider the amplitude ratio and phase angle for such a term.
Substituting s = jw into G(s) = TS + 1 gives
G(jw) = j= +1 (13·47)
from which
AR= jG(jw)I = Jw'Lr2 + 1 (13·48a)
q> = LG(jw) = +tan-l(WT) (13,48b)

Therefore, a process zero contributes a positive phase angle that varíes between O and +90º. The am-
plitude ratio has a high-frequency asymptote
AR=wT (13-49)
that is, an upward slope of + 1 beginning at Wb = 1/'i. This result implies that the output signa! arnpli-
tude becomes very large at high frequencies (i.e., AR - oo as co - oo), which is a physical impossibility.
 13.3 Bode Diagrams 347

Consequently, a process zero is always found in combination with one or more poles. The order of the
numerator of the process traosfer function must be less than or equal to the order of the denominator,
as noted in Section 6.1.
Suppose that the numerator of a transfer function contains the term 1 - TS, with T >O.As shown in
Section 6.1, a right-half plane (RHP) zero is associated with an inverse step response. The frequency
response characteristics of G(s) = 1 - TS are
AR= Jw1T2 + 1 (13-
<!> = -tan-1(wT) 50a)
(13-
50b)
Hence, the amplitude ratios of LHP and RHP zeros are identical. However, an RHP zero contributes
phase lag to the overall frequency response. Processes that contain a right-half plane zero or time
delay are sometirnes referred to as nonminimum phase systems because they exhibit more phase lag
than another transfer function that has the same AR characteristics (Franklin et al., 2002).1 Exercise
13.11 illustrates the importance of zero location on the phase angle.
For the third-order process of Example 13.3, the slope of the amplitude ratio curve for large w was
seen to be -3. For any general transfer function (Eq. 4-34), the high-frequency asymptote of the AR
curve is given by
high-frequency slope = (numerator order) - (denominator order)
or
SHF = m - 11 (13-51)
The phase angle for large w (<!>HF) approaches a limit
li.m <!>HF = (111RHP + 11 - 111LHP)(-90º)
c.,->O)
(13-52)
where mRHP and rilLHP represent the number of right-half plane and left-half plane zeros, respectively,
aad the n poles are stable (located in the left-half plane). Note that Eq. 13-52 holds only for rational
transfer functions; it is not valid for the case where a time delay is present in the transfer functioo,
which is discussed below.

13.3.S Time Delay

Toe time delay e-6s is the remaining important process element to be analyzed. Its frequency response
characteristics can be obtained by substituting s = jw
G(jw) = e-jwa (13-53)
which can be written in rational forro by substitution of the Euler identity
G(jw) = cos w8 - j sin w8 (13-54)
From (13-54)
AR= IG(¡w)I = Jcos2 wa + sin2w8 = 1 (13-55)
008
<!> = LG(jw) = tau-1(-sin
cos we)
or

<!> = -w6 (13-56)

Because ro is expressed in radians/tirne, the phase angle in degrees is -180úl8'1r. Figure 13.6 illus-
trates the Bode plot for a time delay. The phase angle is unbounded, that is, approaches -oo as w

1Sorne textbooks define a trans[er function to be nonminimum pbase if ít contains either an RHP zero or RHP pole (Kuo, 1995).
348 Chapter 13 Frequency Response Analysis

AR

O.l'---'-.I...LLI-IJ""---'---'-'u..r..u..u.--1'--'-_....._.......,
0.01 0.1 1 10
wO

<I>
{deg}
-360>--���-+-���-+���-ri

-540'---'-J...LLI-IJu.L--L--'-'u.LLU.L--l'--L..L.LLllll
0.01 0.1 10
w9 Figure 13.6 Bode diagram for a time delay, e-es.

becomes large. By contrast, the frequency response angle of all other process elements is restricted
to be smaller in magnitude than sorne multiple of 90°. This unbounded phase lag is an important
attribute of a time delay and is detrimental to closed-loop system stability, as is discussed in
Chapter 14.

In Chapter 6 we presented polynomial approximations for the irrational time-delay transfer

function, e-8•. For the 1/1 and 212 Padé approximations of e-e., given in Eqs. 6-35 and 6-37, plot
the Crequency responses. Compare the accuracy of these approximatioos.

SOLUTION Note that the amplitude ratio for each of the Padé approximations is exact, Toe phase angles of
the two approximations are plotted in Fig. 13.7 together with the exact results for e-e, obtained
in Eq. 13-56. Tbe 111 Padé approxímation gives accurate <!> values for w0 s 1, wbereas tbe 212
approximation is somewbat better, yielding satisíactory results for w0 s 2. •

In the discussion of Padé approximations in Section 6.2.1, we raised the issue of their accuracy.
Now we consider this topic from a frequency response perspective. The amplitude ratio is exact for

(dtg) -3601------11----.>,-4-=---' _.1,.

-5401--���--1-���--+���--1

-720�������������
0.1 10 100 Figure 13.7 Phase angle plots for e-e, and for the 1/1 and 212
- ,; '·º w9 ::e� Padé approximations (G1 = 1/1, G; = 212).
vJ•v.'é:J'J'· l¿J
.,
-,._
 13.3 Bode Diagrams 349

ali Padé approximations; only the phase angle exhibits error. Figure 13.7 shows that the 1/1 approx- ·
imation is quite good in the frequency range, O < co < 1/0. For the 2/2 approximation, the equiva- ·
lent range is O< w < 2/0. If tbe process model also includes lag terms (e.g., FOPTD model), the
time-delay approximations are accurate provided that the process dynamics are not dominated
by the time delay. For example, a 1/1 Padé approximation is generally suitable for engineering
purposes if 01-r < 0.5, wbere 'T is tbe largest lag time constant. This situation is illustrated in
Fig. 6.7 where the time-delay approximations are reasonably accurate for a FOPTD model with
0/-r = 0.25.

'1
As a final exarnple of the use of Bode plots to represent the frequency response characteristics
of typical processes, generate the Bode plot for the transfer function

_ S(&r + l)e-6r ,1
G(s). - (20s + 1)(4s + 1)

where the time constants and time delay have units of minutes.

SOLUTION The Bode plot is shown in Fig. 13.8. Tbe steady-state gain (K = 5) is the value of AR wben
co --> O. Toe amplitude ratio exhibits an intermediate-frequency region where the slope is ap-
proximately -2, but for higher frequencies the ultimate slope is -1 because of tbe effect of the
LHP zero. The phase angle at higb frequencies is dominated by the time delay. •

Before introducing the frequency response of typical feedback controllers in the next section,
it is useful to summarize the frequency response characteristics of tbe individual transfer func-
tion elements we have just analyzed. They forro the building blocks of more complex process trans-
fer function models. Table 13.4 provides the relevant inforrnation in a compact form. The ARN
and <j) plots are given for each transfer function where ARN = ARIK. The asymptotes are also
shown.

AR

( 1

0.1 1 10 100
w (rad/minl

(d:,, �: ¡g3�:: ¡ : :]
0.01 0.1
<o
1
(rad/min)
10 100 Figure 13.8 Bode plot of the transfer function in
Example lJ.5.
 3
el) •
.2
-1� .-.11-
...
11
:¡¡
.;
;.
- e-
....
o
o
p:;
o ,.. o- o o . o o .
o o o .
o o
�
IO
a,
�
a, 1
a, .....
O) ....
O)

1 1 1

,,....._
A"' �
·go
'-'

s ,¡::- e 1
¿_ s '-'
r
... e
l.t,
r � �
1
.."..' "11 e:: 1 + �
"'A N

�
-e- 1
�"'
'-' 1
'j

� A

"u'
o
"'r
it:
� blJ
3

�o .;
.2
.....
;.
.§
....
z
o �bll
l'.l .Q
·e":' ..o..
·:::,
.....
B o ....
o
..".' s:: .... .... .....
.."e':
u.,
"o'
A

...,.., �
p:: �
ij
� :-.C:13 � ��
e T. �� ��

::l
O' �
"'
� \
� ,, ...
.....
:e"'
r�
'1
+
� ,,....._
� �+ �f.. :.: Í
e::,
.> �- .....
+
��
J:,
g
.,
'O

!�...
Q.,
'O :;

-�.,
·g ... ... 8.
�9
'O
§ "..'.
'O B ;:,,..

"'
¡.r.. o "'a 'O 'O ��
... .!. -� ...., A
o ·:::, g
� ¡¡;"' B "' o� ·vi
oe:
.s A
 i"' ,....¡ N <') .,¡

'l'-0
 .... ,.,.
11

o o
o
o o o o o o o o
a, CXJ o
a,
tt>
..,¡-
o o
o o o
o o o o o o
o
1 .....1 <O
a,
a,
1 a,
1
........ ,......,
¿,
.:'
!"'

f �I

L--..J
i
..... r
É
........

'-'
�
3
........
i
r!!=
r
É
,......,
r
�
,......, �

� �
1
3 3
.......
�
+ 1 '-'
r r
\
�·
1
s= s=+
+
....

+
/
\
'···--··
.....
+.,
�
'+".�,'l..+.�.,.
l �

ci
....
352 Chapter 13 Frequency Response Analysis

13.4 FREQUENCY RESPONSE CHARACTERISTICS

OFFEEDBACKCONTROLLERS
In order to use frequency response analysis to design control systems, the frequency-related character-
istics of feedback controllers must be known for the most widely used forms of the PID controUer dis-
cussed in Chapter 8. They include P, PI, PD, and PID controllers as well as severa! variations of the
PD and PID controllers. Table 13.5 provides a summary of their frequency response characteristics. In
the following derivations, we generally assume that the controller is reverse acting (Kc > O).
Proportional Controller. Considera proportional controller with positive
gain
(13-57)
G,(s) = s,
In this case IG,(jw)I = K,, which is independent of w. Therefore,
(13-58)
AR=Kc

and
4> = o· (13-59)

Proporttonal-Integral Controller. A proportional-integral (PI) controller has the transfer function,

G,(s) = K, 1 ( + ns-1 ) = K, -(T {S

TJS+
- 1) (13-60)

Substituting s = jw gives

G,(jw) = Kc(l + - 1-) = K,(1 -

T/]W
..L)
WT[
= K,(1�w_,=)
JWT/
(13-61)

Thus, the amplitude ratio and phase angle are

AR = IGe(. )1 = K, 1+ 1 -K J(WT1)2+1
(w ¡)2 (13-62)
¡w T - e WT!

cj, = LGc(jw) = tan -1(-1/WTr) = tan -1(w1'J) - (13-63)

90º

Figure 13.9 shows the Bode plot for a PI controller with K, = 2 and Tf = 10 rnin. At low frequencies
the integral action dorninates, and the slope of the AR curve is -1. As ro - O, AR - oo, and
<!> - -90º. At high frequencies AR = K, and cj, = Oº; neither is a function of w in this region (cf. the
proportional controller).

Ideal Proportional-Derivative Controller. For the ideal proportional-derivative (PD) controller (cf.
Eq. 8-11)
G,(s) = K,(l + TOS) (13-64)
The frequency response characteristics are similar to those of a LHP zero:
AR= K,J(wTo)2 + 1 (13-65)
cj, = tan -1(wTo) (13-66)

Proportional-Derivattve Controller with Filter. The PD controller is most often realized by

the transfer function

G,(s) = K,
TOS+
+ 1)
( CJ.Tl)S l (13-57)
 o o
o
o
°' °'
�1 �
-----:::-
1

... �
..
í
B

.....
+

.".'

��
r�.::-
�
,....
+
� 1
n
....---...
�1� ¿" ,.
+
... ..... .. i.::"
i.::"
354 Chapter 13 Frequency Response Analysis

AR 101::-���-t--"""<f��t-���i--���+--��--=1
Low-frequency asymptote

·"··' _::I ....... 1

0.001 0.01
:tJ 0.1
w (rad/min)
l
.......!
10
....... 1
100

Figure 13.9 Bode plot of a PI controller, Gc(s) = 2 �+ .

(lClr 1)

- - - With derivative filter

AR

10º
10-2 10-l

90
80
70
60
<!>
50
(deg) 40

30
20
10
o
10-2
w (rad/mln)
Figure 13.10 Bode plots of an ideal PD controller and a PO
controller with derivatíve filler.
Ideal: G,(s) = 2 (4s + 1)
With Derivative Filteri �,(s) � 2 (
0
�;/+\)
 13.4 Frequency Response Characteristics of Feedback Controllers
355

where a has a value in the range 0.05-0.2. The frequency response for this controller is given
by
AR - K (wTv)2 + 1 (13-68)
- ' (aww)2 + 1
(13-69)
The frequency responses of the ideal PO controller and the PO controller with a derivative filer
are compared in Fig. 13.10 for K¿ = 2 and TD = 4 min. In the low-frequency region, both controllers
ex- hibit an AR asymptote slope of zero (AR = Kc) and phase angle approaching zero. In the ideal
case,
AR and q, are dominated by the derivative action term as w---'> oo , At high frequencies, AR is
un-
bounded with a slope of + 1, while the phase angle approaches
+90º.
The pole in Eq. 13-67 bounds the high-frequency asymptote of the AR, as shown in Fig.
13.10
(a= 0.1):
lim AR = lim I G,(jw) 1 = KJa = 2/0.1 = (13-70)
20
c..,--,+CO (l)-+00

Remember that this feature actually is an advantage because the ideal derivative action in (13-
64) would amplify high-frequency input noise owing to its large value of AR in that region. In
contrast, the PO controller with derivative filter exhibits a bounded AR in the high-frequency
region. Be- cause its numerator and denominator orders are both one, the high-frequency phase
angle returns to zero.

Para/le/ PID Controller: The PIO controller can be developed in both parallel and series forms,
as discussed in Chapter 8. Either version exhibits features of both the PI and the PO controllers.
The sim- pler version is the following parallel form (cf. Eq. 8-14):

G,(s) = Kc(l
+ J_ + TDS) =
'TJS
Kc(l2
+ TJS + T/TDS )
'TJS
(13-71)

Substituting s = jw and rearranging gives

Gc(s) = K,(1 + .,..! + j(l)'fo) = K,[1 + j (wTD - ...!_)]

/W'T( WT[
(13-72)

The arnplitude ratio and phase angle are given in Table 13.5.

Figure 13.11 shows a Bode plot for a PIO controller with and without a derivative filter
(see Table 8.1). The controller settiogs are K, = 2, TT = 10 min, TD = 4 mio, and a = 0.1.
Its low- frequency behavior cooforms with that of the PI controller (see Fig. 13.9), whereas
its high- frequency behavior is similar to that of the PO controller (see Fig. 13.10). The phase
angle varies from -90° (w -4 O) to +90º (w---'> oo).
By adjustiog the values of Tt and TD, one can prescribe the shape aod location of the "notch" in
the
AR curve. Oecreasing TI and increasing TD narrows the notch, while the opposite chaoges
broaden it. Note that Fig. 13.11 indicates that the ceoter of the notch is located at w = 11J,.1,.o
where q, = Oº and
AR= K,. Varying K, merely moves the amplitude ratio curve up or down, without affecting the
width of the notch. Generally, the integral time TJ should be larger thau TD, Typically T/ = 4TD,
yielding a suit-
able notch width.

Series PID Controller. The simplest version of the series PID controller is

Gc(s) = K,(T[S + 1 (TDS + 1) (13-73)

)
ns
This controller trausfer functiou can be interpreted as the product of the transfer functioos for PI
and PO controllers. Because the transfer function in (13-73) is physically unrealizable and amplifies
high frequency noise, we considera more practical version that includes a derivative filtec.
356 Chapter 13 Frequency Response AnaJysis

AR 101

80
60
40
/
/
/ .... -- ,'',
....
20
'/I '
<l>
(deg) o
-20
-40
-60
-80
L-......i-.....�ml�-'-L.L..J...LJ..LJ.L ..J'--1.....L..J..L.U.il-----l.-'--.L.LJlJ..UJ

10-3 10-l
"'(rad/min)
Figure 13.11 Bode plots of ideal parallel PID controller and series
PID controller witb derivative filler (a.= 1).

Ideal parallel: Gc(s) = 2(1 + 1� + 4s)

Series with Derivative Filter: G,(s) = 2( 1�; 1 )(o�:/+\)

Series PID Controller witñ a Derivattve Filter. The series controller with a derivative filler was
de- scribed in Chapter 8.

+
0e(s) = K,(-r1s'TJS 1)( TDS +
CY.'Tl)S +1
1) (13-74)

where 0.05 < a << 1.0. A comparison of the amplitude ratios in Fig. and Table 13.5 indicates
that the AR for the controller without the derivative filter in (13-73) is unbounded at high
13.11
frequen-
cies in contrast to the controller with the derivative filter (Eq. 13-74) which has a bounded AR
at ali frequencies. Consequently, the addition of the derivative filter makes the series PID
controller less sensitive to high frequency noise, For the typical value of a = 0.05, Eq. 13-74
yields at high frequencies:

..
AR.,...,.= .l,.i.r..n!G,(jw)I = Kd« = 20K, (13-75)
When TD = O, the series PID controller with filter is the same as the PI controller of Eq.
13-60.
If a controller is direct-actíng (Kc < O), the AR plots in Figs. 13.9 and 13.11 do not change
because
IKcl is used in calculating the magnitude. However, the phase angle is shifted by -180" when K, is
neg-
 13.5 Nyquist Diagrams 357

ative. Por example, a direct-acting proportional controller (K¿ < O) has a constant-phase angle of
-180º. As a practica! matter, it is possible to use the absolute value of K¿ to calculate <I> when designing
closed-loop control systems because stability considerations (see Chapter 11) require the choice of
K¿ < O only when KvKpKm < O. This choice guarantees that the open-loop gain (KoL = KcKvKpKm)
will always be positive. Use of this convention conveniently yields <I> = Oº for any proportional con-
troller and, in general, eliminates the need to consider the -180º phase shift contribution of the con-
troller gain.

13.5 NYQUIST DIAGRAMS

The Nyquist diagram is an alternative representation of frequency response information, a polar
plot of G(jw) in which frequency co appears as an implicit parameter. The Nyquist diagram for
a transfer function G(s) can be constructed directly from IG(jw)I and LG(jw) for different values
of w. Alternatively, the Nyquist diagram can be constructed from the Bode diagrarn because
AR = IG(iw)j and <I> = LG(jw). Advantages of Bode plots are that frequency is plotted explicitly
as the abscissa and the log-log and semilog coordinate systems facilitate block multiplication.
The Nyquist diagram, on the other hand, is more compact and is sufficient for many important
analyses, for example, determining system stability. Most of the recent interest in Nyquist diagrams
has been in connection with designing multiloop controllers and for robustness (sensitivity) studies
(Maciejowski, 1989; Skogestad and Postlethwaite, 1996). For single-loop controllers, Bode plots are
used more often.
Consider the transfer function

1
G(s) = 2s + l (13-76)

with
AR= IG(iw)I = l (13-77a)
J(2w)2 + 1
and
<!> = LG(jw) = -ta1c1(2w) (13·77b)

Figure 13.12 gives the Nyquist diagram for this system. At w = O, IG(jw)I is a maximum and the phase
angle is Oº. As w increases, IG(iw)I - O while <I>---) -90º. Therefore, the polar plot (Nyquist diagram)
remains entirely within the lower right quadrant. The direction of the arrows on the diagram shows

lmaginary
part
0.8

0.4

o Real
w:::: part
--0.4 OQ
1.0

-0.4

-0.8 Figure 13.U Toe Nyquist diagram for G(s) = l/(2s + 1)

<,) =2 plotting Re(G(jw)) and Im(G(jw)).
358 Chapter 13 Frequency Response Analysis

lmaginary
part O

-1

-2

-3

-4'--�-L��_._� .��-'-��.. �_._��_._� .��-'-�--'

-2 -1 o 2 3 4 5 6 7 8
Real part

Figure 13.13 Toe Nyquist diagram for the transfer function in Example 13.5.

how G(jw) varies as w increases. Values of w corresponding to specific points on the curve are shown,
indicating a nonuniform frequency scale. At w = O, jG(jw)I = l. If the transfer function gain is changed
from one toan arbitrary value K, AR changes by the same factor K, and, consequently, the distance
from the origin to the curve changes.
Combining the frequency response characteristics for individual transfer function components is
more difficult for Nyquist diagrams than it is for Bode diagrams, where amplitude ratios and phase an-
gles can be determined by graphical addition. If the transfer function contains a time delay, there will
be an infinite number of encirclements of the origin. This feature results from the unbounded phase
angle of a time-delay (<!> = -ew) discussed earlier in connection with Bode diagrams.

As a more complicated example of a Nyquist diagrarn, consider the transfer function in Exam-
ple 13.5.

SOLUTION The Nyquist diagram in Fig. 13.13 is obtained directly from the magnitude and phase informa-
tion in the Bode plot of Fig. 13.8. Toe time delay results in an infinite number of encirclements
of the origio. However, because of the rapid attenuation of G(jw) with increasing frequency,
only a portion of this detall can be observed. •

SUMMARY
Toe frequency response characteristics of a process, its amplitude ratio AR and phase angle cj>, charac-
terize the dynamic behavior of the process. Analytical expressions for AR and <!> can be derived from
transfer functions for processes and controllers and plotted as functions of frequency in Bode and
Nyquist diagrams. In the next chapter, we show that frequency response analysis provides very useful
techniques for closed-Joop stability analysis and control system design.
 Exercises 359

REFERENCES
Franklin, O. F., J. D. Powell, aod A. Emami-Naeioi, Feedback Maciejowski, J. M., Multivariable Feedback Design, Addison-
Con- trol of Dynamic Systems, 4th ed., Prentice Hall, Upper Wes- ley, Readiog, MA, 1989.
Saddle River, NJ, 2002. Ogunnaike, B. A., and W. H. Ray, Process Dynamics, Modeling,
Kuo, B. C., Autom<Ític Control Systems, 71h ed., Prentice Hall, and Control, Oxford University Press, New York, 1994.
En- Skogestad, S., and l. Postlethwaite, J\fultivariob/e Feedback
glewood Clifís, NJ, 1995. Con- trol: Analysis and Design, Wiley, New York, 1996.
Luyben, W. L., and M. L. Luyben, Essentials of Process
Control,
McGraw-Hill, New York, 1997.

EXERCISES
13.1 A heat transfer process has the following trans- A test has been made with P' varied sinusoidally as
fer function between a temperature T and an inlet
flow rate q where the time constants have units oí P' = 0.5 sin 0.21
minutes: For these condítions, the measured temperature is

�- 3(1-s)
Q'(s) - s(2s + 1) T:,, = 3.464 sin (0.21 + e!>)
lf the flow rate varies sinusoidally with an amplitude Find a value for the rnaximum error bound between
of 2 IJmin and a period of 0.5 mio, what is the ampli- T and T�. if the sinusoidal input has been applied far

•
tude of the temperature signa! after the transients a long time.
have died out?
13.5 For each of tbe following transfer functions, develop
13.2 Using frequency response arguments, discuss how well both the amplitude and phase angle of the Bode plot.
e-& can be approximated by a two-tenn Taylor series Find AR and e!> for each transfer function at value� of
expansion, that is, by 1 - es. Compare your results with w = 0.1, I, aod 10. '
those giveo in the text far a 111 Padé approximarion. 5
13.3 A data acquisition system for eovironmeotal moni- (a) (5s + l)(s + 1)
toring is used to record the temperature of an 5
airstream as measured by a thermocouple. It shows (b) (5s + l)(s + 1)2
ao essentially sinusoidal variation after about 15 s.
The máximum recorded temperature is 127 ºF, and 5(s + 1)
the rninimum is 119 ºF at 1.8 cycles per mio. It is esti- (c) (5s + 1)(0.2s + 1)
mated that the thennocouple has a time constan! of
4.S s. Estirnate the actual maximum and min.imum air 5(-s + 1)
(d) (5s + 1)(0.2s + 1)
temperatures.

13.4 A perfcctly stirrcd tank is used to heat a flowing liq- 5

(e) s(5s + 1)
'\ uid. Toe dynamics oí the systcm bavc beeo
deter- S(s + 1)
mined to be approximately as shown in Fig. El3.4. (f) s(Ss + 1)(0.2s + 1)

•
Heater Flow system 13.6 A second-order process transfer functioo is
10 given by
s+l 5s + 1
G(s) _ K(TaS + 1)
- ,-2s2 + 2tu +1
Figure E13.4
(a) Find IGI and LG when t = 0.2.
(b) Plot IG(iro)I vs. Cll'I' and LG(iw) vs. Cll'I' far the
wbere: range 0.01 :S Cll'I' :S 100 and values of -r0/,- =

Pis the power applied to the heater

Q is the heating rate of the system (O, 0.1, 1, 10).
T is the actual temperature in the tank Note that you can obtain the wT plots by setting -r = 1
Ts, is the measured temperature aod reparameterizing -rn/-r accordingly.
360 Chapter 13 Frequency Response Analysís

•
13.7 Plot the Bode diagram (0.1 s w s 100) of the third- 13.12 Develop expressions for the amplitude ratio of each
order transfer function, liíi\ of the two forms of the PID controller:
� (a) The parallel controller transfer function of Eq.
4 8-14.
G(s) = (!Os + 1){2s + l)(s + 1) {b) The series controller transfer function of Eq.
8-15.
Find both the value of w that yields a -180° phase

•
angle and the value of AR at tbat frequency. Put the results on a single plot along with asymptotic
representations of each AR curve. You may assume
13.8 Using MATLAB, plot the Bode diagram of the fol- that T1 = 4TD and a = 0.1.
lowing transfer Iunction: For what regioo(s) o( w are the differences signifí-

•
_ 6(s + l)e-2s cant? by how much?
G(s) - (4s + 1){2s + 1) 13.13 A process has the transfer function,
Repeat for the situation where the time-delay term is Y(s) _ K
replaced by a 1/1 Padé approximation. U(s) - TS + e es+ 1

13.9 Two thermocouples, one oí them a known standard, with K = 10, -r = 2 min, 6 "' 1 min.
are placed in an air stream whose temperature is Obtain the Bode plot for this transfer function. Do
varying sinusoidally. The temperature responses of you need to approximate the denominator exponen-
the two thermocouples are recorded at a number of tial term, or can your computer program handle it
frequencies, witb the phase angle between the two directly?
measured as shown below. The standard is known
13.14 Appelpolscher has just left IGC's Monday rnoming
to follow first-order dynamics and to have a time
status meeting with Smarly, Quirk, and the opera- tions
constant of 0.15 roin when operating in the air
personnel. Over the weekend, a major unit in the
stream. From the data, show that tbe unknown
hydrogen production process, a centrifugal com-
thermocouple also is first order and find its time
pressor supplying carbon monoxide to a reactor,
constant.
failed. The operations people were able to switch in a
standby unit, an old reciprocating compressor that
Frequency Phase Difference can barely meet mínimum requirernents for through-
(deg) put and discharge pressure.
(cycles/min)
Smarly has questioned the possibility that dis-
o.os 4.5 charge pressure oscillations from the temporary unit
0.1 8.7 may damage the reactor product during the interim
0.2 16.0 operating period by causing flow rate changes in the
0.4 24.5 feed to tbe reactor. Because Appelpolscber has sug-
0.8 26.5 gested the idea of putting one ar two surge tanks be-
1.0 25.0 tween the compressor and reactor to damp out any
2.0 16.7 oscillations, his group was chosen to come up witb a
4.0 9.2 quick design. Sorne assumptions can be made:
(i) The proposed piping arrangemeot would be sim-
ilar to that shown in Exercise 2.5. If two tanks
13.10 Exercise 5.19 considered whetber a two-tank Iiquid
are used, they should be sized identitally.
surge system provided better damping of step dis-
(ii) The valves beíore the surge tanks exhibit ap-
turbances than a single-tank system with tbe same
proxiroate linear pressure/flow relarions.
total volume. Reconsider this situation, this time
{ili) The ideal gas law holds approxiroately.
with respect to sinusoidal disturbances; tbat is, de-
(iv) Pressure perturbations are caused by recip-
termine which system better damps sinusoidal in-
rocating action of the pistons in the compres-
puts o( frequency w. Does your answer depend on sor. These perturbations are approxiroately
tbe value of w?
sinusoidal.

13.ll For the process described in Exercise 6.5, plot the
composite amplitude ratio and phase angle curves on
a single Bode plot Ior eacb of the four cases of nu-
merator dynamics. What can you conclude concern-
ing the importance of the zero location for the
amplitude and phase characteristics of this second-
order system?
(v) Because of the present low-pressure Iimitations,
no more than 10% of the compressor's nominal
discharge pressure Ps can be dissipated in the
surge system, excluding the valve at the entrance
to the reactor. If two surge tanks are included in
the desigo, the aUowable drop across each tank
is5% of Pd,

(vi) Tbe pressure cootroller for tbe reactor is able to
maintain its pressure esseotially constant.
Available Data
(i) Tbe nominal discharge pressure Ps is 200 psig
(gauge). Toe nominal tbroughput of carbon
moncxide is 6000 lb/h.
(ii) The compressor contains four cylinders (pistons)
driven by a common shaft that rotales at 600
rpm. The cylinders are spaced equally around
the shaft to balance shaft loading.
(íii) The máximum (estimated) pressure fluctuation
is 2 psig; that is, the amplitude of the fluctuation
caused by reciprocation of tbe pistons is 2 psig.
(iv) Operating personnel want no more than a
valve
0.02-psig variation in pressure before tbe
al the entrance to the reactor.
(v) The nominal discbarge temperature of the com-
pressor is 300 ºF. The surge tanks probably will
operate isothermally,
Appelpolscher is always worricd aboul losing his
year-end bonus; this time he thinks it will disappear if
any of the reactor product is out of spec and has to be
bumed. He asks you to make a two-step analysis of
the situation:
(a) Would a single tank or two equal-sized surge
tanks be better to damp the pressure ftuctuations?
(If two tanks are used, they would each have to be
about one-fourth the volume of a single tank to
keep total system costs rougbly equivalent.)
(b) How large would the tank or tanks have to be to
reduce pressure fluctuations to an acceptable
leve!?
Your answers should deal with how the proposed
design can incorporate a 20-ft-long, 3-in.-i.d. pipe
tbat now connects the compressor to the reactor.
13.15 Plot a Nyquist diagram for the transfer function in
• Exercise 13.8. Repeat with the exponential term re-
Exercises 361
placed by a 1/1 Padé approximation. Why are these
two plots so different in appearance for large values
ofw?
13.16 Plot the Bode diagram for a composite transfer
(m\ function consisting of the one in Exercise 13.8 mul-
� tiplied by tbat of a parallel-form PID controUer
with K¿ = 0.21, -r1 = 5, and vo= 0.42.
Repeat using the same process transfer function
but with a series PID controller with filler that ero·
ploys the same settings. How different are these two
diagrams? In particular, by how much do the two am-
plitude ratios differ when each is evaluated al a value
•
of w that yields a -180º phase angle?
13.17 Por the process described by the transfer function
12
G(s) = (8s + 1)(25 + l)(0.4s + 1)(0.ls + 1)
(a) Find two second-order-plus-time-delay models
tbat approximate G(s) and are of the forro ·
A Ke-es
G(s) - -,---�-�
- (w + l)(-r2s + 1)
One of the approximate models can be found by
using tbe method discussed in Section 6.3; tbe
other, by visual inspection of the frequency re-
sponses of G and G. . ':
(b) Compare all three models (exact and approxi-
mate) in the frequency domain and also by plot-
ting tbeir impulse responses. ·
13.18 Obtain Bode plots Ior both the transfer function
• given as follows: (
_ 10(2.r + l)e-21
G(s) - (20s + 1)(4s + l)(s + 1)
and a FOP1D approximation obtained using the
method discussed in Sectíon 6.3. What do you con-
elude about the accuracy of the approximation rela-
tive to the original transfer functioo? ·
Chapter 14
Control System Design
Based on Frequency
Response Analysis

TABLE OF CONTENTS
14.1 Closed-Loop Behavíor
14.2 Bode Stability Crlterion
14.3 Nyquist Stabllity Critcrion
14.4 Gain and Phase Margins
14.S Closed-Loop Frcquency Response and Scnsitivity Functions
14.5.1 Sensitivity Functions
14.5.2 Bandwidth
14.S.3 Closed-Loop Performance Criteria
14.5.4 Nichols Chart
14.6 Robustness Anolysls
14.6.1 Scosltivity Analysis
14.6.2 Effect of Feedback Control oo Sensitivity
14.6.3 Robust Stability
Swnmary

Frequency response concepts and techniques play an important role in control system design and
analysis. In particular, they are very useful for stability analysis, control system design, and robustness
analysis. Historically, frequency response techniques provided the conceptual framework for early
control theory and important applications in the field of communications (MacFarlane, 1979).
This chapter begins by presenting useful background information in Section 14.1. The Bode and
Nyquist stability criteria in Sections 14.2 and 14.3 are generally applicable and, unlike the Routh sta-
bility criterion of Chapter 11, provide exact results for systems with time delays. These stability crite-
 14.1 Closed-Loop Behavior 363

ria also provide measures of relative stability, how clase a system is to becoming unstable. Two useful
metrics for relative stability, the gain and pitase margins, are introduced in Section 14.4. The frequency
response of closed-loop systems is considered in Section 14.5, followed by an introduction to robust-
ness analysis in Section 14.6. This last topic addresses the important question of the sensitivity of a
control system to process variations and to uncertainty in the process model used to design the control
system.

14.1 CLOSED-LOOP BEHAVIOR

In Chapters 10 and 12 we observed that control system design involves tradeoffs between conflicting
objectives such as performance and robustness. Now we consider these design objectives in more de-
tail and introduce another issue, the effect of measurement noise. In general, a feedback control sys-
tem should provide the following desirable characteristics (see Chapter 12):
l. Closed-loop stability
2. Good disturbance rejection (without excessive control action)
3. Fast set-point tracking (without excessive control action)
4. A satisfactory degree of robustness to process variations and model uncertainty
5. Low sensitivity to measurement noise
The block diagram of a general feedback control system is shown in Fig. 14.1. It contains three exter-
na! input signals: set point Ysp, disturbance D, and additive measurement noise, N. The noisy output Yn
is the sum of the noise N and the noíse-free output Y. The following analysis illustrates the fundamen-
tal limitations and engineering tradeoffs that are inherent in achieving these characteristics.
The dynamic behavior of the closed-loop system in Fig. 14.1 can be described by the following set of
equations where G is defined as G � GvGpGm:

-
Y- Gd D- G,G N KmG,GvGp (14-1)
1 + G,G 1 + G,G + 1 + G,G sp
y
E= GdGm D Cm N Km y (14·2)
1 -1 ,G +1 ,G p

+ G,G +G +G s

GdGmGcGv D - GmG,Gv N + KmGcGv (14-3).

U=
y i
..
·
1 + G,G 1 + G,G 1 + G,G sp

These equations can be derived easily using the block diagram algebra of Chapter 11. They illustrate
how the three externa! inputs (D, N, and Y,p) affect three output variables: the actual output Y, the ·

l N
Figure 14.1 Block diagram with a disturbance D and measurement noise N.
364 Chapter 14 Control System Design Based on Frequency Response Analysis

error E, and the controller output, U. The nine transfer functions in (14-1) to (14-3) completely charac-
terize the closed-loop performance of the control system. Because each transfer function has the same
denominator, there is a single characteristic equation, 1 + GcG = O. Consequently, ali nine closed-loop
transfer functions have identical stability cbaracteristics.
We consider two examples to illustrate the insight provided by tbese equations. Example 14.1
demonstrates that the control system design should never include the cancellation of a pole by a zero
when both are located in the unstable region (that is, to the right of the imaginary axis). This undesir-
able situation is referred to as an unstable pole-zero cancellation.

Consider the feedback system in Fig. 14.l and the followiog transfer functions:

Suppose that controller G, is designed to cancel the uostable pole in Gp, as in the IMC design
method,
_ -3(1 - 2s)
Ge - s +1
Evaluate closed-loop stability and characterize the output response !ora sustained disturbance.

SOLUTION The characteristic equation, 1 + G,G = O, becomes:

l + 3(1 - 2s) 0.5 = O
s+l l-2s
or
s+2.5=0
In view of the single root at s = -2.5, it appears that the closed-loop system is stable. However,
if we consider Eq. 14-1 for N = Y,p = O,
_ Gd _ -0.5(s + 1)
y - 1 + G,G D - (1 - 2s)(s + 2.5) D
This transíer function has an unstable pole at s = +0.5. Thus, the output response to a distur-
bance is unstable. Furthermore, other transfer functions in (14-1) to (14-3) also have unstable
poles. This apparent contradiction occurs because the characteristic equation does not include
ali of the inforrnation, namely, the unstable pole-zero cancellation. •

This example has demonstrated that even an exact cancellation of an unstable pole leads to instability.
Consequently, an unstable pole should neve, be canceled with a right-half plane zero. In coatrast, open-
loop unstable systems can be stabilized with feedback control, as was demonstrated in Example 11.8.
The potential problem of instability arising from an unstable pole-zero cancellation motivates the
concept of intemal stability. A closed-loop system is said to be internally stable if ali of the closed-loop
transfer functions in (14-1) to (14-3) are stable (Goodwin et al., 2001). This definition is equivalent to
requiring that ali signals in the feedback loop are bounded (Morari and Zafiriou, 1989). For the rest of
this book, we assume that no unstable pole-zero cancellations occur in the formation of Gou.
The next example demonstrates that design tradeoffs inherent in specifying Ge can be inferred from
Eqs. 14-1 to 14-3.

Suppose that Gd = Gp, Gm = Km and that Ge is designed so that the closed-loop system is stable
EXAMPLE 14.2
and IGGcl >> 1 over the frequency range of interest. Evaluate this control system desígn strat-
egy for set-point changes, disturbances, and measurement noise. Also consider the behavior of
the manipulated variable, U.
 14.2 Bode Stability Criterion 365
SOLUTIO Because IGGcl >> 1,
N 1
1 + GcG =O
The first expression and (14-1) suggest that the output response to disturbances will be
very good because YID ""O. Next, we consider set-point responses. From Eq. 14-1,
_! = KmGcG,Gp
Y1p 1 + GcG
Because Gm = Km, G = G,GpKm and the above equation can be written as
Y GcG
Y1p = 1 + GcG
For IGGcl >> 1,

_! "" 1
Ysp

Thus, ideal (instantaneous) set-point tracking would occur.

Choosing Ge so that IGGcl >> 1 also has an undesirable consequence. The output Y becornes
sensitive to noise because Y"" -N (see the noise term in Eq. 14-1). Thus, a design tradeoff is
required between the set-point responses and sensitivity to noise. •

In the next section, we consider one of the most important and useful frequency response results,
the Bode stability criterion.

14.2 BODE STABILITY CRITERION

The Bode stability criterion has two important advantages in comparison with the Routh stability cri-
terion of Chapter 11:
l. It provides exact results for processes with time delays, while the Routh stability criterion pro-
vides only approxirnate results due to the polynomial approximation that must be substituted for
the time delay.
2. The Bode stability criterion provides a measure of the relative stability rather than merely a yes or
no answer to the question, "Is the closed-loop system stable?"
Before considering the basis for the Bode stability criterion, it is useful to review the General Stability
Criterion of Section 11.1:
A feedback control system is stab/e ifand 011ly ifatl roots of tire characteristic equation lie to the left of
the imaginary axis in the complex plane.
Thus, the imaginary axis divides the complex plane into stable and unstable regions. Recall that the
characteristic equation was defined in Chapter 11 as
1 + GoL(s) = O (14-4)
where the open-loop transfer function is GoL(s) = G,(s)Gv(s)Gp(s)Gm(s).
The root Jocus diagrams of Fig. 14.2 and Section 11.5 show how the roots of the characteristic equa-
tion change as controller gain K¿ changes. By definition, the roots of the characteristic equation are the
numerical values of the complex variables that satisfy Eq. 14-4. Thus, each point on the root Jocus sat-
isfies (14-5), which is a rearrangement of (14-4):
GoL(s) = -1 (14-5)
The corresponding magnitude and argumentare:
IGoL(jw)I = 1 and LGoL(jw) = -180º (14-6)
 '\

366 Chapter 14 Control System Design Based on Frequency Response Analysis

lmaginary
part

Real
part

Figure 14.2 Root locus diagram for a process with three potes
and no zeroes.

In general, the ith root of the characteristic equation can be expressed as a cornplex number,
rt = a; ± bij. Notethat complex roots occur as complex conjugare pairs. When a pair is located on the
imaginary axis, the real part is zero (a; = O) and the closed-Ioop system is at the stability Iimit. As indi-
cated in Chapter 11, this condition is referred to as marginal stability or conditional stability. When the
closed-loop system is marginally stable and b¡ :f. O, the closed-loop response exhibits a sustained oscil-
Iation after a set-point change or a disturbance. Thus, the amplitude neither increases nor decreases.
However, if K, is increased slightly, the closed-loop system becomes unstable because the complex
roots on the imaginary axis move into the unstable region.1
For a marginally stable system with bt ,f. O, the frequency of the sustained oscillation, w,, is given by
w, = b;. This oscillatory behavior is caused by the pair of roots on the imaginary axis at s = ±w,j (see
Chapter 3). Substituting this expression for s into Eq. 14-6 gives the fo!lowing expressions for a condi-
tionally stable system:
ARoL(w,) = 1 GoL(jw,) 1 = 1 (14-7)
<l>oL(wc) = LGoL(jw,) = -180º (14-8)

for sorne particular value of w, > O. Equations (14-7) and (14-8) provide the basis for both the Bode
stability criterion and the Nyquist stability criterion of Section 14.3.
Before stating the Bode stability criterion, we need to introduce two important definitions:
l. A critical frequency w, is defined to be a value of w for which <!>oL(w) = -180º. This frequency is
also referred to as a phase crossoverfrequency,
2. A gain crossover frequency w8 is defined to be a value of w for which ARoL(w) = l.
For a marginally stable system, Wc = w8.
For many control problems, there is only a single Wc and a single w8• But multiple values can occur, as
shown in Fig. 14.3 for w,. In this somewhat unusual situation, the closed-loop system is stable for two
different ranges of the controller gain (Luyben and Luyben, 1997). Consequently, increasing the ab-
solute value of Kc can actually improve the stability of the closed-loop system for certain ranges of K,.

tFor K¿ < O, instability occurs if K¿ becomes more negative, that is, if IKcl becomes larger. In the subsequent analysis, we assume
that K; is positive, but the results are also valid for K¿ < O, if K¿ is replaced by IKcl.
 14.2 Bode Stability Criterion 367

o.----�--�---.-----.----.
-90¡--�.........t:---r----t----t----j
(f
deogL) -180 t-----+---""'-c,-----,7""'-+-�...--+-----t

-270t-----+---+---+----\-----t
-360���������������
0.001 0.01 0.1 10 100 Figure 14.3 Bode plot exhibiting multiplc critica)
"'(radians/time) frequencies.

Next we state one of the most important results of frequency response analysís, the Bode stability
criterion. It allows the stability of a closed-loop system to be determined from the open-loop transfer
function.
Bode Stability Criterion. Consideran open-loop transfer function GoL = GcGvGpGm that is strictly
proper (more pales than zeros) and has no poles located on or to the right of the imaginary axis, with
thepossible exception ofa single pole at the origin. Assume that the open-loop frequency response has
only a single critica/ [requency Wc and a single gain crossover frequency Wg. Then the closed-loop
system is stable if ARo¿(w,) < l. Otherwise, it is unstable.
Sorne of the importan! properties of the Bode stability criterion are:
l. It provides a necessary and sufficient condition for closed-loop stability based on the properties
of the open-loop transfer function.
2. Unlike the Routh stability criterion of Chapter 11, the Bode stability criterion is applicable to sys-
tems that contain time delays.
3. The Bode stability criterion is very useful for a wide variety of process control problems. How-
ever, for any GoL(s) that does not satisfy the required conditions, the Nyquist stability criterion
of Section 14.3 can be applied.
4. For systems with multiple w, or w8, the Bode stability criterion has been modified by Hahn et al.
(2001) to provide a sufficient conditíon for stability.
In arder to gain physical insight into why a sustained oscillation occurs at the stability lirnit, consider
the analogy of an adult pushing a child on a swíng, The child swings in the same are as long as the adult
pushes at the right time and with the right amount of force. Thus, the desired "sustained oscillation"
places requirements on both timing (that is, phase) and applied force (that is, amplitudes. By contrast,
if either the force or the timing is not correct, the desired swinging motion ceases, as the child will
quickly exclaim. A similar requirement occurs when a person bounces a ball.
To further illustrate why feedback control can produce sustained oscillations, consider the following
"thought experiment" for the feedback control system in Fig. 14.4. Assume that the open-loop system
is stable and that no disturbances occur (D = O). Suppose that the set point is varied sinusoidally at the
critica! frequency, y,p(t) = A sin(wc t), for a long 'period of time. Assume that during this period the
measured output, Ym, is disconnected so that the feedback loop is broken befare the comparator. After
tbe initial transient dies out, Ym will oscillate at the excítation frequency w, because the response of a
linear system to a sinusoidal input is a sinusoidal output at the same frequency (see Section 13.2). Sup-
pose that two events occur simultaneously: (i) the set point is set to zero, and (ii) Ym is reconnected. lf
368 Chapter 14 Control Systern Desi� Based on Frequency Response Analysis

Figure 14.4 Sustained oscillation in a feedback control system.

the feedback control system is marginally stable, the controlled variable y will then exhibit a sustained
sinusoidal oscillation with amplitude A and frequency Wc,
To analyze why this special type of oscillation occurs only when w = wc, note that the sinusoidal sig-
nal E in Fig. 14.4 passes through transfer fuuctions Ge, Gv, Gp, and Gm before returning to the corn-
parator. In order to have a sustained oscillation after the feedback loop is recoonected, signal Y,n must
have the same amplitude as E anda -180º phase shift relative to E. Note that the comparator also pro-
vides a -180° phase shift due to its negative sigo. Consequently, after Ym passes through the compara-
tor, it is in phase with E and has the same amplitude, A. Thus, the closed-loop system oscillates
indefinitely after the feedback loop is closed because the conditions in Eqs. 14-7 and 14-8 are satisfied.
But what happens if K, is increased by a small amount? Toen, ARoL(wc) is greater than one, the oscíl-
lations grow, and the closed-loop system becomes unstable. In contrast, if K, is reduced by a small
amount, the oscillation is "damped" and eventually dies out.

A process has the third-order transfer function (time constaot in minutes),

(.
Gp(s) =
2
(0.5s + 1)3
.,
l t
\;
l.,
Also, G» = 0.1 and G,,. = 10. For a proportional controller, evaluate the stability of tbe closed-
loop control systern using the Bode stabilíty criterion and tbree values of K,: 1, 4, and 20.

SOLUTION For Ibis example,

Go1, = G,GvGpGm = (K,)(0.1) (o.s/+ !)3 (10) = (0.5;K_; !)3

Figure 14.5 shows a Bode plot of GoL for three valúes of K,. Note that all three cases have tbe
sarne phase angle plot because tbe phase lag of a proportional controller is zero Cor K, > O.
From the phase angle plot, we observe that w, = 3.46 rad/m.in. This is the frequeocy of the
sustained oscillation that occurs at the stability lirnit, as discussed above. Next, we consider
the amplitude ratio AR01, Cor eacb value of K,. Based oo Fig. 14.5, we make the following
classifications:

AR01, (Cor w = w,) Classification

1 0.25 Stable
4 1 Marginally stable
20 5 Unstable
 14.2 Bode Stability Criterion 369
lOO,---.,.....,..,....,mn---,--.,.,..,,mr--r-""TTT,.,..,,...--.-,.,.,"TTm
101:.;:.-==-=:=l-=-=a-===::-t-��=-+-'------i
ARoL

�11----�1--��--l��f-ll..->w.:.-���

0.01 �--�---�--.--����

.!�i> =:!I r ::--,--i

--i-tt-,- - -.-
; =.

r---
..-
..-
. t=

.-,1 Figure 14.S Bode plots for Gol = 2K,l(0.5s + 1)3.

.,.
0.01 0.1 Wc 10 }00
w (radfmin)

In Section 12.5.1 the concept of the ultimate gain was introduced. For proportional-only
con- trol, the ultimate gain Kw was defined to be the largest value of K¿ that results in
a stable closed-loop system. The value of Kw can be determined graphically from a Bode
plot for trans- fer function G = GvGpGm. For proportional-only control, Gol = K,G. Because
a proportional controller has zero phase lag, w, is determined solely by G. Also,
ARol(w) = KcARG(w) (14-9)
where ARG denotes the amplitude ratio of G. At the stability limit, w = w,, ARol(w,) = 1
and K, = K,u. Substituting these expressions into (14-9) and solving for Kw gives an important
result:
1
Kcu = ARG(wc) (14-10)°

The stability limit for K, can also be calculated for PI and PID conlrollers, as demonstrated
by
Example 14.4. •

Consider PI control of an overdamped second-order process (time constants in minutes),

5
Gp(s) = (s + 1)(0.Ss + 1)
Gm = Gv= 1
Determine the value of Kcu. Use a Bode plot to show that controller settings of K, = 0.4
and
'</ = 0.2 min produce an unstable closed-loop system, Find Ko«. the maximum value of K,
.. that
• 1 can be used with '<t = 0.2 min and still have closed-loop stability. Show that '<1 = 1 mio results
in
a stable closed-loop system for all positive values of K,.

SOLUTION In order to determine Km, wµet..Gc....=-K,� Toe open-loop transfer function is Gol =
K,G
. where G = GvGpGm. BecaÚse a proportional controller does not introduce any phase lag,
G and Gol have identical phase angles. Consequently, the critical frequency can be
determined graphically from the phase angle plot for G. However, curve A in Fig. 14.6
indicares that w,
does not exist because <l>ol is always greater than -180º. As a result, K,., does not exist,
and
thus K¿ does not have a stability
limit.
Conversely, the addition of integral control action can produce closed-loop instability,
Curve· B in Fig. 14.6 indicates that an unstable closed-loop system occurs for G,(s) = 0.4 (1
+ I/0.2s) because ARol > 1 when <l>ol = -180º.
To find Kcm for -t¡ = 0.2 mín, we note that w,.depends on T/, but not on K¿ because K¿ has
no effect on <!>Ol· For curve B in Fig. 14.6, w� = 2.2 rad/min and the corresponding amplitude
ratio is ARol = 1.38. To find Ko«, multiply the current value of K¿ by a factor, 1/1.38. Thus,
K,m =
0.4/1.38 = 0.29.

. ......
·7.:
r
370 Chapter 14 Control System Desigu Based on Frequency Response Analysis

100
10
AR
0.1
0.01 .\
90 Figure 14.6 Bode plots for Example 14.4:
Curve A:. Gp(s)
o
"'
(deg) -90 Curve B: Go¿(s); Gr(s) = 0.4( 1 +
-180 O.�) Curve C: Go¿(s); Gc(s) = 0.4(1
-270
0.01 0.1 1 "'e + �)
"'
(rad/min)

When TJ is increased to 1 min, curve C in Fig. 14.6 results. Because curve C does not have a--
--..:_. critica! frequency, the closed-loop system is stable for aU positive valúes of K,.
.a·
Find the critica! frequency for tbe foUowing process and PID controller assuming G; = Gm
.EXAM.PLE 14.5 = 1:
e-0.3.s
Gp(s) = (9s + l)(lls + 1) G,(s) =20(1 + z1s +s)
SOLVrJON Figure 14.3 shows the open-loop amplitude ratio and phase angle plots for Gol, Note that
the phase angle crosses -180º at three points. Because there is more than one value of wc,
the Bode stability criterion cannot be applied. However, the Nyquist stabílity criterion
presented in Sec- tion 14.3 can be used to determine stability.
•

14.3 NYQUIST STABILITY CRITERION

The Nyquist stability criterion is similar to the Bode criterion in that it determines closed-loop sta-
bility from the open-loop frequency response characteristics. Both criteria provide convenient mea-
sures of relative stability, the gain and phase margins, which will be introduced in Section 14.4. As
the name implies, the Nyquist stability criterion is based on the Nyquist plol for GoL(s), a polar
plot of its frequency response characteristics (see Chapter 13). Toe Nyquist stability criterion <loes
not have the same restrictions as the Bode stability criterion because it is applicable to open-loop
unstable systerns and to systems with multiple values of Wc or wg, The Nyquist stability criterion is
the most powerful stability test that is available for linear systems described by transfer function
rnodels.
The Nyquist stability criterion is based on two concepts from complex variable theory, contour map-
ping and the Principie of tite Argument. We briefly review these concepts in Appendix D. More de- .
tailed descriptions are available elsewhere (Brown and Churchill, 1996; Franklin et al., 2002). We now
present one of the most irnportant results of frequency domaio analysis:
Nyqutst Stabllity Criterion. Consideran open-loop transfer function GoL(s) that is proper and has
no unstable pole-zero cancellations. Let N be the number of times that the Nyquist plot for GoL(s)
encircles the (-1, O) point in the clockwise direction. Also /et P denote the number ofpotes of GoL(s)
that lie to the right ofthe imaginary axis. Then, Z = N + P where Z is the number of roots (or zeros)
ofthe characteristic equation that lie to the right of the imaginary axis. The closed-loop system is stable
if and only ifZ = O.
 14.3 Nyquist Stability Criterion 371

Sorne important properties of the Nyquist stability criterion are:

l. lt provides a necessary and sufficient condition for closed-loop stability based on the open-loop
transfer function.
2. The reason that the ( -1, O) point is so important can be deduced from the characteristic equation,
1 + GoL(s) = O. This equatíon can also be written as GoL(s) = -1, which implies that ARoL = 1
and <poL = -180º, as noted earlier. This point is referred to as the critica!point.
3. Most process control problems are open-loop stable. For these situations, P = O and thus Z = N.
Consequently, the closed-loop system is unstable if the Nyquist plot for Go¿(s) encircles the criti-
ca! point, one or more times.
4. A negative value of N indicates that the critica! point is encircled in the opposite direction (coun-
terclockwise). This situation implies that each countercurrent encirclement can stabilize one un-
stable pole of the open-loop system.
5. Unlike the Bode stability criterion, the Nyquist stability criterion is applicable to open-loop un-
stable processcs.
6. Unlike the Bode stability criterion, the Nyquist stability criterion can be applied when multiple
values of Wc or w8 occur (cf. Fig.14.3).
Control system design based on Nyquist plots is described elsewhere (Kuo, 1995; Luyben and Luyben,
1997).
Example 14.6 illustrates the application of the Bode and Nyquist stability criteria.

Evaluate the stability of the closed-loop system in Fig. 14.1 far:

4e-' I"·"-·
Gp(s) = 5s + 1 ��

The time constant and time delay have units of minutes and,
.,,,.
G,
//
,,..,2; G» = 0,25,
/ '
Ge = K¿ '
.:,../
Obtain'éi.; aod Ke« from a Bode plot. Let K¿ = 1.5Kcu and draw the Nyquist plot for the result-

ing o�=tbop·sy�fom. · , - , ··

SOLUTION The Bode plot for Gol and K¿ = 1 is showo in Fig. 14.7. For wc = 1.69 rad/min, cf>ol = -180º
and ARoL = 0.235. For K¿ =·í,ARoL = ARG and Kcu can be calculate<l from Eq, 14-lÓ. Thus,
Kcu = 1/0.235 = 4.25. Setting K¿ = ·1.5Kcu gives 's, = 6.38.

o ;r� �·:�i±
..
ARot """I
90�����--�-.---�--
�
Ot---�;;;;;;:;=t-----t---:----t------1
�OL -901------1---=----1-c:..-.----l---�
(deg) -180 1-------1----+�---+-------l
-2701-----4----+-+-�--+------l
-360'--'--'--'-'-Ull<----"'-'-.LL.Ll.W....+J---'--"I.U./JL.LI.-....L...L..LJ'-LWJ
0.01
"'e = 1.69
rad/min
"'(rad/min) Figure 14.7 Bode plot for Example 14.6, K¿ = l.
372 Chapter 14 Control System Design Based on Frequency Response Analysis

lmaginary
part
2

Real
part
-4 2

-2

Figure 14.8 Nyquist plot for Exarnple 14.6: K¿ = 1.SKcu = 6.38.

Toe Nyquist plot for K, = 6.38 is shown in Fig. 14.8 (!ow-frequency data for O < w < 0.4
have been omitted). Note that the -1 point is encircled once. Applying the Nyquist
stability criterion
gives N = 1, P = O, and Z = l. Thus, the larger value of K¿ causes the closed-loop system to
be-
come unstable. Only values of K¿ less than Kcu result in a stable closed-loop system.
•

At this point, it is appropriate to summarize the relative advantages and disadvantages of tbe
Bode and Nyquist plots. The Bode plot provides more information than the Nyquist plot because
the fre- quency is shown explicitly. In addition, it facilitates analysis overa wide range of frequencies
due to its logarithmic frequency scale. Another advantage of the Bode plot is that it allows the
open-loop fre- quency response characteristic to be graphically constructed from the characteristics
for the individual transfer functions, Ge, Gv, Gp, and G»; as shown in Chapter 13. The chief
advantage of the Nyquist plot is that the Nyquist stability criterion is more widely applicable than
the Bode stability criterion.

14.4 GAJN AND PHASE MARGINS

If the process changes or the controller is poorly tuned, the closed-loop system can become
unstable. Thus, it is useful to have quantitative measures of relative stability that indicate how close
the system is to becorning unstable. The concepts of gain margin (GM) and phase margin (PM) provide
useful met- rics for relative stability.
Let AR, be the value of the open-loop amplitude ratio at the critica! frequency wc, Gain margín
GM
is defined as:
GMg,_l_ (14-11)
A.Re
According to the Bode stability criterion, A.Re must be less than one for closed-loop stability.
An
equivalent stability requirement is that GM > l. The gain margin provides a measure of relative
stabil-
ity because it indicates bow much any gain in the feedback loop can increase before instability
occurs. For example, if GM == 2.1, either process gain Kp or controller gain K¿ could be doubled
and the closed-loop system would still be stable, though probably very oscillatory.
Next, we consider the phase margin. In Fig. 14.9, <!>, denotes the phase angle at the gain-
crossover frequency wg where ARoL = l. Phase margin PM is defined as
A
PM = 180 + q>g (14-U)
The phase margin also provides a measure of relative stability. In particular, it indicates how much
ad-
ditional time delay can be included in the feedback loop before instability will occur. Denote the
addi-
 14.4 Gain and Phase Margins 373

ARoL
AR = _1_ 1------ii-----------1
e GM

<!>OL <Pg
(deg)

Figure 14.9 Gaio and phase rnargins on a Bode plot.

tional time delay as ó6max, For a time delay of u6max, the phase angle is -ó6maxW (see Section
13.3.5). Thus, u6max can be calculated from the following expression,

PM = ó6maxWc (180º)
---;- (14·13)

or

(14·14)

where the ('rr/180º) factor converts PM from degrees to radians. Graphical representations of the
gain and phase margins in Bode and Nyquist plots are shown in Figs.14.9 and 14.10.
The specification of phase and gain margins requires a compromise between performance and
ro- bustness. In general, large values of GM and PM correspond to sluggish closed-loop responses,
while smaller values result in less sluggish, more oscillatory responses. The choices for GM and
PM should also reflect model accuracy and the expected process variability.
Guideline: In general, a well-tuned controller should have a gain margin between 1.7 and 4.0
and a phase margin between 3(/' and 45º.

lmaginary
part
,,,,,,..,,...,,..-- --- <, ,

/ ''
f
I/
/
' \
\
I I

-_l-<1,... ,;.,,:::::..-�---- .;_ Real

part
I
I
I
f
I
/
//

i ,,,,,,,.
---

l Figure 14.10 Gain and phase margins on a Nyquist plot.

374 Chapter 14 Control System Design Based on Frequency Response Analysis

Recognize that these ranges are approximate and that it may not be possible to choose PI or PID con-
troller settings that result in speci.fied GM and PM values. Tan et al. (1999) have developed graphical
procedures for designing PI and PID controllers that satisfy GM and PM specifications. The GM and
PM concepts are easily evaluated when the open-loop system does not have multiple values of Wc or
wg, However, for systems with multiple oog, gain margins can be determined from Nyquist plots (Doyle
et al., 1992).
For the FOP1D model ofExample 14.6, calculate PID controller settings for the two tuning re-
·EXAMPLE.14. lations in Table 12.6:
7.
(a) Ziegler-Nichols
(b) Tyreus-Luyben

Assume that the two PID controllers are implemented in the parallel forrn with a derivative fil-
ter (a = 0.1) in Table 8.1. Plot the open-loop Bode diagram and determine the gain and phase
margins for each controller.

-- Ziegler-Nichols
- - - Tyreus-Luyben
1021,-��...,..��---,1--�����-+������-cl

10-}��. ,._._._._._,_,_,_�...._ ._... .,�LLL�-'----'--'-''-L->-'-'-'

10-2 10-1 10° 101
ú) (rad/mln)

----- ----
-toor:=:=:::::,,,,, ..._��1---��----�-�"T--l-����� j

-300��"'---'--'-'-'-'"""""''--�'---'-�................�_._ ,�_._L..LI..LJ
10-2
ú) (rad/min)
Figure 14.11 Comparison of Gol Bode plots for Example 14.7.
 14.4 Gain and Phase Margins 375

For the Tyreus-Luyben settings, determine the maximum increase in the time delay Aemu lhat
can occur while still maintaining closed-loop stability.

SOLUTION From Example 14.6, the ultimate gain is Kcu = 4.25 and the ultimate period is P11 = 2'7T/l.69 =
3.72 min. Therefore, the PID controllers bave the following settings:

Controller T/ TD
Settings K, (min) (min)

Ziegler-Nichols 2.55 1.86 0.46

Tyreus-Luyben 1.91 8.18 0.59

Toe open-loop transfer functioo is:

2e-s
Go1. = G,G,GpGm = G,
55
+l

Figure 14.11 shows tbe frequency response oí Go1. for the two controllers. The gain and phase
margins can be determined by inspection o( the Bode diagram or by using tbe .MATLAB corn-
mand, margin.

Controller GM PM Wc (rad/min)
Ziegler-Nichols 1.6 40° 2.29
Tyreus-Luyben 1.8 76º 2.51

The Tyreus-Luyben controller settings are more conservative owing to tbe larger gain and
phase margins. The value o{ Aemu is calcuJated from Eq. 14-14 and the information in the
above table:
_ (76º)('1T rad) _ .
Aemu - (0.79 rad/min)(180º) - 1.7 nun
Thus, time delay 6 can increase by as muchas 70% and still maintain closed-loop stability. •

Although the gain and phase margins provide useful metrics for robustness, they can give
misleading results for unusual situations. For example, the Nyquist plot for a stable open-loop
process in Fig.
14.12 exhibits large GM and PM values, but the Nyquist curve passes very close to the critica!
point.

lmaginar
y part

,.
.>" --- --- ........ ,
''
I
I
/
/
'\ '
I \
-lf \ Real
---"<;�=--y----,,1------;- part
I
I
I
/

,. ,.
/

.>

Figure 14.12 Nyquist plot where the gain and phase rnargins
are misleading.
 '1

376 Chapter 14 Control System Design Based on Frequency Response Analysis

Thus, the closed-loop system is not very robust and a small process perturbation could cause instabil-
ity. This potential shortcoming of the gain and phase margins can be avoided by considering the sensi-
tivity functions that are introduced in the next section.
The open-loop transfer function and its frequency response characteristics play a key role in control
system design and anaJysís. In the loop-shaping approach, the controller is designed so that the open-
loop transfer function has the desired characteristics. These techniques are described in books on ad-
vanced control theory (Doyle et al., 1992; Skogestad and Postlethwaite, 1996).

14.5 CLOSED-LOOP FREQUENCY RESPONSE r

AND SENSITIVITY FUNCTIONS

The previous sections have demonstrated that open-loop frequency response characteristics play a key
role in the analysis of closed-loop stability and robustness. However, the closed-Ioop frequency re-
sponse characteristics also provide important information, as discussed in this section.

14.5.1 Sensitivity Functions

The following analysis is based on the block diagram in Fig. 14.1. We define Gas G � GvGpGm andas-
sume that Gm = Km and Gd = l. Two important concepts are now defined:
I!,. 1
S = 1 + G,G sensitivity function (14-15a)

T� G,G compJementary sensitivity function (14-15b)

1 + G,G

Comparing Fig. 14.l and Eq.14-15 indicates that Sis the closed-loop transfer function for disturbances
(Y/D), while T is the closed-Ioop transfer function for set-point changes (Y!Ysp). lt is easy to show that:
S+ T=l (14-16)
As will be shown in Section 14.6, S and T provide measures of how sensitive the closed-loop system is
to changes in the process.
Let IS(jwl and ITUwl denote the amplitude ratios of S and T, respectively. The maximum values of
the amplitude ratios provide useful measures of robustness. They also serve as control system design
criteria, as discussed below. Define Ms to be the maximum value of IS(iwl for all frequencies:
Ms � max!S(iw)I (14-17)
"'
The máximum value Ms also has a geometrical interpretation (Ástrom and Hagglund, 1995). . ·,
Let
GoL(s) = G,(s)G(s). Then Ms is the inverse of the shortest distance from the Nyquist plot for
Gotis)
to the critica! point. Thus, as Ms decreases, the closed-loop system becomes more robust.
The second robustness measure is MT, the maximum value of IT(iw)I:
..MT � max]
w
T(jw) 1 (14-18)
MT is also referred to as the resonant peak. Typical amplitude ratio plots for S and Tare shown in Fig.
14.13. If the feedback controller includes integral action, offset is eliminated for set-point changes or
sustained disturbances. Thus, at low frequencies, IT(iw)l -1 and IS(jw)I � O, as shown in Fig, 14.13.
It is easy to prove that Ms and Mr are related to the gain and phase margins of Section 14.4 (Skoges-
tad and Postlethwaite, 1996):

GM � M�� l' PM :2: 2 sin-1(2 �s) (14-19)

GM :2: 1 + JT PM :2: 2 sin-1(2 .iJ 7

) (14-20)
 14.5 Closed-Loop Frequency Response and Sensitivity Functions 377

AR

Figure 14.13 Typical S and T magnitude plots.

(Modified from Maciejowski (1989)).

Designing a controller to have a specified value of Mr or Ms establishes lower bounds for GM and PM.
For example, when Mr = 1.25, (14-20) indicates that GM � 1.8 and PM � 47°, numerical values that
satisfy the guidelines given in Section 14.4. Equations 14-19 and 14-20 imply that better performance
(corresponding to a larger value of Mr) is achieved at the expense of robustness because the lower
bounds for GM and PM in Eqs. 14-19 and 14-20 become smaller. In general, a satisfactory tradeoff be-
tween performance and robustness occurs for the following range of values (Ástrom et al., 1998): ,
Guidelines. For a satisfactory control system, Mr should be in tire range of1.0-1.5 and Ms
should be in the range of1.2-2.0.

14.S.2 Bandwidth
In this section we introduce an important concept, the bandwidth. A typical amplitude ratio plot for T
and the corresponding set-point response are shown in Fig. 14.14. The bandwidth wsw is defined as the
frequency at which IT(fw)I = 1)2 = 0.707. The bandwidth indicates the frequency range for which sat-
isfactory set-point tracking occurs. In particular, WB\V is the maximum frequency for a sinusoidal set
point to be attenuated by no more than a factor of 0.707. The bandwidth is also related to speed of re-
sponse. In general, the bandwidth is (approximately) inversely proportional to the closed-loop settling
time. ·
An alternative definition of the bandwidth is sometimes used. The bandwidth can also be defined as
the frequency for which IS(fw)I = 0.707 when IS(fw)I crosses 0.707 from below, as shown in Fig. 14.13
(Skogestad and Postlethwaite, 1996). Fortunately, the two bandwidth definitions typically result in

Closed-loop Set-point
amplitude ratio step response
1.4
MT� l.25 2
1.2

1.0 1 y
IT<jwll 1
1
1
0.8 i1
1
0.6 1
i 1
1
j

l
"'P "'BW
Frequency Time
Figure 14.14 Typical closed-loop amplitude ratio IT(jw)I and set-poinr response.
378 Chapter 14 Control System Design Based on Frequency Response Analysis

similar nurnerical values. Furthermore, the values of wsw and the gain crossover frequency w8 are usu-
ally very close.
The curves in Fig. 14.14 are similar to the corresponding plots for an underdamped second-order
system (see Sections 5.4 and 13.3.3). For Fig. 13.3 a peak amplitude ratio of 1.25 corresponds to a
darnping coefficient of i = 0.5 for an underdamped second-order system. This analogy provides sup-
port for the previous guideline that My should have a value between 1.0 and 1.5.

14.5.3 Closed-Loop Performance Criteria

Ideally, a feedback controller should satisfy the following criteria.

l. In arder to eliminate offset, jT(jw)I -1 as w-+ O.

2. IT(jw)I shóuld be maintained at unity up to as high as frequency as possible. This condition en-
sures a rapid approach to the new steady state during a set-point change.
3. As indicated in the guidelines, MT should be selected so that 1.0 < MT < 1.5.
4. The bandwidth wsw and the frequency wr for Mr should be as large as possible. Large values re-
sult in fast closed-loop responses.
Satisfying these criteria typically requires a comprornise. For example, the requirement that MT < 1.5,
meaos that the controller gain K, cannot be too large. However, smaller values of K, result in srnaller
values of wsw and tar.
Next, we consider the desired closed-loop frequency response characteristics for disturbances. lde-
ally, we would like to have the closed-loop amplitude ratio jS(jw)I, be zero for all frequencies. How-
ever, this ideal situation is physically impossible for feedback control, and thus a more realistic goal is
to rninimize jS(jw)I overas wide a frequency range as possible. According to the Guideline, the con-
troller should be designed so that 1.2 < Ms < 2.0.

14.5.4 Nichols Chart

The closed-loop frequency response can be calculated analytically from the open-loop frequency re-
sponse. Again consider Fig. 14.1 and assume that the sensor dynamics are negligible so that Gm(s) =
Km, From Fig. 14-1 and Eq. 14-15b, it follows that the closed-transfer function for set-point changes
Y!Ysp is a function of Gol,

_r_= T= Gol (14-21)

Ysp 1 + GoL
where Gol i G,GvGpKm, Denote the open-loop frequency response characteristics by ARol and
<!>ol, and the correspondíng closed-loop quantities, ARcL and <!>CL, Equation 14-21 and the analytical
techniques of Chapter 13 can be employed to derive analytical expressions for ARcL and <!>el as func-
tions of ARol and <!>oL (Luyben and Luyben, 1997).
The Nichols chart in Fig. 14.15 provides a graphical display of the closed-loop frequency response
characteristics for set-point changes when Gm(s) = Km, Contours of constant ARCL and <!>CL are shown
on a plot of ARol vs. <l>ol- In a typical Nichols chart application, ARoL and <!>ol are calculated from
Gol(s) and plotted on the Nichols chart as a series of points. Then ARCL and <!>CL are obtained by in-
terpolation. Por example, íf ARol = 1 and <l>ol = -100° at a certain frequency, then interpolation of
Fig. 14.15 gives ARCL = 0.76 and cl>cl = - 50º for the same frequency. The Nichols chart can be gener-
ated in the MATLAB Control Taolboxñ by a single command, nichols.
The Nichols chart served as a useful graphical technique for designing control systems prior to 1960,
but it has become less irnportant now that control software is widely available. Control system designs
based on Nichols charts, including applications where Gm(s) *Km, are described elsewhere (Kuo,
1995; Franklin et al., 2002).
The following example illustrates PID controller design for an Ms robustness constraint.
 14.5 Closed-Loop Frequency Response and Sensitivity Functions
379
20

- ARel /!� � r-- /

<,
1.0 <l>cL

10 1.10 - / � --�- -- -- -s•-

- '
'
o , AR et
�
..;

1.2q . ......
_..,,.
'

=
I

--- 23--
-c ....-
- f -10·
- ..... , ,,,...-
. ' -,
'\ I J 1/ ..,,/

- 1.60:
.....:..... - �

o f / �e- I
�., 1.40 ....
- 1.80 �� > -,-- -\ 20°
T 71 /!-------..�
- �i
"O
2.0

-c�. ..--2.5
6 -':S; _.:J
-
-
-); ¡:;,--
�� t.>< .
Jo._:.:: �
"'
o, - ..
->:� �
�
, , ,, _.. �
E >--4.0 3�
�-
� � ::::o.s ·
.s- - � '"lo;;----

� �o. � �
' '

--
»·.
«:
' '

��
..... �
�- _.-.-- ;>r

""' - -
[ -
.... ... -
r-
', '

o '- �<;

-,
'
-, ·, ''"b-. �
'
' ' ' ' o

°' \ ·,
'Po 'f� l��o. Ko. \ I'
��Oo- � �
' '
 -
1
<l>cL ..... , .....
1 ,'... \� '¡., \;.., \� 't \� \ \ b, -,

,, \
\
1 ' "\ \
' ' '
1 'f=i \u >
(X)
�
\11' ,o. 'q
� I

0.1 o I� 10. 1 •
-190° -170° -150° -130° -110° -90° -70° -50°
Open-loop phase angle, c!>oL
Figure 14.15 A Nichols chart. (Toe closed-loop amplitude ratio ARcL (--) and
phase angle <l>cL (- - -) are shown as families of curves.]

Consider a fourth-order process with a wide raoge of time constants tbat have units of
minutes
(Ástrcm et al., 1998): ·
1
G = GvGpGm = (s + 1)(0.2s + 1)(0.04s + 1)(0.008s + (14-22)
1)

Calculate PID cootroller settings based on following tuning relations in

Chapter 12: (a) Ziegler-Nichols tuning (Table 12.6)
(b) Tyreus-Luyben tuning (Table
12.6)
(e) IMC tuning with 'Te= 0.25 min (Table 12.1)
(d) Simplified IMC (SIMC) tuning with 'fe= 0.25 min (Table
12.5)
For parts (e) and (d) use a second-order plus tirne-delay model derived frorn
Skogestad's model approximation method (Section 6.3.1).
Determine sensitivíty peaks Ms and Mr for each controller. Compare the closed-loop
responses to step changes in the set-point and the disturbance using tbe parallel fonn of
the PID con- troller without a derivative filler:

G,(s) = «, [1 + ..!... + 'T'DSJ

'TJS
(14-23)

Assume that Gd(s) = G(s).

SOLUTION The Bode plot for G(s) is not shown but indicates that w, = 11.18 rad/min, Pu = 21'(/w,
=
0.562 min, and Kw = 1/ARc = 30.2. Applying Skogestad's model approximation
procedure gives:
_ e-o.02&r
G(s) = (s + 1)(0.22s + 1) (14-24)

Toe SIMC controller in Table 14.1 was calculated using the tuning relations in Table 12.5
for
'T't < 89. The PID controller settings are compared in Table 14.1. The Ms and Mr
values
380 Chapter 14 Control System Design Based on Frequency Response Analysis

Table 14.1 Controller Settings and Peak Sensitivities for Example 14.8
Controller Kc TI {min) TD (ro.in) Ms Mr
Ziegler-Nichols 18.1 0.28 O.Q70 2.38 2.41
Tyreus-Luyben 13.60 1.25 0.089 1.45 l.23
Th1C 4.39 1.22 0.180 1.13 1.00
Simplifled IMC 21.8 1.22 0.180 1.58 1.16

determined Irom the sensitivity plots are also summarized in Table 14.1. The controller settings
and sensitivity values in Table 14.1 indicate that the Z-N controller settings are the most aggres-
. sive and the IMC settings are the most conservatíve. In fact, the sensitlvity values for Z-N con-
troller are unacceptably large, according to the guidelines.
Toe closed-loop responses for a uoit set-point change at r = O anda step disturbance (d = 5)
at r = 4 min are shown in Fig. 14.16. The ZN controllcr provides the worst set-point response
but the best disturbance response. The IMC controller is quite sluggish because the approxi-
rnate SOPTD model has a relatívely small time delay. Similar results were obtaioed for the
FOYID model in Example 12.4. The SIMC and T-L controllers provide the best overall perfor-
mance of these four controllers. However, a PID controller with improved performance for this
example (not shown) can be obtained using a Direct Synthesis approach based on disturbance
rejection (Chen and Seborg, 2002). •

Example 14.8 has demonstrated that a variety of PID controllers can have approximately the same
Ms value but different performance characteristics. Thus, although the guidelines for Ms and MT pro-
vide useful limits concerning controller robustness, controller performance should also be considered.
Robustoess metrics such as Ms, MT, GM, and PM should be evaluated in conjunction with controller
design methods, especially the model-based techniques of Chapter 12.

14.6 ROBUSTNESS ANALYSIS

In order for a control system to function properly, it should not be unduly sensitive to small changes in
the process or to inaccuracies in the process model, if a model is used to design the control system. A
control system that satisfies th.is requirement Is said to be robust or insensitive. It is very important to
consider robustness as well as performance in control system design. First, we explain why the S and T
transfer functions in Eq. 14-15 are referred to as "sensitivity functions."

2
l.8 r»; Z-N
1.6 I ' T-L
I ' IMC
I '
1.4
'',..... SIMC
1.2 ......... _
y
0.8
0.6
0.4
0.2
o Figure 14.16 Closed-loop responses Ior Example
o 2 3 4 5 6 7 8 9 14.8. A set-point change occurs at t = O anda step
Time {min) disturbance at t = 4 min.
 14.6 Robustness Analysis 381

14.6.1 Sensitivity Analysis

In general, the terrn sensitivlty refers to the effect that a cbange in one transfer function (or variable)
has on another transfer function (or variable). Suppose that G changes from a nominal value Go toan
arbitrary new value, Go + dG. This differential change dG causes T to change from its nominal value
To to a new value, To + dT. Thus, we are interested in the ratio of these changes, dT/dG, as well as in
the ratio of the relative changes:
dT!T ú ...
(14-25)
dG/G = sensitívíty

Toe sensitivity is of particular interest because it is dimensionless and independent of the units of G
and T. It is evaluated for a specific condition such as Go and To. We can write the sensitivity in an
equivalent form:

dTIT = (ªT) G (14-26)

dG/G dG T
'The derivative in (14-26) can be evaluated after substituting the definition of T in (14-15b):
dT = G,:Sl . (14-27)
dG
Substitute (14-27) into (14-26). Then substituting the definition of Sin (14-15a) and rearranging gives
the desired result:
dTIT _ 1
dGIG - 1 + GeG = S
Equation 14-28 indicates that the sensitivity is equal to S. For this reason, Sis referred to as the sensi-
tivity function. In view of the important relationship in (14-16), T is called the complementary sensitiv-
ity function,

14.6.2 Effect of Feedback Control on Sensitivity

Next, we show that feedback reduces sensitivity by compariog the relative sensitivities for open-loop
control and closed-loop control. By definition, open-loop control occurs when the feedback control
loop in Fig. 14.1 is disconnected from the comparator. For this condition:

(! )
s sp OL
= ToL g, GeG (14-29)

Substituting ToL for Tin Eq.14-25 and noting that dToudG = Ge gives:
dTouToL = (dToL) _Q_ = Ge _Q_ = (14-30)
l
dG/G dG ToL GeG

Thus, the sensitivity is unity for open-loop control and is equal to S for closed-loop control, as índi-
cated by (14-28). Equation 14-lSa indicates that !SI < 1 if IG,Gpl > 1, which usually occurs over the
frequency range of interest. Thus, we have identified one of the most important properties of feedback
control: Feedback control makes process performance less sensitive to changes in the process.
As indicated in the previous section, it would be desirable to make S very small at all frequencies.
But this is not possible owing to the following integral constraint:

The Bode Sensitivtty Integral (Goodwin et al., 2001). Considera stable feedback control system with
an open-loop transfer functíon, GoL(s) = G,(s)G(s)e-es. Assume that G(s) has no poles to the right of
the imaginary axis. Then,

f 0
"' In IS(jw)I dw = O (14·31)

L
382 Chapter 14 Control System Design Based on Frequency Response Analysis

Thus, if S is small for a particular frequency range, it has to be large for other frequencies. This
phe- nomenon is sometimes referred to as the "waterbed effect" because pushing down on one part
of a wa- terbed causes another part to rise, Similar integral constraints are available for specific
situations, including systems with right-half plane poles and zeros (Skogestad and Postlethwaite,
1996).

14.6.3 Robust Stability

The robustness of a control system can be analyzed theoretically if the degree of process variability
(or model uncertainty) can be characterized. The uncertainty can be structured (for example,
parameter variations) or unstructured (for example, variations in the transfer function or frequency
response). In this section, we consider the irnportant concept of robust stability. The related concept
of robustperfor- mance is described elsewhere (Morari and Zafiriou, 1989; Skogestad and
Postlethwaite, 1996).
In robustness analysis, the unstructured uncertainty is often expressed as bounds on a
nominal process model. For exarnple, suppose that G is the process model used for control system
design, and G is the transfer function for the actual (but unknown) process. They can be related
by an unstruc- tured model uncertainty, li(s),
G(s) = (1 + ó(s))G(s) (14-32)
Rearranging Eq. 14-32 indicates that ó(s) can also be interpreted as the relative model
error:
(14-33)
i.\(s) = G(s) - G(s)
G(s)
Robustness analysis is typically based on a magnitude bound for the uncertainty
(14-34)
liiU(J))l :S e((J))

where e((J)) is a known function of frequency. The multiplicative uncertainty description in (14-32)
pro- vides the basis for an important robust stability result. The following theorem guarantees
closed-loop stability for a specified degree of model uncertainty.

Robust Stabtltty Theorem, Consider the feedback control system in Fig. 14.1 anda specific controller
G,. Assume that:
(i) The closed-loop system � stable for the nominal closed-loop system that consists of G, and
the nominal process model, G.
(ü) The magnitude of the unstructured uncertainty ó(s) in (14-33) is bounded by e((J)) in (14-
34).
Also assume that each G(s) in (14-32) that is generated by this uncertainty structure has
the
same number of right-half-plane potes.
Then the closed-loop system is robustly stable for all G that satisfy (14-33) and (14-34), if and only
if
- . 1
l T (J(J))I < e((J)) for ali w > o (14-35)

or equivalently,
I T(iw)le(w) < 1 for au (J) > o (14-36)

where T denotes the nominal complementary sensitivity function, T = G,G/(1 + G,G).

The proof for this theorem is available in textbooks on advanced control such as Bélanger (1995).
Sorne important consequences of this theorem are (Skogestad and Postlethwaite, 1996):
L The theorem has a graphical interpretation that is shown in Fig.14.17. Robust stability is guaranteed
if the uncertainty region at each frequency w, a circle with radius IGoLUw)[e(w), does not include the
critical point at (-1, O). Figure 14.17 shows the uncertainty region for one frequency; the circles for
the other frequencies have been omitted. In many practica! problems, not all of the perturbations al-
 14.6 Robustness Analysis 383
lmaginary
part

Figure 14.17 Graphical interpretation of the Robust

Stability Theorem.

lowed by Eq. 14-34 are possible. For example, upper and lower limits on model parameters such as a
gain or time constant produce uncertainty regions that are not circular, as they appear inFig.14.17.
For these situations, the robust stability theorem provides sufficient (but not necessary) conditions.
Consequently, the results tend to be conservative as will be demonstrated in Example 14.9.
2. In general, wh�n the uncertainty bound !(co) is large, the controller design must be more conserv-
ative so that I T(jw) 1 will be small enough to satisfy Eqs. 14-35 and 14-36.
In surnmary, if the conditions of this theorem are satisfied, robust stability is guaranteed for the as-
sumed process uncertainty description in Eqs. 14-32 to 14-34. The theorem also allows determination
of how much modeling error can be tolerated for a particular controller design, as is demonstrated by
the following example.

Consider the nominal process model of Exercise 12.2:

G(s) = -1.6(1 - 0.5s)

+ 1)
s(3s

The PID controller settings for the IMC method and Te = 1.25 are: K¿ = -1.22, -r1 = 6 min, and
w = 1.5 mio. This value of Te was chosen to provide a peak sensitivity value of 1.96 (Chen and
Seborg, 2002).
(a) Suppose that a small measurement time constan! Tm was neglected in developing the nomi-
nal model. Tbus, the actual process transfer function Gis gíven by
G(s) = G(s)
'T,,,S +1
Use the robust stability theorem to determine the largest value of 'Tm for which robust sta-
bility can be guaranteed.
(b) Repeat part (a) using an exact stability analysis.
(e) Do tbe answers of part (a) and (b) agree? If not, explain which estimate is more accurate
and why the discrepancy occurs.

i SOLUTION (a) The complementary sensitivity function for the nominal model and the PID controller is:

(-1.6(1- 0.5s))(-l 22)(l + .!_ + 1.Ss )

s(3s + 1) · 6s (0.1)1.Ss +1
1 T = G,G -

l
1 + GeG 1 + (-1.6(1 - O.Ss))(-1.22)(1 + .!. + 1.5s )
s(3s + 1) 6s (0.1)1.Ss + 1
384 Chapter 14 Control System Design Based on Frequency Response Analysís

ITüw)lt(w)

O.l'--..A..._....,._..............,�...._���������
0.01 0.1 10 Figure 14.18 Robustness results for
"'(rad/min) Example 14.9.

Simplifying gives:
-9.66s3 + 13.3s2 + 11.0s + 1.95
T = 2.75? + 9.24s3 + 19.3s2 + 11.0s + 1.95
Toe model uncertainty expression is given by (14-33):

A(s) = G(s)_- G(s) = -'T,,,s (14-37)

G(s) 'T,,.r +1
Toe bound f(w) in (14-34) can be set equal to the magnitude of A(s):

f(w) = jA(jw)I = 'TmW (14-38)

,/'Tt,,w2 +1
The robust stability theorem can be used to determine the maximum value of 'Tm for wbich
robust stability can be guaranteed. For example, a plot of IT(jw)le(w) vs. w can be con·
structed for a small value of 'Tm, and the condition in Eq. 14-35 can be checked to con.firm
that it is satisfied. Then 'Tm is gradually increased until (14-35) is no longer valid. Represen-
tative results for three values of 'Tm are showa in Fig.14.18. By additional tria! and error, the
maximum value oí 'Tm = 1.88 min.
(b) Toe characteristic equation is given by 1 + GcG = O. Substituting gives,

1 + (-l;�i; �)Ss) )(-1.22)(l + � + (o.1A.;: + 1)( 'Tm/+ 1) O

=

Toe Routh stability criterion of Chapter 11 can be used because the characteristic equation
is a polynomial in s. By tria! and error, the maximum value is found to be 'Tri! = 1.93.
(e) For tbis example, the robust stability theorem provides a slightly conservative estímate
owing to the suffícient-only condition that is a consequence of the structured model uncer-
tainty in (14-33). Even though the approximate value of 1.88 min is very close to tbe exact
value of 1.93 min, sucb excellent agreement does not always occur in other situations. •

SUMMARY
Frequency response techniques are powerful tools for the design and analysis of feedback control sys-
tems. The Bode and Nyquist stability criteria provide exact stability results for a wide variety of con-
trol problems, including processes with time delays. They also provide convenient measures of relative
stability, such as gain and phase margins. Closed-loop frequency response concepts such as sensitivity
functions and bandwidth can be used to characterize closed-Ioop performance.
Control system design involves tradeoffs between control system performance and robustness, Mod-
ero control systems are typically designed using a model-based technique such as tbe ones described in
 Exercises 385

Chapter 12. The robustness of a control system can then be analyzed using the concepts of the
sensitiv- ity and complementary sensitivity functions tbat were introduced in tbis chapter. The
robust stability theorem can be used to ensure that the control system can accommodate a specified
degree of model uncertainty, orto determine how much uncertainty can be tolerated before the
closed-lcop system be- comes unstable, Frequency response analysis is greatly facilitated by
software tools such as MATLAB and Simulink. which generate the required graphícal and numerical
results.

REFERENCES
Ástrom, K. J. and T. Hagglund, PID Controllers: Theory, Design,
and Tuning, 2d ed., ISA, Research Triang\e Park, NC, 1995.
Astréim, K. J., H. Panagopoulos, and T. Hagglund, Design of PI
Conlrollers Based on Non-Convex Optimizatioo, Automatica, 34,
585 (1998).
Bélanger, P. R., Control Engineering: A Modem Approoch, Saun-
ders College Pub., Fort Worth, TIC, 1995.
Brown, J. W. and R. V. Churchill, Complex Variables ond Appllca-
tions, 6th ed., McGraw-Hill, New York, 1996.
Cben, D., and D. E. Seborg, PIIPID Controller Design Based on
Direct Syntbesis and Disturbance Rejection, Ind. E11g. Chem.
Res., 41, 4807 (2002).
Doyle, J. C., B. A. Francis, and A. R. Tannenbaum, Feedback Con·
trol Theory, Macmillan, New York, 1992.
Franklin, G. F., J. D. Powell, and A. Emami-Naeioi, Feedback Con-
trol of Dynamic Systems, 4th ed., Preotice Hall, Upper Saddle
River, NJ, 2002.
Goodwio, G. C., S. F. Graebe, aod M. E. Salgado, Control System
Design, Prentice Hall, Upper Saddle Ríver, NI, 2001.
Hahn, J., T. Edison, and T. F. Edgar, A Note on Stability Analysis
Using Bode Plots, Chem. E11g. Ed11c. 35(3), 208 (2001).
Kuo, B. C., Automatic Control Systems, 7tb ed., Prentice Hall, En-
glewood Cliffs, NJ, 1995, Chapter 9.
Luyben, W. L., and M. L. Luyben, Essentials of Process Control,
McGraw-Hill, New York 1997, Chapter 11.
MacFarlaoe, A. O. J., Toe Development of Frequency Response
Methods in Automatic Control, IEEE Trans. Auto. Control, AC·
24, 250 (1979).
Maciejowski, J. M., Multivarlable Peedbock Design, Addiscn-Wes-
ley, New York, 1989.
Morari, M., and E. Zafíriou, Rabust Process Control, Prentice Hall,
Englewood Cliffs, NJ, 1989.
Skogestad, S., Simple Analytic Rules for Model Reduction and PID
Controller Tuning,J. Process Control, 13, 291 (2003).
Skogestad, S., and l. Postlethwaite, Multivariable Feedback Design:
An11/ysis and Design, Wiley, New York, 1996.
Tan, K. K., Q.-G. Wang, C. C. Haog, and T. Hagglund, Advances in
PJD Control, Springer, New York, 1999.

EXERCISES

14.1 In Chapter 12 we considered a stability analysis

based on substituting s = jw into the characteristic
equation. Demonstrate that this approach is (or is
not) ídentical to the Bode stability criterion. Hint:
Consider a special example such as GoL(s) =
Ke-8'/(-rs + 1).
14.2 A process that can be modeled as a time delay is con-
trolled using a proportional feedback controller. The
control valve and measurement device have negligí-
ble dynamics and steady-state gains of Kv = 0.5 and
Km = 1, respectively. After a small set-poínt change is
made, a sustained oscillation occurs, which has as pe-
ríod of 10 min.
(a) What controller gain is being used? 0.1 1 10 50
(b) How large is the time delay? "'(rad/min)
14.3 A Bode diagram far a process, valve, and sensor is
shown in Fig. E14.3.
(a) Determine an approximate transfer function far
thís system. <l>oL o�
(deg) -90º
(b) Suppose that a proportional controller is used -180º��
. .........a_

and that a value of K¿ is selected so as to pro· 0.01 0.1 1 10 50

vide a gain margin of 1.7. What is the phase w{rad/mln)
margin? Figure El4.3
386 Chapter 14 Control System Design Based on Frequency Response Analysis

Q;

l
Figure E14.4

14.4 Consider the storage taok with sightglass in Fig. (a} U the PI controller is adjusted so tbat -r1 = 0.4
El4.4. The parameter values are R: = 0.5 min/ft2, min, what value oí K¿ will result in a phase mar-
R2 = 2 mio/ft2, A1 = 10 ft2, Kv = 2.5 cfm/mA, gin of 45º?
A2 = 0.8 ft2, Km = 1.5 mA/ft, and -rm = 0.5 min. (b) If the controller settings in part (a) are used, what
(a) Suppose that R2 is decreased to 0.5 min/ft2. Com- is the gain margin?
pare the old and new values of the ultimate gain (e) If the Ziegler-Nichols settings for a PI controller
and the critica! frequency. Would you expect the are used, estímate Mr, the maximum amptitude
control system performance to becorne better or ratio for the complementary sensitivity funcrion,
worse? Justify your answer. T(s).
(b) If PI controller settings are calculated using the
Ziegler-Nichols rules, what are the gain and w (rad/min) AR 4> (deg)
phase margins? Assume Ri = 2 min/ft.

•
14.S A process (including valve and transmitter) has the 0.1 2.40 -3
approxirnate transfer function, G(s) = 2e-0.2sf(s + 1)
0.10 1.25 -12
with time constant and time delay in minutes. Deter- 0.20 0.90 -22
mine PI controller settings and the corresponding 0.5 0.50 -41
gain margins by two methods: 1.0 0.29 -60
(a) Direct synthesís (-re= 0.3 min). 2.0 0.15 -82
(b) Phase margin = 40º (assume -r1 = 0.5 ruin). 5.0 o.os -122
(e) Simulate these two control systems for a unit step 10.0 0.017 -173
change in set point. Which controUer provides 15.0 0.008 -230

• •
the better performance?

14.6 Consider the feedback control system in Fig. 14.1, 14.8 For the process in Exercise 14.S, the measurement is to
and the foUowing transfer functioos: be flltered using a noise filter with transfer function
G¡:(s) = 1/(0.ls + 1). Would you expect thís change to
2 result in better or worse control system performance?

Ge= K{0�/+\) G; = 0.5s +1 Compare the ultimate gains and critical frequencies
with and without the filter. Justify your answer.
G _ 0.4 3
P - s(5s + 1) Gd = 5s + 1 14.9 The block diagram of a conventicnal feedback con-
o;» 1 • trol system contains the following transfer functions:

(a) Plot a Bode diagram for the open-loop transfer
function.
(b) Calculate the value of K¿ that provides a phase
margin of 30º.
(c) Wbat is the gain rnargin when K¿ = 10?
14.7 Frequency response data for a process are tabulated
below. These results were obtained by íntroducing a
sinusoidal change in the cootroller output (under
manual control) and recording the measured re-
sponse of the controlled variable. This procedure was
repeated for various frequencies,
Ge = Kc(1 + is) o, "" 1
1
Gm = s + 1
(a) Plot the Bode diagram for the open-loop transfer
function.
(b) For what values of K; is the system stable?
(e) If K¿ = 0.2, what is the phase margin?
(d) What value of K¿ will result in a gaín margin of l.7?
 Exercises 387

Steam (a) Plot the Bode diagram for the transfer function,
G = GvGpGm.
(b) Designa PI controller for this process and sketch
p
the asymptotic Bode diagram for the open-loop
transfer function, GoL = G,G.
(c) Analyze the stability of the resulting feedback
control system.
(d) Suppose that under open-loop conditions, a sinu-
soidal set-point change, Jsp(t) = 1.5 sin(O.St), is in-
troduced. What is the amplitude of the measured
output signa! Ym(t) that is also sinusoidal in nature?
(e) Repeat the same analysis for closed-loop
Process fluid T conditions,
w,T¡ (f) Compare and discuss your results of parts (d)
and (e).
14.U Hot and cold liquids are mixed at the junction of two
Condensate ríii) pipes. Toe temperature of the resulting mixture is to
� be controlled using a control valve on the hot stream.
The dynamics of the mixing process, control valve,
and temperature sensor/transmitter are negligible
and the steady-state gain is 6 mA/rnA. Because the
Figure E14.10
temperature sensor is located well downstream of the
junction, an 8 s time delay occurs. There are no heat
losses/gains for the downstream pipe.
14.10 The dynamic behavior of the heat exchanger shown (a) Draw a block diagram for the closed-loop system.
1111\ in Fig. E14.10 can be described by the following (b) Determine the Ziegler-Nichols settings (continuous
� transfer functions (H. S. Wilson and L. M. Zoss, ISA cycling method) for both PI and PID controllers.
J., 9, 59 (1962)): (e) Por each controller, simulate the closed-loop re-
sponses for a unit step change in set point.
Process:
(d) Does the addition of derivative control action
T' 2 ºF/lb min provide a signi.ficant improvement? Justify your
'Wr = (0.43� + 1)(0.017s + 1) answer.
Control valve: 14.13 Consider the control problem of Exercise 14.10 and a
X' = 0.047 in/psi ¡íii) PI controtler with K¿ = 5 and 1'/ = 0.3 ruin.
w:=112�
� (a) Plot the Nyquist diagrarn for the open-loop system.
e; 0.083s + 1 X' rmn m
(b) Determine the gain margin frorn the Nyquist plot.
Temperature sensor-transmitter:
14.14 Considera process model,
P]- = 0.12 psil°F
T' 0.024s + 1 - IOe-s
G(s) = 2s + l
The valve lift x is measured in inches. Other symbols
are defined in Fig. E14.10. and a PI controller with the Hagglund-Ástrom set-
(a) Find the Zíegler-Nichols settings for a PI con- tings of Table 12.4. Use the robust stability limit of
troller. Eq. 14-35 to determine how robust this control system
(b) Calculate the corresponding gain and phase is to the assumed va!ue of the process gain. Assume
margíns, that the time constan! and time delay do not change.
14.11 Consider a standard feedback control system wíth the 14.15 For the process in Exercise 14.5, a PI controller is de-
• following transfer functions: signed so that the peak value of the complementary
-10 sensitivity function is Mr = 1.25. Use the robust sta-
Gm = e-0.5s G, = s +l bility limit of Eq. 14-35 to estimate the values of the
time constant for which closed-Ioop stability is guar-
G 1.5
P = lOs + 1 anteed. Assume that the process gain and time delay
do not cbange.