Sei sulla pagina 1di 19

An Interpretative Solution to Set-17

Control Engineering Course Proje t


Aditya Gaonkar ( EE11B005 )
Surajkumar H ( EE11B075 )
Manikandasriram ( EE11B127 )
November 27, 2013

Abstra t

This is an interpretative take on the set-17 problems, as part of the Control Engineering Course
proje t. The questions ranged from as simple as plot mat hing, to verifying stability riteria, undershoot
derivations, and ontrolling for es. Workloads were split, dierent lines of reasoning pursued, and we
hammered out neat and elegant solutions to the problems here, under reasonable assumptions. A detailed
look at the questions, and their solutions follows below.
Contents
1

Appetizers

Assorted Sandwi hes

2.1

Cu umber-Butter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Ham Brie, Apple, and a Ra e to the Finish. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3

2.2.1

Steady State Position Error Tuning

2.2.2

Root Lo us of Open Loop System

2.2.3

Minimum Settling Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Grilled Shrip-Ham, Margins, and Stability Thrills . . . . . . . . . . . . . . . . . . . . . . . . .

11

2.3.1

Routh-Hurwitz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

2.3.2

Frequen y Domain - The Nyquist Approa h . . . . . . . . . . . . . . . . . . . . . . . .

12

2.3.3

Frequen y Domain - The Bode Approa h, Troubled Waters

13

Main Course - A Pendulum, A ollar, and a mandate to not os ilalte

14

3.1

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

15

The equations governing the system


3.1.1

. . . . . . . . . . . . . . .

Motion of the ollar

3.2

Choosing the state spa e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.3

Finding the for e

F = k

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Takeaway - On Undershoot and a Proof


4.1

4.1.1
4.1.2
4.2

4.3

4.4

On G (s)

17

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

On G (s)

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Observations on the step response

A small derivation

15
15

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

17
17
18

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18

4.2.1

Initial slope of G (s) step response . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18

4.2.2

Initial slope of G (s) step response . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18

A denition of Undershoot and some small derivations . . . . . . . . . . . . . . . . . . . . .

18

4.3.1

Framing our system is a desirable manner.

. . . . . . . . . . . . . . . . . . . . . . . .

18

4.3.2

The values of S and D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

The Final Frontier.

1
2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

Appetizers

We were asked to mat h the following set of Pole-Zero plots with their orresponding Bode plots.

For a quantitative solution, we approa hed the problem by estimating an approximate transfer fun tion

G(s)

for the system from its Pole-Zero plot. Using this, we obtained the bode plot using MATLAB and then

subsequently mat hed it with the set of plots given.


An approximate transfer fun tion for pole-zero plot A is

GA (s) =

(s + 10)
s(s + 90)

The Bode plot for this system an be obtained using the following MATLAB ode snippet:

num = [1 10;
den = [1 90 0;
Ga = tf(num, den);
bode(Ga);
This produ es the following bode plot whi h an then be seen to mat h with Plot 3 in the given set (shown
in Figure 1).

Figure 1: Bode plot for G (s) and the orresponding mat hing bode plot.

The above on lusion an also be made qualitatively by analysing the Pole-Zero plot. Due to the presen e
of a pole at the origin, the magnitude plot should have20db slope for low frequen ies. The presen e of a

zero at

s = 10

auses an in rease in slope by

20db

whi h produ es a net slope of

0db

between the zero and

the se ond pole. The magnitude plot again en ounters a net20db slope for higher frequen ies due to the

ombined ee t of the two poles and a zero. Similarly, the phase plot is also a sum of the individual ee ts

of the poles and zero. For very low frequen ies, the pole at origin dominates and for es the phase to tend to
900. The presen e of a zero pulls up the phase towards 00 and as frequen y in reases, the other pole pulls
0
the phase ba k down towards90 .
Using a similar approa h, the rest of the plots were mat hed and the results are tabulated as follows.
Approximate G(s)

Bode Plot

(s+10)
s(s+90)

Mat hing Bode Plot

Approximate G(s)
1
(s+100+i10)(s+100i10)

Bode Plot

(s+300)
(s+40)

(s+10)
s(s+20+i100)(s+20i100)

(s+10)
(s+150)

(s+8+i60)(s+8i60)
(s+1+i10)(s+1i10)

Mat hing Bode Plot

Assorted Sandwi hes

2.1

Cu umber-Butter

We were asked to mat h the following Pole-Zero plots with their orresponding Step response plots

For a quantitative solution, we approa hed the problem by estimating an approximate transfer fun tion
of the system from its Pole-Zero plot. Using this, we obtained the Step response plot using MATLAB and
then subsequently mat hed it with the set of plots given.
An approximate transfer fun tion for the Pole-zero plot would be

GA (s) =

1
(s + 1 i10)(s + 1 + i10)

The Step response plot for this system an be obtained using the following MATLAB ode snippet:

num = [1;
den = [1 2 101;
G = tf(num, den);
step(G);
This produ es the following step response plot whi h an then be seen to mat h with Plot 3 in the given set
(shown in Figure 2).

Figure 2: Step response plot for

GA (s)

and the orresponding mat hing plot

The above on lusion an be justied qualitatively as follows. The system represents a se ond order system
with imaginary onjugate poles near the

axis. Comparing the transfer fun tion to the standard form, we

2
get 0.1 and n 10. This implies that it is an underdamped system with Mp = e 1 = 73.04% and
1
the steady state value is Lims0 sG(s) = 0.01. Hen e the orresponding step response plot should be Plot
s
3.
Using a similar approa h, the step responses of the remaining systems were evaluated and the results are
tabulated below.
Approximate G(s)

Step Response

1
(s+1i10)(s+1+i10)

s+300
s+25

Mat hing Step Response

Approximate G(s)

Step Response

s10
(s+20+i20)(s+20i20)

1
(s+5)(s+2+i20)(s+2i20)

(s+40)
(s+300)

s
(s+1+i10)(s+1i10)

Mat hing Step Response

2.2

Ham Brie, Apple, and a Ra e to the Finish.

The given open loop transfer fun tion is

Gc (s)Gp (s) = H1 (s) =

s3

+ (1 +

(s + )
+ ( 1)s + (1 )

)s2

It is embedded in a unity feedba k loop.

2.2.1

Steady State Position Error Tuning

It is required to tune

su h that the steady state position error for a step input is

the steady state error for the step input in terms of a

Now we know that

2.2.2

Kp

requirement as

Steady State Error

Kp

Kp = lims0 H1 (s) = lims0

< 10%.

We an write

1
0.1
1 + Kp
9

(s+)
s3 +(1+)s2 +(1)s+(1)

9 = 0.9

Root Lo us of Open Loop System

This is basi ally nding the variation of open-loop poles with

In order to ompute this, we require the

Chara teristi equation of the Open Loop Transfer expressed in a ni e form.


3
2
3
2
2
So we take the denominator of H1 (s) whi h is s +(1+)s +(1)s+(1) = (s +s s+1)+(s +s1).
(s2 +s1)
We an now dene a new polynomial H2 (s) = 1 +
(s3 +s2 s+1) . We easily see that the roots of H1 (s) are

also the roots of

H2 (s).

We now plot the root lo us of

H2 (s)

using the

of the variation of the open-loop poles with

rlo us() fun tion in MATLAB. This will give us a fair idea

Figure 3: Open Loop Root Lo us

Note that we require open loop root lo us to see the variation of open loop poles with
system performan e, we still need to nd the losed loop root lo us

To observe

To do this, we note that it is in a unity feedba k stru ture

H(s) =

H1 (s)
1+H1 (s) .The losed loop poles are the
0 = s3 +(1+)s2 +s+1 = 0. As in

1+H1(s) = 0 = 1+ s3 +(1+)s2(s+)
+(1)s+(1) =
3
2
2
the previous ase, we an group the terms together to bring as a parameter to give (s +s +1)+(s +s) = 0.
2
s +s
We an now dene H3 (s) = 3
s +s2 +1 , giving 1 + H3 (s) = 0 as the modied hara tristi polynomial. We
roots of the equation

an now plot the losed loop root lo us using the rlo us() ommand in MATLAB.

Figure 4: Closed Loop Root Lo us

The interse tion with the real-axis is

0.469 and is given by = 4.48. We do


= 4.48, with the other pole being at

dominant pole assumption does hold at

note additionally that the


greater than a fa tor of 5

away. Note that this ase yields a riti ally damped system under dominant pole assumptions

2.2.3

Minimum Settling Time

It is desired to nd the value of alpha giving the minimum settling time.
4
nd
We do know for a 2 order system that the 2 Settling time is =
n . So assuming we an approximate
the system to a 2nd order one, we would expe t the minimum settling timeto o ur at the maximal value of

n ,

whi h we know to be the real interse tion,

= 4.48,

whi h we plotted as below.

What did surprise us though, was running some ba kground tests to verify this fa t. If indeed

= 4.48

was the best settling time response, we should expe t surrounding values of alpha to settle later, that is have
larger settling time. To he k this, we made a qui k and dirty MATLAB s ript to plot the step response for
surrounding values of alpha, stored in the array.

al=[4.9 4.48 4.2 4.0 3.95 3.9 3.87 3.85 3.82 3.8 3.75 3.7 3.6 3.5 3.3 3.1 2.9
2.5 1.9
for i=1:length(al),
ad=al(i) b=tf([1 ad,[1 1+ad ad-1 1-ad) step((b/(1+b))) hold on
end
legend(strtrim( ellstr(num2str(al'))'))
Whi h is plotted as below.

Figure 5: Step responses for various values of

10

For values of

> 4.48,

it behaves as expe ted, sin e the system is underdamped giving a longer step

response. What we didn't expe t that for values of

just < 4.48,

the response seemed to speeden up. Note

that this orresponds to a slightly underdamped system, with the roots having both a real and an imaginary
part.
The best settling time a ording to pure observation seemes to be around

3.8,

whi h we veried using

a zoomed in graph, and MATLAB's own settling time al ulator, whi h gave a settling time of

= 3.82.

For values of

< 3.82,

11.3s

for

settling time began to gradually in rease, ultimately giving rise to an

overshoot system. The system be am more os iallatory as we de reased

whi h again is expe ted as the

poles get loser to the imaginarry axis. The zoomed in plot is shown below.

Figure 6: Zoomed in Graph

2.3

Grilled Shrip-Ham, Margins, and Stability Thrills

1
(s+1)3 system, embedded in a unity feedba k loop. We are required to
nd the stabilizing gain using several methods, and give small adjustments to ea h to make these shake-hands.

In this question, we are given a

2.3.1

G(s) =

Routh-Hurwitz

To nd stability by Routh-Hurwitz, we need the hara teristi equation. This is realised using a feedba k
KG(s)
H(s) = 1+KG(s)
= (s+1)13 +K . So the Chara teristi
equation is
3
2
loop and a generi gain term K. The transfer fun tion is

Characteristic Equation = s + 3s + 3s + (K + 1)

We want to nd the range of K su h that this remains stable. This an be done with a Routh Table as
below

Table 3: Routh Stability Derivation


s3
1
3
0
s2
3
1+K
0
8K
s
0
0
3

3(1+K)

We see know that Routh stability of the system is interpreted as the rst olumn in the above table having
no sign hanges ( no right half plane poles ). This is a hieved if

11

0<K <8

Thus the range of proportional gain, using Routh-Hurwitz was found to be


The limit of proportional Gain while maintaining stability is

2.3.2

0<K <8

K=8

Frequen y Domain - The Nyquist Approa h

The next part of the question was to nd the proportional gain using the on ept of Gain and Phase Margin.
The rst was to obtain this interpretation is to use the mother of all frequen y stability riteria, the Nyquist
plot. We an obtain the Nyquist plot using MATLAB, and it is shown below.

Figure 7: Nyquist Plot

It an easily be seen that the Nyquist plot interse ts the real axis at0.124. To arrive at the same value

theoreti ally, we onsider the fun tion

1
(1 j)3
(1 3 2 ) + j(3 + 3 )
=
=
3
2
3
(j + 1)
(1 + )
(1 + 2 )3

intersea tion, we set Imaginary part to zero, yielding =


3, 0.
G(j) =

T oobtain the real


putting ba k in

G(j),

3,

and

we get

G(j) =

8
(1 3 3)
=
= 0.125
(1 + 3)3
64

We see that the theoreti al estimate of the real axis interse tion is
Now we go to the Nyquist stability riterion, whi h
en ire lements about

Taking

s = 1, P

N = P Z,

a,a = 0.125.

where N is the number of lo kwise

is the number of open-loop poles in the right half plane, and Z is the number

of losed loop zeroes in the RHP.


We easily see that for
know that

Z = 0.

G(s),

there are no RHP poles, so

This an only be enfor ed if

N = 0,

P = 0.

In order for the system to be stable, we

i.e there are no en ir lements about

s = 1.

If we a gain term K, the nyquist plot gets amplied by K. The new Real axis inerse tion is at

Ka, a = 0.125.

We an keep in reasing the gain K until

number of en ir lements, giving rise to an unstable system.


1
Thus for stability, Ka = 1, or K =
a = 8.
Note that a Gain K = 8 orresponds to a gain margin of

Ka = 1,

18.06 dB

This seems to agree with the Routh-Hurwitz derived idea, bolstering the
harge forth into the Bode derivation of the same.

12

s =

as beyond this there are a non-zero

K =8

idea. With this, let us

2.3.3

Frequen y Domain - The Bode Approa h, Troubled Waters

We an now target the Bode plot approa h to the same problem.

We approa h stability from the Bode

diagram, using Gain margin to indi ate stability. We know that Gain margin is a measure of stability, of
how mu h proportional gain we an add until the system starts to be ome unstable.
MATLAB has the

margins() fun tion, whi h reated a bode plot indi ating the Gain and Phase margins.

We an also pass these values to variables as will be done shortly. The bode plot obtained is shown below

Figure 8: Bode Plot

The Gain margin given to us by MATLAB is

18.1 dB .

Taking

20 log(K) = 18.1 dB ,

we get

K = 8.035,

whi h is quite lose to our original value. Sin e these plots tend to round-o, we aptured these values in
Variables.
[g,p,wg,wp=margin(gh)
g = 8.0011
p = -180
wg =
1.7322
wp = 0
It is learly seen that the Gain K required to bring the system to the edge of stability is

K = 8.0011,

whi h is very lose to the estimated value in the other two. So, by various methods, we have arrived at the
same proportional gain range..

Range of P roportional Gain K [0, 8]


Note that we used the bode plot used given by MATLAB. A hand-drawn Asymptoti bode plot would
give a very dierent answer indeed.

13

Main Course - A Pendulum, A ollar, and a mandate to not


os ilalte

3.1

The equations governing the system

The system looks a bit ompli ated with the two springs and the ollar and all.

However here we did a

simple minded analysis worrying ourselves about the angular motion of the rod as it will be unae ted by
the motion of the ollar! The motion of the ollar is a bit ompli ated and we just give the equations here.
Hen e we essentially designed a simplisti model for the required for e F by taking

kx

in

F = kx x + k

to

be zero.
Angular motion of the rod
Let I be the moment of inertia of the rod. Let
respe tively. We take

and be the angular position, velo ity and a elration

to be measured anti- lo kwise from the verti al dire tion to be positive.

Thus the equation for the angular a eleration be omes

I = F L bv
Under the small angle approximation, we an write

mgLsin()
2
sin() approximately

as

and then the equation

be omes

I = F L bv

mgL
2

(1)

Also letting T be the tension in the rod whi h arises due to the intera tion of the ollar and the rod.
Then applying the entripetal for e ondition to the rod we get (taking at the entre of the mass of the rod):

m 2 L
= T mgcos()
2

14

3.1.1

Motion of the ollar

The only useful equation available is the ondition along the horizontal. Thus we get

2kx = T sin()
3.2

Choosing the state spa e

We hoose

and

as the states of the system thus from equation (1) we get the state representation as

X =

 


 
1
0
X
+
L F
bIv
I

0
mgL
2I

The other equations involving x and T are not of mu h interest in our analysis as we have taken F to
depend only on

3.3

Finding the for e

F = k

We now set that F=KX where K is the row ve tor


0 .

Here we have set the se ond term as zero sin e

. Inserting this in our state spa e formulation




 



0
1
0
0 X
X = mgL
bv X + L k

I
2I
I

we dont want the for e to depend on

we get

(2)

Thus after multiplying out the matri es and adding them we get

0
X = mgL
+
2I

0
A = mgL
+
2I

1
bIv

k L
I


1
X
bIv

(3)

. The eigenvalues of A are the poles of this losed loop system. solving for
k L
I
|sI-A|=0 we get the quadrati in s as
Let

s2 + s

bv
mgL k L
+

=0
I
2I
I

(4)

As we need the system NOT TO OSCILLATE, this means we need the poles to be purely real. This leads
to an overdamped system, and demands the dis riminant of (4) to be positive. Thus we get

15

mgL k L
b2v
4(

)
I2
2I
I
Thus solving for

k we

get

b2
mg
v
2
4IL

Therefore if we need the system to be riti ally damped we an hoose

k =

mg
b2
v
2
4IL

k >

mg
b2
v
2
2IL

And if we need overdamping then

or any su h value whi h satises the inequality onstraint.


Note that we did not feature the for e as

F = kx x + k .

This is be ause the ollar ontraint auses

x and

to be linked together. We annot vary one without varying the other, that is they are not independently

ontrollable. Any hange in


equations.

is modelled as a hange in

with the additional help of the Tangential for e

Takeaway - On Undershoot and a Proof

For this question we used the ommand step(sys) in MATLAB to get the step response. To reate the transfer
fun tion we use the ommand tf(num,denom). Hen e this problem is a simple appli ation of the 2 ommands
over the given transfer fun tions.

4.1

On G1 (s)

Given

G1 (s) =

(1s)
(1+s)(1+ 2s ) it an be written as

G1 (s) =

(1s)
0.5s2 +1,5s+1 . Next in the MATLAB ommand line I

s=tf([-1,1,[0.5,1.5,1) and this reates the system fun tion s. Next I give the ommand
step(s) and it produ es the graph of the step response as the output. The graph obtained was

type the ommand

4.1.1
Given

On G2 (s)

G2 (s) =

(1s)(1 2s )
(1+s)(1+ s2 )(1+ 3s ) it an be written as

G2 (s) =

3s2 9s+6
s3 +6s2 +11s+6 . Next in MATLAB ommand line I

type the ommand s=tf([3,-9,6,[1,6,11,6) and this reated the system fun tion s. Next I give the ommand
step(s) and it produ es the graph of the step response as the output. The graph obtained was

17

4.1.2

Observations on the step response

G1(s)

For

we see that the step response initially de reases in magnitude before in reasing and settling in a

steady steady state after hitting a global minimum. This phenomenon is alled as an undershoot response
whi h is analysed in some detail in the last part of the Takeaway se tion. For

G2(s)

we see that the step

response rst in reases then de reases and nally in reases again to a nal steady value.

4.2

A small derivation

The Initial Value theorem states For a given fun tion


value of

f(0) is given by

limt0 f (0) = lims sF (s).

f(t) with its one sided Lapla e transform F(s) the


H(s) the

Thus for a system with transfer fun tion

H(s)
s .

As we are looking at the derivative of the step response


H(s)
the Lapla e transform of the derivative of the step response is s
whi h is H(s) again. Thus we get
s

limt0 0f (0) = lims sH(s).


lapla e transform of the step response is

4.2.1

Initial slope of G1 (s) step response

s-> sG1 (s).

Thus
(1s)
0.5s2 +1,5s+1
Simplifying this we get that g '(0)=-2.

We get g '(0)= lim

s->

g '(0)= lim

As the initial slope i.e. the derivative of the step response at t=0 is negative we see that the step response
de reases initially.

4.2.2

Initial slope of G2 (s) step response

2 (s). Thus
s-> sG
3s2 9s+6
s-> s s3 +6s
2 +11s+6
Simplifying this we get that g2 '(0)=3.
2

We get g '(0)= lim

g '(0)= lim

As the initial slope i.e. the derivative of the step response at t=0 is positive we see that the step response

in reases initially.

4.3

A denition of Undershoot and some small derivations

We read through the paper mentioned in the problem statement and in it the denition of undershoot for a
proper and stable system was given as the ondition where the steady state value of the step response and
rst non zero derivative (whi h will be of the order n-m where n is the number of poles and m is the number
of zeros) at time t=0 will both be non zero and have opposite signs. We have used this denition in solving
this problem.
Next is proving 2 important formulae we will use in this problem.

Let's onsider a proper and stable

system with a transfer fun tion H(s) and with m zeros in total and n poles and m<n. Let S be the steady
H(s)
and the value of
state value of the step response. Then the Lapla e transform of the step response is
s
H(s)
S is given by S=lim
s
H(s). This gives us the steady state value of the step
s whi h gives us S=lim
response.

s->0

s->0

th derivative of the step response is given by sr U (s) where U(s)

Let r=n-m. The Lapla e transform of the r

H(s)
is the Lapla e transform of the step response. We know that U (s) =
and thus the Lapla e transform of
s
r1
U (s). Let us denote the value of this derivative at time t=0 by D. Then by
the r derivative is given by s
r
the initial value theorem the value of D is given by D=lim
s H(s).

th

s->

4.3.1

Framing our system is a desirable manner.

We an ast the transfer fun tion of our system in almost any form. Here we will put it in a form whi h
separates out the onjugate pairs of poles and zeros ( we onsider real valued systems) and then the real
valued poles whi h lie in the left half plane and the real valued zeros whi h an lie on both right and left half

plane. Note that in this analysis we dont onsider poles to lie on the imaginary axis and origin and the zeros
to lie at the origin. Thus our transfer fun tion will have the following form:

H(s) = K

Q
Q
(s zi )(s zi ) (s zj ) (s + zk )
Q
Q
(s pm )(s pm ) (s + pn )

Where i ranges over a whi h is the number of onjugate pairs of zeros, j ranges over b, the number
of positive real zeros, k ranges over , the number of negative real zeros, m ranges over d, the number of
onjugate pairs of poles (whi h lie of the left half plane) and n ranges over e, the number of negative real

j, zk, pn are all positive numbers and d+e>a+b+ .

poles. Also z

4.3.2

The values of S and D

In the rst part of this se tion we gave the formulae to evaluate S and D. We have the transfer fun tion of
the system dened. Hen e we an give expressions for these quantities.

Q
Q
Q
(szi )(szi ) (sz
Qj ) (s+zk ) . This
K Q(sp
The expression for S was S=lim
H(s) whi h gives us S=lim
)(sp
)
(s+p
m
m
n)
Q
Q
Q
2
( |zi |Q
)(1)b ( Qzj )( zk )
redu es to S = K
. Here note that ea h produ t appearing in the formula is a positive
( |pm |2 )( pn )
quantity. Thus the sign of S will depend on the sign of K and on whether b is odd or even.

s->0

s->0

Let r=d+e-a-b- where d+e represents the total number of poles and a+b+ represents the total number
r
s H(s) whi h gives the below formula

of zeros. The expression for D was D=lim

s->

D = lims> sr .K

Q
Q
(s zi )(s zi ) (s zj ) (s + zk )
Q
Q
(s pm )(s pm ) (s + pn )

We an see that the degree of the numerator produ t terms is a+b+ and it is further in reased by
r=d+e-a-b- . Thus the total degree of the numerator be omes d+e whi h is the same as the degree of the
denominator. Also on e multiply out the numerator and denominator terms we will see that the oe ient
of s

d+ewill be 1.

4.4

As we take limit s-> the value of D be omes K.1=K. Thus D=K.

The Final Frontier.

We have the values of S and D whi h are the steady state step response value and the value of the rst non
zero derivative of the step response at time t=0. The ondition for undershoot is that S and D should be
non zero and should have opposite signs. Thus mathemati ally this ondition be omes SD<0
Q
Q
2
b Q
2 ( |zi |Q)(1) ( Qzj )( zk )
Substituting for S and D the ondition be omes K
< 0. Here note that K and
2
( |pm | )( pn )
ea h produ t that appears has been ensured to be positive. Thus the inequality is satised when b is odd

where b is the number of real zeros on the right half plane. Thus we see that a stri tly proper and stable
system has undershoot response when it has odd number of real zeros on the right half plane. Q.E.D

19

Potrebbero piacerti anche