Second order transfer functions 1
w t norris 2nd order transfer functions Monday 3 March 2008
2nd Order transfer function - Summary of results
The canonical 2nd order transfer function is expressed as
H(s) =
n
2
s
2
+ 2
n
s +
n
2
1
n
is the natural frequency; is the damping coefficient.
2
The Poles, s
1
, s
2
are
s
1
,s
2
=
n
( )
2
1 3
On the right is the root locus for fixed
n
and varying .
s
j
0
sin
0
<1
<1
>1 >1
1
j
n
-
n
j
n
-j
n
The transfer function in the frequency domain is
H() =
n
2
n
2
w
2
+ 2
n
w
=
n
2
(
n
2
2
)
2
+ 4
2
n
2
2
e
j
with tan =
2
n
n
2
2
The Bode plot is :
For =
1
2
we have the sharpest
corner without a maximum.
For <
1
2
there is a peak in the plot of
magnitude
|H|
max
=
1
2
1
2
5
This is above the level at = 0.
For the H(s) given at t= 0 then |H| = 1
The roll-off rate at high frequencies
is 40 dB per decade ( in frequency).
The overall phase change is -180.
For low the swing in phase is all
close to the frequency at the maximum
|H|.
The peak itself is at
p
=
n
1 2
2
Beware; later we introduce
d
=
1
2
-0.1 1 10
-40
-30
-20
-10
0
10
H, dB
Frequency, rad s
-1
2
0.1
1
2
0.01 0.1 1 10
-180
-135
-90
-45
0
50
Frequency, rad s
-1
( log scale )
Phase,
degrees
2
1
2
0.1
Second order transfer functions 2
w t norris 2nd order transfer functions Monday 3 March 2008
Impulse Responses
= 0 Undamped
y(t) =
n
sin(
n
t)
1/
n
-1/
n
y(t) 2/
n
> 0 Damped
y(t) =
1
2
n
2
1
( ) exp(s
1
t) exp(s
2
t)
1 Critically damped
y(t) =
n
2
t exp( -
n
t)
Curve is very similar to damped response
2 3 4 5
-0.3
-0.2
-0.1
0.1
0.2
0.3
t
y(t)
exp( s
2
t)
exp( s
1
t)
< 0 Underdamped
y(t) =
n
2
d
exp(-
n
t) sin(
d
t)
The logarithm of the ratio of successive
peaks is the logarithmic decrement or log-
dec for short
=
2
1
2
2 if << 1
5
10
15 20
25
-0.6
-0.4
-0.2
0.2
0.4
0.6
0.8
t
y(t)
Second order transfer functions 3
w t norris 2nd order transfer functions Monday 3 March 2008
Step responses
= 0 Undamped
y(t) =
( )
1 cos(
n t
)
2 4 6 8 10
1
2
t
y(t)
> 0 Damped
y(t) =
_
1 +
1
2
1
[ s
2
exp(s
1
t) s
1
exp(s
2
t) ]
10 15
0.2
0.4
0.6
0.8
1
t
y(t)
1 Critically damped
y(t) =
( )
1 exp(
n
t ) [1 +
n
t]
2 4 6 8 10
1
t
y(t)
< 0 Underdamped
y(t) =
'
1 exp(
n
t)
_
cos(
d
t) +
d
sin(
d
t)
First peak
*
,
P
= 1 + exp
1
2
Proportional overshoot:
P
= y 1 = exp
1
2
Time to be within , T
s
=
ln()
n
For = 0,02, i.e., 2% : T
s
4
n
Root Total square error =
1+ 4
2
4
n
* Some authors call
M
P
the overshoot .
10 20
1
t
y(t)
M
P
T
p
T
s
Overshoot,
O
P
T =
2
Period,
Frequency of oscillation:
d
=
n
1
2
Period of oscillation:
T =
2
d
=
2
n
1-
2
For least root square total deviation = 1 / 2.
Then O
P
16%
Second order transfer functions 4
w t norris 2nd order transfer functions Monday 3 March 2008
Ramp response Now set f(t) = t as
illustrated.
F =
1
s
2
f= t
t
Perhaps unkindly I leave derivations to the reader. The general response is:
Y = H(s)
1
s
2
=
n
2
s
2
(s
2
+ 2
n
s +
n
2
)
= 0 Undamped
Y =
_
1
s
2
1
s
2
+
n
2
y =
_
t
sin(
n
t )
n
Plot is for
n =1
; y = t sin(t)
Note: s
1
s
2
=
n
2
; s
1
+ s
2
= 2
n
2 4 6 8 10
2
4
6
8
10
t
y(t)
Response
y=t
> 0 Damped : s
1
, s
2
=
n
( ) -
2
1
Y =
]
1
1
1
1
s
2
+
s
1
+s
2
s
n
2
+
1
s
1
s
2
_ s
2
s-s
1
s
1
ss
2
y =
]
1
1
1
t
2
n
+
1
s
1
s
2
( )
s
2
e
s
1
t
s
1
e
s
2
t
Plot is with
n
= 1, = 2
2 4 6 8 10
-4
-2
2
4
6
t
y(t)
Response
t -
n
2
= 1 Critically Damped
Y =
1
n
s
2
2
s
+
2
s+
b
+
n
(s+
n
)
2
y =
1
n
( )
n
t 2 + (2 +
n
t ) e
n
t
The plot is for
n
= 1.
6 8 10
-2
2
4
6
8
Y(t)
t
Response
2 4
t -
n
2
Second order transfer functions 5
w t norris 2nd order transfer functions Monday 3 March 2008
< 0 Underdamped:
d
=
n
1
2
Y =
1
n
s
2
2
s
+
2(s+
n
)
d
(12
2
)
d
(s+
n
)
2
s+
n
2
(1
2
)
y =
1
n
t 2+e
n
t
[2cos(
d
t)
d
(12
2
) sin(
d
t)]
The plot is with
n
= 1 and = 02
2 4 6 8 10
2
4
6
8
y(t)
t
Response
t -
n
2
Second order transfer function analysis
As an example consider the mass, damper and spring acted on by a force, f(t). We had for
the displacement, y :
m y
"
+ c y
'
+ k y = f
Take the Laplace transform ( with no initial values ) :
m s
2
Y + c s Y + k Y = F
or s
2
Y +
c
m
s Y +
k
m
Y = F/m
which with
n =
k
m
and =
c
2
km
becomes If we redefine F F/
n
2
m = F/k:
s
2
Y + 2
n
s Y +
n
2
= F
n
2
: or (s
2
+ 2
n
s +
n
2
) Y = F
n
2
and so H(s) =
F
Y
=
n
2
s
2
+ 2
n
s +
n
2
1
This equation 1 is a standard form for any second order transfer function. Whatever the
physical variables it helps to turn the expression to this format early in the analysis.
n
is the natural frequency; is the damping coefficient. 2
Poles
Poles, s
1
, s
2
are the values of s for which the denominator is zero.
s
1
,s
2
=
2
n
4
2
n
2
4
n
2
2
=
n
( )
2
1 3
Second order transfer functions 6
w t norris 2nd order transfer functions Monday 3 March 2008
Cases;
(1) = 0: poles at s = j
n
,
(2) < 1: two complex poles. These lie on a circle of radius
n
(3) = 1: two coincident poles at s =
n
(4) > = 1: two poles on the negative real axis.
These are illustrated on the root locus diagram on the right for
constant
n
.
s
j
0
0
<1
<1
>1 >1
1
j
n
-
n
j
n
-j
n
Second order transfer functions 7
w t norris 2nd order transfer functions Monday 3 March 2008
Bode Plot
From eq. 1 H(s) =
n
2
s
2
+ 2
n
s +
n
2
and putting s j H() =
n
2
n
2
2
+ j 2
n
4a
= |H()| exp( j) 4b
with |H()| =
n
2
(
n
2
2
)
2
+ 4
2
n
2
2
4c
and tan =
2
n
n
2
2
4d
0 =
n
| H | 1 1
2
n
2
2
For <
1
2
there is a peak in the | H | curve. It is easier to find the peak of |H|
2
which is at
the same value of as the peak of |H|. And easier still to find the minimum of
1
|H|
2
whic
occurs at teh same value of as the peak in |H|
n
4
d
d
_
1
|H|
2
=
d
d
( )
(
n
2
2
)
2
+ 4
2
n
2
2
= 2(
n
2
2
)(-2) + 8
2
n
2
= 4
( )
2
n
2
( 12
2
)
so
d
d
_
1
|H|
2
= 0 if = 0; or if =
n
1 2
2
In the latter case |H| =
1
( 1 1+ 2
2
)
2
+ 4
2
( 12
2
)
=
1
4
2
( 1
2
)
=
1
2
1
2
5
Note that if = 0 then the peak of | H | . This is the undamped case. Note also that the
peak is at a frequency close to but slightly less than
n
. There is no peak if >
1
2
since then
the frequency is imaginary.
This is illustrated in the next figures with
n
= 1. Note that the frequency scale has been
extended in the phase plot to show the approach to the limits.
In the magnitude plot note (i) the roll of rate is 40 dB per decade.
Second order transfer functions 8
w t norris 2nd order transfer functions Monday 3 March 2008
(ii) the peak is at about 14 dB for = 01. Eq. 5 gives a peak of
a factor of 5: so 20 log
10
(5) = 13.97. The theory pans out correctly.
-0.1 1 10
-40
-30
-20
-10
0
10
H, dB
Frequency, rad s
-1
2
0.1
1
2
0.01 0.1 1 10
-180
-135
-90
-45
0
50
Frequency, rad s
-1
( log scale )
Phase,
degrees
2
1
2
0.1
Impulse Responses
Consider the Dirac impulse, (t), as an input.
= 0. Y =
n
2
s
2
+
n
2
1 =
n
s
2
+
n
2
so y(t) =
n
sin(
n
t)
There is no damping and the system rings
forever.
t
f(t)
(t)
n
-
n
y(t)
t
2/
n
Second order transfer functions 9
w t norris 2nd order transfer functions Monday 3 March 2008
> 1 The roots are real. We had eq. 3:
s
1
,s
2
=
n
( )
2
1 ; s
1
> s
2
3
Thus the response is Y =
n
2
(ss
1
)(s-s
2
)
1
=
n
2
s
1
s
2
_
1
s-s
1
1
ss
2
and y(t) =
n
2
2
1
( ) exp(s
1
t) exp(s
2
t)
The response below is for
n
= 1 and = 2. The coefficients of the component exponentials
is omitted in the figure. The formula is : y = 0289( e
027 t
e
37 t
)
2 3 4 5
-0.3
-0.2
-0.1
0.1
0.2
0.3
t
y(t)
exp( s
2
t)
exp( s
1
t)
= 1 Y =
n
2
(s +
n
)
2
1 =
n
2
(s +
n
)
2
so y(t) =
n
2
t exp( -
n
t)
The plot is very like the previous one.
< 1 This is a bit more complicated. We need to complete squares in the denominator.
Y = H(s) 1 =
n
2
s
2
+ 2
n
s +
n
2
=
n
2
(s +
n
)
2
2
n
2
+
n
2
=
n
2
(s +
n
)
2
+
d
2
6
with
d
=
n
1
2
7
This frequency appears often and should be noted.
Second order transfer functions 10
w t norris 2nd order transfer functions Monday 3 March 2008
Thus Y =
n
2
d
(s +
n
)
2
+
d
2
so, from the tables y(t) =
n
2
d
exp(-
n
t) sin(
d
t) 8a
=
1
2
exp(-
n
t) sin(
d
t) 8b
The response for
n
= 1 and = 01 is below. It is an exponentially decaying sine wave.
The formula is : y = 1.005 e
0.1 t
sin( 0995 t)
5
10
15 20
25
-0.6
-0.4
-0.2
0.2
0.4
0.6
0.8
t
y(t)
The period of the oscillation, T =
2
d
. The ratio of succeeding positive peaks is then:
exp(-
n
t)
exp
'
_
t +
2
d
= exp
_ 2
n
d
The logarithm of this ratio is =
2
n
d
=
2
n
n
1
2
so =
2
1
2
2 if << 1 9
is called the logarithmic decrement or log-dec for short. It is a useful way of
measuring the damping in a highly resonant system.
Note we use the symbol with another meaning later.
Second order transfer functions 11
w t norris 2nd order transfer functions Monday 3 March 2008
Step function responses
The input is f = u(t)
so
F =
1
s
f= u(t)
t
1
The response is, using H(s) from eq. 1
Y =
n
2
s
2
+ 2
n
s +
n
2
1
s
10
Cases
= 0 Y =
n
2
s( s
2
+
n
2
)
A
s
+
Bs + C
s
2
+
n
2
Thus 1 A(s
2
+
n
2
) + s (Bs+C)
Put s = 0:
n
2
= A
n
2
: A = 1
Coeff. s
2
: 0 = A + B : B = A = 1
Coeff, s : 0 = C : C = 0
Y =
_
1
s
s
s
2
+
n
2
and y(t) =
( )
1 cos(
n t
)
The plot for
n
= 1 is on the right.
The system oscillates about the mean
indefinitely.
2 4 6 8 10
1
2
t
y(t)
> 1 The transfer function is factorisable and the roots are real. As before eq. 3 gives the
roots:
s
1
,s
2
=
n
( )
2
1 ; s
1
> s
2
3
and the response is Y =
n
2
(ss
1
)(s-s
2
)
1
s
We have Y =
n
2`
s(ss
1
)(s-s
2
)
A
s
+
B
s-s
1
+
B
s-s
2
Second order transfer functions 12
w t norris 2nd order transfer functions Monday 3 March 2008
whence 1 A(s s
1
)(s s
2
) + Bs(ss
2
) + Cs(s s
1
)
Put s = s
1
:
n
2
= B s
1
(s
1
s
2
) B =
n
2
s
1
(s
1
s
2
)
=
s
2
s
1
s
2
Put s = s
2
:
n
2
= C s
2
( s
2
s
1
) C =
n
2
s
2
(s
1
s
2
)
= -
s
1
s
1
s
2
Put s = 0 :
n
2
= A s
1
s
2
A =
n
2
s
1
s
2
=1
So Y =
1
s
+
1
2
1
_ s
2
ss
1
s
1
ss
2
) ; since s
1
s
2
=
n
2
; s
1
s
2
=
n
2
1
and y(t) = 1 +
1
2
1
[ s
2
exp(s
1
t) s
1
exp(s
2
t) ])
noting s
1
,s
2
=
n
( )
2
1 ; s
1
> s
2
With
n
=1 and = 2 this becomes
y(t) = 1 + 0077 e
373 t
1077 e
027 t
which is illustrated on the right.
10 15
0.2
0.4
0.6
0.8
1
t
y(t)
= 1 Y =
n
2
(s +
n
)
2
1
s
=
n
2
s(s +
n
)
2
A
s
+
B
s+
n
+
C
(s+
n
)
2
, say
Thus
n
2
(s +
n
)
2
+ Bs(s +
n
) + Cs
Put s = 0 :
n
2
= A
n
2
A =1. Put s =
n
:
n
2
= C
n C =
n
Coeff. s
2
: 0 = A + B B = A = 1
Y =
1
s
1
s+
n
n
(s +
n
)
2
Thus from the Tables:
y(t) = 1 exp(
n
t ) [1 +
n
t ]
For
n
= 1 y(t) = 1 e
t
( 1+t)
This is illustrated on the right. 2 4 6 8 10
1
t
y(t)
< 1 The more complicated but more interesting case. From eq. 1 we have
Y =
n
2
s
2
+ 2
n
s +
n
2
1
s
Second order transfer functions 13
w t norris 2nd order transfer functions Monday 3 March 2008
=
n
2
s(s
2
+ 2
n
s +
n
2
)
=
A
s
+
Bs + C
s
2
+ 2
n
s +
n
2
, say.
Thus 1 A (s
2
+ 2
n
s +
n
2
) + s (B s +C)
Put s = 0 :
n
2
= A
n
2
A = 1
Coeff. s
2
:
0 = A + B B = A = 1
Coeff. s : 0 = A 2
n
+ C C = 2
n
A = 2
n
and Y =
1
s
s + 2
n
s
2
+ 2
n
s +
n
2
=
1
s
s + 2
n
(s +
n
)
2
+
d
2
: completing the square as in eq. 6
=
1
s
(s +
n
) +
d
d
(s +
n
)
2
+
d
2
: making the numerator to suit the Tables
with
d
=
n
1
2
Thus y(t) = 1 exp(
n
t)
_
cos(
d
t) +
d
sin(
d
t) 11
For
n
= 1 and = 02
we have
y(t) 1 e
02 t
[ cos( 098 t ) + 02
sin( 098 t) ]
This is plotted on the right.
Note that the gradient is zero at t = 0
5 10 15 20
1
2
t
y(t)
1+ e
-t
1- e
-t
Second order transfer functions 14
w t norris 2nd order transfer functions Monday 3 March 2008
There are properties of interest of the curve.
10 20
1
t
y(t)
M
P
T
p
T
s
Overshoot,
O
P
T =
2
Period,
Overshoot
The derivative
dy
dt
=
n
2
d
exp(
n
t) sin(
d
t)
is remarkably simple. It is equal to the impulse response. Indeed it is generally the case that
the derivative of the step response of any linear system is the impulse response.
dy
dt
= 0 when
d
t = n or when t =
n
d
=
n
n
1-
2
=
nT
2
where the period of the oscillations is T =
2
d
=
2
n
1-
2
The value of y at these extrema is Y = 1 + exp
_
n
1
2
The first of these when n = 1 gives the peak of the overshoot ( often simply called the
overshoot.
P
= 1 + exp
1
2
Express proportional overshoot as the
excess as a proportion of the final steady
state value.
P
= y 1 = exp
1
2
12a
occurring at
T
P
=
n
1-
2
=
d
=
T
2
12b
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
Overshoot, O
p
Damping,
Second order transfer functions 15
w t norris 2nd order transfer functions Monday 3 March 2008
We may be interested in the time to reach and stay within a certain 'distance' from the final
value. The deviation of the nth extremum is
= exp
_
n
1
2
which occurs at t
n
=
n
n
1
2
or
n
1
2
= t
n
n
so = exp
_
n
1
2
= exp(
n
t
n
)
and T
s
= t
n
=
ln()
n
Strictly t
n
should be an integral number of half periods but the discrepancy is usually
ignored.
For = 002, i.e., 2% T
s
4
n
and for = 005, i.e., 5% T
s
3
n
We may also consider an overall expression of deviation from the final value. The root
mean square seems a suitable choice.
5
10 15 20
-1
1
t
t
y(t) - 1
5 10 15 20
0.2
0.4
0.6
0.8
1
(y-1)
2
The deviation and its square are illustrated above for the previously shown response. We
take then as a measure of the overall deviation as
E
rts
= the total square deviation ( integrated from t = 0 to t = )
We require
E
rts
=
]
1
1
1
1
1
( y 1)
2
dt
1/2
=
1+4
2
4
n
This rather simple expression is
valid for all , i.e., 0 < < . It has
a minimum value of 1 when =
1
2
.
1 2 3 4 5
1.2
1.0
1.4
1.6
1.8
2
2.2
E
rts
Second order transfer functions 16
w t norris 2nd order transfer functions Monday 3 March 2008
This minimum square deviation
condition is illustrated on the right
with a 16% overshoot..
10 5 15
1
t
y(t)
1.16
Second order transfer functions 17
w t norris 2nd order transfer functions Monday 3 March 2008
There are other measures of the error. To put emphasis on the approach to the final value on
might weight erros by muptiplying by time. The integral of this de-emphasises the initial
approach to the final value.. Thus we have
The root of the integral of time weighted
total square erroe is
E
rts
=
]
1
1
1
1
1
t ( y 1)
2
dt
1/2
=
1
2
n
1+ 8
4
2
This has a minimum of
1
2
1/4
n
when
=
_ 1
8
1/4
0595
5
10 15 20
0.2
0.4
0.6
0.8
t
t (y - 1)
2
The overshoot is
exp
1
2
= exp
221
98%
Second order transfer functions 18
w t norris 2nd order transfer functions Monday 3 March 2008
Keeping factors constant
Poles are at s =
n
( )
1
2
:
Constant imaginary part.
d
=
n
1
2
= constant.
So Constant frequency in
response
1
t
y(t)
j
s
Constant real part:
n
is constant.
Fixed envelope
y =1 exp(
n
t)
and so time to get within a
given limit is deermined by
n
.
E.g., T
=2%
=
4
n
1
2
t
y(t)
j
s
Constant damping ratio,
Thus al so const an
overshoot:
O
P
= exp
1
2
Note: = sin where is
marked in the figure.
1
t
y(t)
j
There is scarcely any
overshoot when > 60.
20 40 60 90
1.2
1.4
1.6
1.8
2
Fractional overshoot,
O
P
j
s
Second order transfer functions 19
w t norris 2nd order transfer functions Monday 3 March 2008
