University of Cagliari

Dept of Electrical and Electronic Eng.

Describing Function analysis

of
nonlinear systems
Prof. Elio USAI
eusai@diee.unica.it

University of Leicester
Department of Engineering
 System definition and problem statement

d(t) Plant

r(t) + e(t) u(t) m(t) + y(t)

_ C A P
+
controller actuator process

z(t) ym(t) +
F T
+ trasmitter
filter
n(t)
Controller

r: reference variable ym: measure of the output

e: error signal z: feedback signal
u: control variable d: disturbance
m: manipulated input n: measurement noise
y: controlled variable

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

In many cases the system presents a nonlinear phenomenon which is fully

characterised by its static characteristics, i.e., its dynamics can be neglected

m
 Saturated actuators +M

 Relay control

 Gears backlash u

 Hysteresis in magnetic
materials
-M
 Dead zone in electro-
mechanical systems

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

In many cases the system presents a nonlinear phenomenon which is fully

characterised by its static characteristics, i.e., its dynamics can be neglected

e
+U
 Saturated actuators

 Relay control
-E
 Gears backlash e
+E
 Hysteresis in magnetic
materials

 Dead zone in electro- -U

mechanical systems

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

In many cases the system presents a nonlinear phenomenon which is fully

characterised by its static characteristics, i.e., its dynamics can be neglected

m (driven gear)

 Saturated actuators

 Relay control
-U u (driving gear)
 Gears backlash
+U
 Hysteresis in magnetic
materials

 Dead zone in electro-

mechanical systems

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

In many cases the system presents a nonlinear phenomenon which is fully

characterised by its static characteristics, i.e., its dynamics can be neglected

m
 Saturated actuators

 Relay control +M0

 Gears backlash u

 Hysteresis in magnetic -M0

materials

 Dead zone in electro-

mechanical systems

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

In many cases the system presents a nonlinear phenomenon which is fully

characterised by its static characteristics, i.e., its dynamics can be neglected

 Saturated actuators

 Relay control
-U u
 Gears backlash
+U
 Hysteresis in magnetic
materials

 Dead zone in electro-

mechanical systems

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

Nonlinearities are not always a drawback, they can also a have a

“stabilising” effect

r(t) + e(t) u(t) m(t) y(t)

_ C A P

z(t)
H

1
Ps  

s 1  1.4 s  s 2 
0.5  1
H s   As  
1  0.1s   sat  2 

C s   10

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement
S te p re s p o n s e
20
N O T s a tu ra te d a c tu a to r
S a tu ra te d a c tu a to r
15

10

The saturated
5
system
S y s te m o u tp u t

0
oscillates but
does not
-5 diverge

-1 0

-1 5

-2 0
0 2 4 6 8 10 12 14 16 18 20
tim e [s ]

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

B o d e D ia g r a m o f th e d iff e r e n c e tr a n s f e r f u n c t io n
50
S y s te m : s y s
F r e q u e n c y ( r a d /s e c ) : 1 .6 8
M a g n it u d e ( d B ) : - 0 . 0 7 0 8
0 S y s te m : s y s
M a g n itu d e ( d B )

F r e q u e n c y ( r a d /s e c ) : 0 .9 4
M a g n it u d e ( d B ) : 1 2 The NOT
-5 0
saturated
system is not
-1 0 0
stable since its
-1 5 0
stability margin
-9 0
are negatives
S y s te m : s y s
F r e q u e n c y ( r a d /s e c ) : 0 .9 4 1

-1 8 0
Ph a s e (d e g ): -1 8 0 • mg = -12 dB
P h a s e (d e g )

S y s te m : s y s • m = - 47 deg
-2 7 0 F r e q u e n c y ( r a d /s e c ) : 1 .6 8
Ph a s e (d e g ): -2 2 7

-3 6 0
-1 0 1 2
10 10 10 10

F re q u e n c y (r a d /s e c )

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

Many systems can be reduced to a simplified form in which all the linear
dynamics is concentrated in a unique block and the static non linear
characteristics is represented by a separate block

r(t) + e(t) u(t) m(t) y(t)

C(j) f(u) P(j)
_

z(t) H(j)

r(t) r’(t) + u(t) m(t) w(t)

C(j) f(u) P(j)H (j)C (j)
_

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement

If constant (or very slowly varying) reference signals are considered,

under some conditions it is possible to separate the low-frequency, almost
static, behaviour defined by a working nominal condition and a high-
frequency behaviour due to small variations around such a working point
r(t)=R r’(t) + u(t) m(t) w(t)
C(j) f(u) P(j)H (j)C (j)
_

u t   u0  u t 
r’(t)=kCR + u(t)
f(u) m(t) w(t) mt   m0  mt 
G (j)
wt   w0  wt 
_

u0  kC R  kG m0

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement
The non linear characteristics is translated so that the origin of the new
reference Cartesian system is the point (u0, m0) in the original one
m
m

kC
m0  f u0 
kG
R

m0
u  u0  kC R  kG m0
kC R
u
u0
m0  f u 
1

kG f u0   f u0  u   m0

r’(t)=0 + u(t) w(t)

f’(u) m(t)
G (j)
_

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 System definition and problem statement
The aim of the system analysis is to define which is the steady-state
behaviour of the system defined by the variations around the nominal
working point

r’(t)=0 + u(t) w(t)

f’(u) m(t)
G (j)
_

If f’(u) is a passive sector function absolute stability tools allow for sufficient
conditions for global asymptotic stability of the variation system, i.e, the steady
state is characterised by constant values of the system variables

If the origin of the variation system is not stable does the

varia-bles diverge to infinity or some periodic motion can
appear?

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Limit cycles
Limit cycle: a periodic oscillation around a constant working point

10
G s  
r’=5 + u(t)
f(u) m(t)
G (j)
w(t) 
s  1 s 2  1.4s  1 
_

f u   sat 1, 116 , 16

11

2

m
1 1
1 .5 m u
10 2
1

5
u u0 
0 .5
11
m

m
5
0

m0 
0
u0 11
-0 .5

-1
-6 -4 -2 0 2 4 6 8

u
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Limit cycles
Limit cycle: a periodic oscillation around a constant working point
r’(t)=0 + u(t) w(t)
f’(u) m(t)
G (j)
_

7
5
u0 
6 11
5
5 m0 
11
o u tp u t w (t)

w
0
4

50
1 w0   4.545
11
0
0 5 10 15 20 25 30

t im e [ s ]
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Limit cycles

Limit cycle can be a drawback in control systems:

 Instability of the equilibrium point

 Wear and failure in mechanical systems

 Loss of accuracy in regulation

Parameters of the limit cicle can be used to discriminate between

acceptable and dangerous oscillations

 oscillation frequency

 oscillation magnitude

Electronic oscillators can be based on limit cycles

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function - Assumptions

The describing Function approach to the analysis of steady-state oscillations

in non linear systems is an approximate tool to estimate the limit cycle
parameters.

It is based on the following assumptions

 There is only one single nonlinear component

 The nonlinear component is not dynamical and time

invariant

 The linear component has low-pass filter properties

 The nonlinear characteristic is symmetric with respect

to the origin

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function - Assumptions

There is only one single nonlinear component

The system can be represented by a lumped parameters

system with two main blocks:
•The linear part
•The nonlinear part

r’=const. + u(t) m(t) w(t)

N.L. G (j)
_

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function - Assumptions

The nonlinear component is not dynamical and time invariant

The system is autonomous.

All the system dynamics is concentrated in the linear part
so that classical analysis tools such as Nyquist and Bode
plots can be applied.

m
+M

u
u mM
u 

-M

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function - Assumptions

The linear component has low-pass filter properties

This is the main assumption that allows for neglecting

the higher frequency harmonics that can appear when a
nonlinear system is driven by a harmonic signal

G  j   G  jn  n  2, 3, 

The more the low-pass filter assumption is verified the

more the estimation error affecting the limit cycle
parameters is small

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function - Assumptions

The nonlinear characteristic is symmetric with respect to the origin

This guarantees that the static term in the Fourier expansion of the
output of the nonlinearity, subjected to an harmonic signal, can be
neglected

r’(t)=0 + u(t) w(t)

f’(u) m(t)
G (j)
_

Such an assumption is usually taken for the sake of simplicity, and

it can be relaxed

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Fourier expansion - Recall

Consider a periodic function

y (t )  f t , y (t )  y (t  T ), T is a real constant

a0 
y (t )    ak sin k 2
T t   bk cosk 2
T t 
2 k 1

T 
2 2 1 2

ak   f t  cosk 2
t dt   f t cosk 2
t d  2T t 
T T 2 T
  2
T

k  01,2,  b0  0
T 
2 2 1 2
bk   f t sin k 2
t dt   f t sin k 2
t d  2T t 
T T 2 T
  2 T

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Harmonic balance

r’(t)=0 + u(t) w(t)

f’(u) m(t)
G (j)
_

u (t )  U sin  t 


m(t )   ak cosk t   bk sin k t 
k 1


w(t )   Gk ak cosk t   k   bk sin k t   k 
k 1
Gk  G  jk   k  G  jk 

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Harmonic balance

Consider the polar representation of a complex number associated

with the exponential form of harmonic signals

u (t )  Ue j t

 M k  ak2  bk2
m(t )   M k e jk e jk t a
k 1 k  arctan k
bk

Taking into account the low-pass property of the linear part of the
system


w(t )   Gk e j k M k e jk e jk t  G1e j1 M 1e j1 e j t
k 1

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Harmonic balance

r’(t)=0 + u(t) w(t)

N.L. m(t)
G (j)
_

A permanent oscillation in the loop appears if u(t)=-w(t)

j t j1 j1 j t j1 M 1 j1
Ue  G1e M 1e e 1  G1e e 0
U
Harmonic balance equation

1  G  j N U ,    0

1
N U ,    b1  ja1  is the Describing Function of the nonlinear term
U

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Harmonic balance

The armonic balance equation is a necessary condition for the existence of

limit cycles in the nonlinear system

The approximate analysis gives good estimates if the low-pass filter

hypothesis is strongly verified. It is a good tools for engineers

The harmonic balance equation is similar to the characteristic polynomial

function, i.e. it leads to the Nyquist condition for closed-loop stability

The Describing Function is a linear approximation of the static nonlinearity

limited to the first harmonic

In most cases the Describing Function is not a function of the frequency and
this simplifies the verification of the harmonic balance equation by means of
the Nyquist plot of the transfer function

1
G  j   
N U ,  
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Harmonic balance
N y q u is t D ia g r a m
2

= + 
0

-2
Im a g in a ry A x is

( U 0 , 0 )
-4 +  U U=0

-\1 /N (U )

-6

-8

G ( j )
= 0 +
-1 0
-1 .4 -1 .2 -1 -0 .8 -0 .6 -0 .4 -0 .2 0 0 .2

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Harmonic balance

N y q u is t a n d D F D ia g r a m s
1

0 .5 S y ste m : sy s
R e a l : -2 .0 1
Im a g : -0 .0 2 1
F re q (ra d /se c ): 0 .9 9 7
0 = 1 0 0

-0 .5
= 1 7 . 7 8 2 8
= 0 . 1
Im a g in a r y A x is

-1

U
-1 .5 S y ste m : sy s
R e a l : -1 .1 4
Im a g : -2 .0 2
F re q (ra d /se c ): 0 .6 5 1
-2

= 3 . 1 6 2 3
-2 .5

-3 
= 0 . 5 6 2 3 4
-3 .5
-8 -7 -6 -5 -4 -3 -2 -1 0 1

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

The DF computation can be performed by means of its definition

Usin(t) m(t)
f’(u)

T
2 2
a1   mt  cost dt
1 T T 2
N U ,    b1  ja1 
U T
2 2
b1   mt sin t dt
T T 2

The evaluation of coefficients a1 and b1 can be performed by means of both

analytical calculation and numerical integration, depending on the type of
nonlinearity involved

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Ideal relay
3 2
u
m
1 .5
+M
2

+M 1

1
0 .5
 m (t)

0 0
T
-0 . 5
-1

-M -1

-2

-M -1 . 5

-3
-3 -2 -1 0 1 2 3 -2
0 0 .5 1 1 .5 2 2 .5 3
 u (t) T im e

T
2 2
a1   mt  cost dt  0 Because of the odd symmetry of the m(t) signal
T T 2

4M
T 

N U ,   
2 2 2 2 4M
b1   mt sin t dt   M sin  d 
T T 2  0 
U

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Ideal relay N y q u is t a n d D F D ia g r a m s
0 .5

U
The negative reciprocal 0

of the DF is the (  ; 4 M / ) -1 /N (U )
-0 .5 c
negative real axis in
backward direction -1

Im a g in a r y A x is
-1 .5

-2 
A limit cycle can exist
if the relative degree of -2 .5

G(j) is greater than -3

G ( j )
two
-3 .5

-4
-1 .6 -1 .4 -1 .2 -1 -0 .8 -0 .6 -0 .4 -0 .2 0 0 .2

R e a l A x is

The oscillation frequency is the critical frequency c of the linear system and the
oscillation’s magnitude is proportional to the relay gain M
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Pure hysteresis
2
2
u
M m
1 .5
1 .5

1
1

0 .5
0 .5
m

0
0

-0 .5 -0 .5

-1 -1

-M
-1 .5 -1 .5

-2 -2
-2 .5 -2 -1 .5 -1 -0 .5 0 0 .5 1 1 .5 2 2 .5 0 0 .5 1 1 .5 2 2 .5 3
u T im e
T
2
2
b1 = ∫ Δm ( t ) sin ( ωt ) dt =0
T −T Because of the even symmetry of the m(t) signal
2

T 
2 2 2 2 4M
a1   mt  cost dt    M cos d  
 0  4M
N U ,     j
T T 2

U

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation
N y q u is t a n d D F D ia g r a m s
Pure hysteresis 0 .2

The negative reciprocal

of the DF is the -0 .2

negative imaginary -0 .4

axis in backward
direction Im a g in a r y A x is
-0 .6

-0 .8
G ( j )
A limit cycle can exist -1

if the relative degree -1 /N (U ) 

of G(j) is greater
-1 .2

than one and G(j) is -1 .4

U
a type-0 system
-1 .6
-0 .4 -0 .2 0 0 .2 0 .4 0 .6 0 .8 1 1 .2 1 .4

R e a l A x is

The oscillation frequency is lower than the critical frequency c of the linear
system and the oscillation’s magnitude is proportional to the relay gain M and to
the modulus of the transfer function at phase -/4
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Hysteretic relay
2 .5
2
u
m
2
+M
1 .5

1 .5

1
1

0 .5 0 .5

-  0
m

0
t

-0 .5
-0 .5

-1
-1
-1 .5

-1 .5
-M -2

-2 -2 .5
-2 .5 -2 -1 .5 -1 -0 .5 0 0 .5 1 1 .5 2 2 .5 0 0 .5 1 1 .5 2 2 .5 3
u T im e

If U the hysteretic relay behaves as pure hysteresis

t T 2
8 8 4 4M  
b1   M  sin t dt  M  sin t dt  1  
T 0 T t  U 
 4M
8
t
8
T
4
 j U U 
a1   M  cost dt  M  cost dt  
T 0 T t N U ,     2
 4 M    4 M
1    j U 
The imaginary part of N(U) is proportional to  U U  U 2
the hysteresis area
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Hysteretic relay N y q u is t a n d D F D ia g r a m s
0 .5

The negative reciprocal 0

of the DF is the parallel -0 .5

to the negative real axis U -1 /N (U )

and with constant -1

Im a g in a r y A x is
negative imaginary part -1 .5

-2

If  is larger than U
-2 .5

the relay could not -3

G ( j )
behave as a pure

hysteresis -3 .5

-4
-1 .4 -1 .2 -1 -0 .8 -0 .6 -0 .4 -0 .2 0 0 .2

R e a l A x is

The oscillation frequency is lower than the critical frequency c of the linear
system and the oscillation’s magnitude is proportional to the relay gain M
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Saturation 1
u
m
0 .8
4

+M 0 .6
3
+M
0 .4
2

k 0 .2
1

0
m

0
t

-1 -0 .2

-2 -0 .4
-M
-3 -0 .6
-M

-4 -0 .8
-2 .5 -2 -1 .5 -1 -0 .5 0 0 .5 1 1 .5 2 2 .5
u
-1
0 0 .5 1 1 .5 2 2 .5 3
T im e

If U the saturation behaves as pure gain

T
2 2
a1   mt  cost dt  0 Because of the odd symmetry of the m(t) signal
T T 2
2kU  M  
t T 2
8 8 4  M  M  
b1   kU sin t dt  M  sin t dt 
2
arcsin   1   
T 0 T t    kU   kU   kU  
 
 M
k U
 k
N U    
 M  
2
2k
 M  M  M
 arcsin   1   U
    kU   kU   kU   k


Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Saturation N y q u is t a n d D F D ia g r a m s
0 .5

-1 /N (U )
The negative reciprocal 0

U -1 /k
of the DF is part of the -0 .5

negative real axis in

backward direction
-1

Im a g in a r y A x is
-1 .5

-2

A limit cycle can exist -2 .5

if the relative degree

of G(j) is greater
-3


than two and gain k is -3 .5 G ( j )
sufficiently high
-4
-1 .6 -1 .4 -1 .2 -1 -0 .8 -0 .6 -0 .4 -0 .2 0 0 .2

R e a l A x is

The oscillation frequency is the critical frequency c of the linear system and the
oscillation’s magnitude depends on the saturation parameters M and k

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation
2 .5
u
Dead zone 2 m
3

1 .5

2
1

k
1 0 .5

- 0
m

0
t
+ 
-0 .5

-1
-1

-2
-1 .5

-2
-3
-3 -2 -1 0 1 2 3
u
-2 .5

If U the dead zone has no output 0 0 .5 1 1 .5

T im e
2 2 .5 3

T
2 2
a1   mt  cost dt  0 Because of the odd symmetry of the m(t) signal
T T 2

8 4
T
2kU         
2

b1   k U sin t    sin t dt    arcsin     1    

T t  2 U  U   U  


  U sin t 
0 U 
 
N U       
2
   
arcsin 
2k
k     1   U 
   U   U   
U
 
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation
N y q u is t a n d D F D ia g r a m s
Dead zone 1

The negative reciprocal 0 .5

of the DF is part of the -1 /N (U ) - 1 /k

negative real axis in 0

forward direction U=+

Im a g in a r y A x is
-0 .5

-1

A limit cycle can exist G ( j )

if the relative degree -1 .5

of G(j) is greater 

than two and gain k is -2

sufficiently high
-2 .5
-3 .5 -3 -2 .5 -2 -1 .5 -1 -0 .5 0

R e a l A x is

The oscillation frequency is the critical frequency c of the linear system and the
oscillation’s magnitude depends on the dead zone parameters  and k

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation
3

Dead zone
2 k

The nonlinear
s a t u r a t i o n  ( u )
characteristics of 1

the Dead Zone

can be computed d e a d z o n e ( u )
m 0
by subtracting the
Saturation
characteristics -1

from a linear one

 u   k   u 
-2

-3
-3 -2 -1 0 1 2 3
u

N  U   k  N  U 

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

The Describing function of a nonlinear characteristics can be computed as the

combination of the Describing Functions of the elementary constituting
nonlinear characteristics

Nt(U,)
N1(U,)
r’(t)=0 + u(t) m(t) w(t)
G (j)
_
N2(U,)

N t U   N1 U   N 2 U 

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

A number of Describing Function can be computed by particularisation of the

function
   
2


arcsin     1   2 
in which the parameter defines a peculiar point of the nonlinear
characteristics m

+M

k
k  1
-
u/U N    
+ k      1

-M

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

+M k2

k1  1
-
k1
N    
k 2  k1  k 2       1
u/U
+

m
-M

k
0  1
- u/U N    
+ k 1      1

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation
m
+M k2
0  1

- - k1
u/U
N    k  1      1 
+ + 
k           1
-M

+M k
4M
u/U N    k 
-M 

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation
m
+M

4M
u/U N   


-M

m
+M

 0  1

u/U N     4 M

  1   2
 1

-M

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

k 0  1

- N     k 4k
 2 1   2  1  j  1      1
u/U
+

k 0  1

- N     k 4k
+
u/U
 2 1   2  1  j 1      1


Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

m
+M

0  1

-
u/U N U ,     4 M 4 M
 U 1   2
 j  1
+ U

-M

m
+M

-
u/U 0  1

+ N U ,     4 M 4M
 U 1   2
 j  1
U

-M

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Describing Function – Computation

Nt(U,)
r’=0 + u m w
G (j)
_
1/j

Negative inverse of the Describing Function

0.5

U
1
N t U   N 1 U   N 1 U 
0
=100

j -0.5
=17.7828 U

 =0.1
Imaginary axis -1

4M 4M
N U ,    j -1.5 U

U U -2
U
=3.1623
-2.5

1 U
   j 
N U ,    
-3

4M 1   2
=0.56234
-3.5
-8 -7 -6 -5 -4 -3 -2 -1 0 1
Real axis

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Stability of limit cicles

r’(t)=0 + u(t) w(t)

N.L. m(t)
G (j)
_

If the block N.L. is a pure constant k, the stability of the feedback system
can be performed by means of the Nyquist criterion, which gives the
number of roots with positive real roots of the Harmonic balance equation
1+G ( jω ) k =0

The Nyquist criterion looks at the relative position of the transfer function
G(j) with respect to the point (-1/k, 0) in the complex plain.
By extension the criterion can be applied to any point – of the complex
plain with reference to the Harmonic balance equation

α+G ( jω) =0 , α∈ C

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Stability of limit cycles

If the transfer function G(j) represents a stable system, the reduced

Nyquist criterion can be applied, i.e. the closed loop stability can be stated
if the reference point – of the complex plain remains on the left-side
when running along the Nyquist plot of G(j) from =0+ to  =+.

N y q u i s t p l o t o f G ( j )

1
1

G  j  
0 .5

0
= +   2

j  j   0.5 j  1
-1 /k
-0 .5

The closed loop is not stable

Im a g in a r y A x is

-1

-1 .5 with respect to the point (-1/k,0).

-2

-
The closed loop is stable with
-2 .5

respect to the point -

-3

-3 .5
+
= 0
-4
-3 -2 .5 -2 -1 .5 -1 -0 .5 0

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Stability of limit cycles

Can the generalization of the Nyquist criterion be used to analyse the

stability of a limit cycle?

YES

What does it mean that a limit cycle is stable?

If a perturbation of the magnitude of the periodic oscillation occurs, it
tends to the original value as time passes.

How can be considered a limit cycle from the Nyquist criterion point of
view?
It is a marginally stable condition.
How can the magnitude of a limit cycle represented?
By a point in the Describing Function plot.

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Stability of limit cycles

N y q u is t a n d D F p l o t s
2

1 .5
Points B, B’,
B”, A and A’ A’
A
1

represent 0 .5

possible
Im a g in a r y A x is

magnitudes of
0

the -0 .5 U
oscillation. -1 B B”
B’
-1 .5

A and B re- -2

G ( j )
present two -2 .5 
possible limit
cycles -3
-3 -2 .5 -2 -1 .5 -1 -0 .5 0

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 Stability of limit cycles
Nyquist and DF plots
Applying the reduced 2

Nyquist criterion with 1.5

respect to: A’
A
1

0.5

point B’: oscillations tend 0

to decrease (stable system)

Imaginary Axis
-0.5
U
B B”
point B”: oscillations tend -1

B’
to increase (unstable -1.5

system) -2

G(j  )
-2.5 

point A’: oscillations tend -3

-3 -2.5 -2 -1.5 -1 -0.5 0

to decrease (stable system) Real Axis

B: Stable limit cycle A: Unstable limit cycle

The linear system G(j) is assumed to be stable in order to apply the reduced Nyquist criterion
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

Consider a DC motor with permanent magnets

+24 V

M
0V d.c.

-24
V

The position of the motor shaft is measured by means of a rotational variable

resistance.

The voltage on the resistance drives the position of a relay that switches the
motor supply voltage between +/- 24 V d.c.
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

The linear approximate model of a DC motor is the following

dir t 
vr t   Rr ir t   Lr  et 
dt Rr: rotor resistance
d t 
J  Cem t   B t  Lr: rotor inductance
dt
d t  ke: voltage feedback constant
 kt: torque constant
dt
et   ke t  Jm: motor inertia
Jl: load inertia
Cem t   kt ir t  Bm: motor friction coefficient
Bl: load friction coefficient
J  J m  Jl

B  Bm  Bl vr: rotor supply voltage

ir: rotor wound current
Cem: electromagnetic torque
 24   d : rotational speed
vr t    : shatf angular position
 24   d

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

The motor transfer functions are the following

kt
W s  
sLa  Ra sJ  B   k t k e

kt
W ϑ ( jω )= =ℜ( jω ) +j ℑ( jω )
jω (( jωL a +R a )( jωJ+B ) +k t k e )
k t ( Ra J+La B )
=- 2
2
ω 2 ( Ra J+L a B ) +( R a B+k t k e−ω 2 L a J )
k t ( Ra B+k t k e−ω2 La J )
−j
[ 2 2
ω ω ( R a J+La B ) +( Ra B+k t k e−ω La J ) 2 2
]

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

Taking into account the parameters of the motor and of the load
N y q u is t a n d D F p lo ts
R=0.4; % rotor resistance 0

L=0.001; % rotor inductance -2

U

ke=0.3; % voltage feedback constant -4

kt=0.3; % torque constant -6

I m a g in a r y A x is
Jm=0.01; % motor inertia
-8

-1 0
Jl=0.09 % load inertia
-1 2

Bm=0.05; % motor friction coefficient

-1 4

Bl=0.05; %load friction coefficient -1 6

J=Jm+Jl; -1 8

B=Bm+Bl; -2 0
-2 -1 .5 -1 -0 .5 0 0 .5 1

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

Taking into account the parameters of the motor and of the load
x 10
-4 N y q u is t a n d D F p lo ts
R=0.4; % rotor resistance 5

4
L=0.001; % rotor inductance
3

ke=0.3; % voltage feedback constant

2

kt=0.3; % torque constant S y ste m : W p

Im a g in a r y A x is
1 R e a l : -0 .0 0 5 6 8

Jm=0.01; % motor inertia Im a g : -8 .7 3 e -0 0 7

F re q (ra d /se c ): 3 6 .8
0

Jl=0.09 % load inertia -1

U

Bm=0.05; % motor friction coefficient -2

Bl=0.05; %load friction coefficient -3


J=Jm+Jl; -4

B=Bm+Bl; -5

-2 0 -1 5 -1 0 -5 0
-3
R e a l A x is x 10

Ra B  k t k e
 W p  j cr   0.1759 rad
4M
 W   cr   36.056 rad/s U-
p  j  0
La J 
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

Taking into account the parameters of the motor and of the load
x 10
-4 N y q u is t a n d D F p lo ts
R=0.4; % rotor resistance 5

4
L=0.001; % rotor inductance
ke=0.3; % voltage feedback constant
3
STABLE limit cycle!!!!
2

kt=0.3; % torque constant S y ste m : W p

Im a g in a r y A x is
1 R e a l : -0 .0 0 5 6 8

Jm=0.01; % motor inertia Im a g : -8 .7 3 e -0 0 7

F re q (ra d /se c ): 3 6 .8
0

Jl=0.09 % load inertia -1

U

Bm=0.05; % motor friction coefficient -2

Bl=0.05; %load friction coefficient -3


J=Jm+Jl; -4

B=Bm+Bl; -5

-2 0 -1 5 -1 0 -5 0
-3
R e a l A x is x 10

Ra B  k t k e
 W p  j cr   0.1759 rad
4M
 W   cr   36.056 rad/s U-
p  j  0
La J 
Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

80

a n g le
The system 60
speed
presents a c u rre n t
periodic 40

steady-state
oscillation 20

-2 0

-4 0

-6 0

-8 0
0 1 2 3 4 5 6 7 8 9 10
T im e

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

The system 2 .5
presents a
periodic
steady-state 2

oscillation
A n g le

1 .5

0 .5

0
0 1 2 3 4 5 6 7 8 9 10
T im e

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

1.75 Tc=2/c
The system
1.7
presents a
periodic 1.65

steady-state
Angle [rad]

oscillation 1.6

1.55
2U=8M|G(jc)|/
1.5

1.45

1.4

8.4 8.45 8.5 8.55 8.6 8.65 8.7 8.75 8.8

Time [s]

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

What does it happen if the same control law is applied to the speed
control problem?

ang
24 V

S te p S ig n G a in w

S cop e
C r
i

D C M o to r

kt
W s  
sLa  Ra sJ  B   k t k e
The linear plant is characterised by an all-pole transfer function with relative
degree two, therefore there is no crossing of the Nyquist plot with the real
negative axis, but the origin at =+ (corresponding to U=0 in the DF plot)

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

N y q u is t a n d D F p lo ts
U=0
= + 
0
U = 0 The Nyquist plot of the
-0 .2 linear system is tangent
to the negative
-0 .4
reciprocal of the DF at
the origin, i.e., U=0 and
Im a g in a r y A x is

-0 .6
=+

A sliding mode
-0 .8

-1
behaviour is established
asymptotically
-1 .2


-1 -0 .5 0 0 .5 1 1 .5 2 2 .5 3

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

7
angle
speed

Possibly a sort of 5

dead-bit control,
or finite time
4
stabilisation
seams to appear
3

Can it be a 2
sliding mode
behaviour?
1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application
7

By reducing the 6

integration step
of the simulation 5
it is apparent that
the stabilisation
is asymptotic
Speed [rad/s]

Can it be a 2
sliding mode
behaviour?
1

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

Time [s]

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

60

50
By reducing the
integration step 40

of the simulation
it is apparent that 30

the stabilisation
is asymptotic 20
Rotor current [A]

10

The rotor current 0

tends to a
constant value, -10

i.e. an
-20
asymptotic (2nd
order) sliding -30

sliding mode
appears -40
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

Time [s]

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

The DC motor is a second order system whose state variables are the rotor speed and
current
d t  B k
  t   t ir t 
dt J J
dir t  k R 1
  e  t   r ir t   v r t 
dt Lr Lr Lr
B k
y t    t   t ir t   2 t 
J J

The system transfer function has a zero in –2, and if the system output is steared to
zero in a finite time, than the system behaves as a first order system with time
constant =0.5

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application
3 .5

3
It is apparent that the shaft speed tends
to the steady state value  as a first
2 .5
order system with =0.5
S p e e d [r a d /s ]

1 .5

0 .5

2 .5

0
0 0 .5 1 1 .5 2 2 .5 3 3 .5
T im e [s ]
2

Apart from a very short transient, the

R o to r c u rre n t [A ]
1 .5

state trajectory tends to the steady-state

values sliding on a linear manifold of 1

the state space

0 .5

0
0 0 .5 1 1 .5 2 2 .5 3 3 .5
S p e e d [ra d /s ]

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008
 DF Analysis – Example of application

N y q u is t a n d D F p lo ts
2

1 The Nyquist plot of the

U
linear system cross the
0
negative reciprocal of
the DF at the origin, i.e.,
Im a g in a r y A x is

U=0 and =+

-1

A sliding mode
-2
behaviour is established
in a finite time
-3

-4
-2 -1 0 1 2 3 4 5 6 7 8

R e a l A x is

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008