2.

Limit cycle analysis

Introduction
Nonlinear systems can display oscillations of fixed amplitude and fixed
period without external excitation. These oscillations are called limit cycles,
or self-excited oscillations. Limit cycles represent an important
phenomenon in nonlinear systems. They can be found in many areas of
engineering and nature.

• Describing function analysis

During WWII, servo system developed rapidly, but with limit cycle
problem caused by nonlinear elements in system. Classical control theory
could not deal with this problem. Daniel proposed method of description
function analysis in 1940, realized analysis and suppression of limit cycles.
• Phase plane analysis
In the stage of modern control theory, research of linear system had
made rich achievements, but slow progress for nonlinear system. Phase
plane analysis method was used to determine existence of limit cycles in
nonlinear systems.
 2. Limit cycle analysis
2.1 Describing function analysis – Concepts
Frequency response method is a powerful tool for analysis and design of
linear control systems. Frequency domain analysis cannot be directly
applied to nonlinear systems because frequency response functions
cannot be defined for nonlinear systems.
For some nonlinear systems, an extended version of frequency
response method (P.J. Daniel, 1940, harmonic linearizing method), can be
used to approximately analyze and predict nonlinear behavior (limit cycles).
First assume existence of a limit cycle, then determine whether system
equation can indeed sustain such a solution.
E.g.: Van der pol equation. x   ( x 2  1) x  x  0
Assume: x(t )  Asin(t )
Output of nonlinear block:

w   x 2 x   A2 sin 2 (t ) A cos(t )

A3
 (cos(t )  cos(3t ))
4
 2. Limit cycle analysis
2.1 Describing function analysis – Concepts
3 2
Linear element with low-pass property: w   A  cos(t )  A d [ A sin(t )]
In frequency domain: 4 4 dt
A2
w  N ( A, )(  x) , N ( A, )  ( j) ( depend on A)
4
Nonlinear block can be approximated by equivalent quasi-linear block.
x  Asin(t )  G( j)w  G( j) N ( A, )(  x)
A2 p  A2 ( j ) 
1  1   0  A 2 , 1
4 p2   p  1 4 ( j )2   ( j )  1

1 1 2 2
1,2    ( A2  4)   ( A  4)2  1   j
8 64
 2. Limit cycle analysis
2.1 Describing function analysis – Concepts
Basic assumptions:
• There is only a single nonlinear component
If there are two or more nonlinear components in a system, one either
has to lump them together as a single nonlinearity, or retain only primary
nonlinearity and neglect others.
• Nonlinear component is time-invariant
Consider only autonomous nonlinear systems. Nyquist criterion applies
only to linear time-invariant systems.
• Corresponding to a sinusoidal input sin(t ) , only fundamental
component w1 (t ) in output w(t ) has to be considered.
Higher-frequency harmonics can all be neglected in analysis, as
compared with fundamental component. (filtering hypothesis)
• Nonlinearity is odd
f ( x ) is symmetric about origin. For simplicity, common nonlinearities
discussed all satisfy assumption.
 2. Limit cycle analysis
2.1 Describing function analysis – Concepts
Basic Definitions:
Consider a sinusoidal input to nonlinear element. Output of nonlinear
component w(t ) is often a periodic. Using Fourier series:
a0 
w(t )   [an cos(nt )  bn sin(nt )]
2 n 1
1  1  1 

  
a0  w(t )d (t ) ; an  w(t )cos(nt )d (t ) ; bn  w(t )sin(nt )d (t )
    

Due to assumption a0  0 and:

w(t )  w1 (t )  a1 cos(t )  b1 sin(t )  M sin(t   )

M ( A, )  a12  b12 ;  ( A, )  arctan(a1 / b1 )

j ( t  ) jt Me j (t  ) M j 1
w1 (t )  Me  (b1  ja1 )e  N ( A, )  jt
 e  (b1  ja1 )
Ae A A
2. Limit cycle analysis
2.1 Describing function analysis – Concepts
E.g.: A hardening spring w  x  x3 / 2
w(t ) is odd:
a1  0
1  3 3
b1         
3 3
[ A sin t A sin ( t ) / 2]sin( n t ) d ( t ) A A
   8
3
w1 (t )  a1 cos t  b1 sin t  w1 (t )  ( A  A3 )sin t
8
3
N ( A,  )  N ( A)  1  A2
8

Due to odd nature of nonlinearity, describing function is real, being a

function only of amplitude of sinusoidal input.
 2. Limit cycle analysis
2.1 Describing function analysis – Concepts
Applications Domain:
Any system which can be transformed into configuration in figure can be
studied using describing functions.
• Linear systems: systems contain hard nonlinearities in control loop but
are otherwise linear.
• Nonlinear systems: example in previous section.

Knowledge of limit cycle's existence is critical. Describing function

method can be used for this purpose. It can also guide design of
compensators so as to avoid limit cycles.
 2. Limit cycle analysis
2.1 Describing function analysis – Concepts
Computing describing functions:
• Analytical calculation
When nonlinear characteristics w  f ( x ) of nonlinear element are
described by an explicit function and Fourier coefficients can be easily
carried out.
• Numerical integration
For nonlinearities, w  f ( x ) is given by graphs or tables, it is convenient
to use numerical integration to evaluate describing functions.
• Experimental evaluation
Suitable for complex nonlinearities and dynamic nonlinearities. When a
system nonlinearity can be isolated and excited with sinusoidal inputs of
known amplitude and frequency, experimental determination of describing
function can be obtained by using a harmonic analyzer on output of
nonlinear element. Similar to linear elements. Difference is that not only
frequencies, but also amplitudes of input sinusoidal should be varied.
 2. Limit cycle analysis
2.1 Describing function analysis – Hard nonlinearities
Hard nonlinearities are commonly found in control systems, both in
small range operation and large range operation. Whether a system in
small range operation should be regarded as nonlinear or linear depends
on magnitude of hard nonlinearities and on the extent of their effects on
system performance.

1) Saturation
Saturation can have complicated effects
on control system performance.
• In a linearly stable system, saturation
tends to slow down response of system,
because it reduces effective gain.
• If system is unstable in linear range,
divergent behavior may be suppressed
into a self-sustained oscillation.
 2. Limit cycle analysis
2.1 Describing function analysis – Hard nonlinearities
Nonlinearity is single-valued, output is symmetric over four quarters of a
period.
In first quarter: kA sin(t ) 0  t  
w(t )   ;   sin 1 (a / A)
ka   t   / 2

4  /2 4  4  /2 2kA a a2
b1   w(t )sin(t )d (t )   kA sin (t )d (t )   ka sin(t )d (t )  [  1 2 ]
2

 0  0   A A

b1 2k 1 a a a2
N ( A)   [sin  1 2 ]
A  A A A
 2. Limit cycle analysis
2.1 Describing function analysis – Hard nonlinearities
2) Dead-zone
Output is zero until magnitude of input exceeds a certain value.
• Most common effect is to decrease static output accuracy.
• May also lead to limit cycles or system instability because of lack of
response in dead zone.
• May actually stabilize a system or suppress self-oscillations.

0 0  t  
w(t )   ;   sin 1 ( / A)
k ( A sin(t )     t   / 2

4  /2


b1  k ( A sin(t )   )sin(t )d (t )
0

2kA  1   2
 (  sin  1 2 )
 2 A A A

2k  1   2
N ( A)  (  sin  1 2 )
 2 A A A
 2. Limit cycle analysis
2.1 Describing function analysis – Hard nonlinearities
3) Backlash and hysteresis
Often occurs in transmission systems. It is caused by small gaps which
exist in transmission mechanisms.
A critical feature is its multi-valued nature. Which one occur depends on
history of input.
Multi-valued nonlinearities lead to energy storage in system. Energy
storage is a frequent cause of instability and self-sustained oscillation.
 2. Limit cycle analysis
2.1 Describing function analysis – Hard nonlinearities
w(t ) is neither odd nor even.
In one cycle:
w(t )  ( A  b)k  / 2  t    
w(t )  ( A sin(t )  b)k     t  3 / 2   sin 1 (1  2b / A)
w(t )  ( A  b)k 3 / 2  t  2  
w(t )  ( A sin(t )  b)k 2    t  5 / 2

4kb b
a1  (  1) 1 2 2
 A N ( A)  a1  b1
A
Ak  2b 2b
b1  [    (  1) 1  (  1) 2 ] N ( A)  tan 1 (a1 / b1 )
 2 A A

A phase lag (up to 90°) is introduced. It is

reflection of time delay of backlash. May create
stability problems.
 2. Limit cycle analysis
2.1 Describing function analysis – Existence
1) Nyquist criterion and its extension
Characteristic equation (loop TF) of linear system:

 ( p)  1  G( p) H ( p)  0  G( p) H ( p)  1

Nyquist criterion ( GH has no poles or zeros on j axis):

• In p plane, draw a Nyquist path enclosing right-half plane.
• Map this path into another complex plane through G( p) H ( p) .
• Determine N , number of clockwise encirclements of plot of G( p) H ( p)
around point ( 1,0) .
• Compute Z , number of zeros of  ( p) in right-half p plane, by

Z  N  P (P: number of unstable poles of  (p))

Then value of Z is number of unstable

poles of closed-loop system.
 ( p)  1  KG( p) H ( p)  0  G( p)H ( p)  1/ K
2. Limit cycle analysis
2.1 Describing function analysis – Existence
2) Existence of limit cycles
Assume that there exists a self-sustained oscillation of amplitude A and
frequency  in system. Variables in loop satisfy following relations:

x   y ; w  N ( A, ) x ; y  G( j)w
1
G( j ) N ( A,  )  1  0  G( j)  
N ( A,  )
If equation has no solutions, then nonlinear system has no limit cycles.
It is difficult to solve equations analytically, a graphical approach is
taken. If two curves intersect, there exist limit cycles. If curves intersect n
times, system has possible n limit cycles. Which one is actually reached
depends on initial conditions. Im
G ( j )

K A Re

1 N ( A)
2. Limit cycle analysis
2.1 Describing function analysis – Existence
3) Stability of limit cycles
Limit Cycle Criterion: Each intersection point of curve G( j ) and 1/ N ( A, )
corresponds to a limit cycle. If points near intersection and along
increasing A side of curve 1/ N ( A, ) are not encircled by curve G( j ) ,
then corresponding limit cycle is stable. Otherwise, limit cycle is unstable.
G ( j ) Im

Graphical Conditions(reliability): If G( j ) locus is
tangent or almost tangent to 1 / N locus, L2 1 N ( A,  ) Re
A
conclusions from a describing function analysis might
L1
be erroneous. Conversely, if 1 / N locus intersects
L1
G locus almost perpendicularly, then L1
results of describing function are usually Im Im
good. G ( j ) G ( j )
 
Re A Re
1 N ( A)
1 N ( A)

A
 2. Limit cycle analysis
2.1 Describing function analysis – Existence
E.g.: Analysis of control systems with saturated nonlinearity.
2k  a a a 2
N ( A)  arcsin  1 ( )  , A  a
  A A A 

a dN (u) 2k  1 u 2  4k
u     1 u 
2
  1  u2
A du   1  u2 1  u2  
1 1
a 1 , k  2    0.5 ,   
N (a ) N ( )
1 0.1  0.2  15
K  15  x   7.07  G( jx )   1    7.07 , A  2.5
0.1  0.2 0.1  0.2

0.02  K 0.5  0.3 j

 0.5  K max   7.5
0.3 0.02 ①
1 ②

N ( A) -0.5
K -1
r (t )  0 e(t ) 0 c(t )
a 1 s (0.1s  1)(0.2 s  1)
k 2
 
 2. Limit cycle analysis
2.2 Phase plane analysis – Concepts
Graphical method for studying second-order systems, introduced by
mathematicians Henri Poincare (1885).
One of the most important techniques, since there is usually
no analytical solution for a nonlinear system.

A number of useful properties.

 As a graphical method, it allows us to visualize what goes on in a
nonlinear system starting from various initial conditions, without having
to solve nonlinear equations analytically.
 It is not restricted to small or smooth nonlinearities, but applies equally
well to strong nonlinearities and to hard nonlinearities.
 Some practical control systems can indeed be adequately approximated
as second-order systems, and phase plane method can be used easily
for their analysis.
 2. Limit cycle analysis
2.2 Phase plane analysis – Concepts
1) Phase portraits
A second-order nonlinear system:
x1  f1 ( x1, x2 ) , x2  f 2 ( x1, x2 ) ; x1  x2 , x2  f ( x1, x2 )

Phase plane: geometrically, state space of this system is a plane having

x1 , x2 as coordinates.
Trajectory: given a set of initial conditions x(0)  x0 , system defines a
solution x (t ) . With time t varied from 0 to infinity, solution can be
represented geometrically as a curve in phase plane.
Phase portrait: a family of phase plane trajectories corresponding to
various initial conditions.

E.g.: mass-spring system.

x  x  0  x(t )  x0 cos(t )  x 2  x 2  x02

 2. Limit cycle analysis
2.2 Phase plane analysis – Concepts
2) Singular points
A singular point is an equilibrium: f1 ( x1, x2 )  0 , f 2 ( x1, x2 )  0
Slope of phase trajectory: phase trajectories will not intersect.
dx2 f 2 ( x1 , x2 ) dx2 f 2 ( x1 , x2 ) 0
    , (singular point)
dx1 f1 ( x1 , x2 ) dx1 f1 ( x1 , x2 ) 0

E.g.: Nonlinear system. x  0.6 x  3x  x 2  0

Note that t does not explicitly appear in phase plane.

x  xt  t  x / x
x
x  dx / dt  dt  dx / x  t  t0   (1 / x)dx
x0

A phase portrait may have a priori known symmetry properties by

examining system equation(odd, even, origin). (symmetry of slopes)
dx2 f ( x1 , x2 )
x  f ( x, x )  0  
dx1 x
 2. Limit cycle analysis
2.2 Phase plane analysis – Roughly sketch
1) Analytical method
• Solving equations x1  f1 ( x1, x2 ) , x2  f 2 ( x1, x2 ) for x1 (t ), x2 (t ) , and then
eliminating time t leading to a trajectory g ( x1, x2 , c)  0 .
• Directly eliminating time variable, and then solving for a functional
relation between x1 , x2 .
dx2 f 2 ( x1 , x2 )

dx1 f1 ( x1 , x2 )
E.g.: mass-spring system.
dx
x  x  0 ; x  (dx / dx)(dx / dt )  x  x  0  x 2  x 2  x02
dt
E.g.: satellite control system. (piecewise linear)
 d Ud
 U   2  2U  c1 ;   U   2  2U  c1

Marginally stable
 2. Limit cycle analysis
2.2 Phase plane analysis – Roughly sketch
2) Method of isoclines
An isocline is defined to locus of points with a given tangent slope.

dx2 f 2 ( x1 , x2 )
   f 2 ( x1 , x2 )   f1 ( x1 , x2 )
dx1 f1 ( x1 , x2 )
dx2 x
E.g.: mass-spring system. xx0    1    x1   x2  0
dx1 x2
E.g.: Van der Pol equation
dx 0.2( x 2  1) x  x
x  0.2( x  1) x  x  0 
2
    0.2( x 2  1) x  x   x  0
dx x

Two step:
• Field of directions.
• Form trajectory.
 2. Limit cycle analysis
2.2 Phase plane analysis – Linear system
x (t )  k1e1t  k2e2t
Second order linear system:
x  ax  bx  0  x (t )  (k1  k2t )et
1) Stable or unstable node
t
1 , 2 are both real and have same sign. x ( t )  e (k1 cos  t  k2 sin  t )

 Stable node (sinks): 1 , 2 negative, converge to zero exponentially.

 Unstable node (sources): 1 , 2 positive, diverge from zero exponentially.

2) Saddle point (unstable)

1 , 2 are both real and have opposite signs. Almost all of trajectory
diverge to infinity. (converge only when k2  0 )
 2. Limit cycle analysis
2.2 Phase plane analysis – Linear system
3) Stable or unstable focus (spiral)
1 , 2 are complex conjugates with non-zero real parts.
A stable focus occurs when real part of eigenvalue is negative. If real part
is positive, singularity point is called an unstable focus.

4) Center point (periodic)

1 , 2 are complex conjugates with real parts equal to zero.
All trajectories are ellipses and singularity point is center of these ellipses.
2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
Phase plane analysis of nonlinear systems is related to that of linear
systems (approximated). Nonlinear systems can display much more
complicated patterns in phase plane, such as multiple equilibrium points
and limit cycles.

x  0.6 x  3x  x 2  0 x  0.2( x 2  1) x  x  0
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
1) Local behavior of nonlinear systems
Similarity to a linear system in local region of each singular point can be
formalized by linearizing nonlinear system.
 x1  f1 ( x1 , x2 )  x1  f1 ( p1, p2 )  a11 ( x1  p1 )  a12 ( x2  p2 )  H .O.T
  
 x2  f 2 ( x1 , x2 )  x2  f 2 ( p1, p2 )  a21 ( x1  p1 )  a22 ( x2  p2 )  H .O.T

f1 ( x1 , x2 ) f1 ( x1 , x2 )
a11  x1  p1 , x2  p2 , a12  x1  p1 , x2  p2
x1 x2
f 2 ( x1 , x2 ) f 2 ( x1 , x2 )
a21  x1  p1 , x2  p2 , a22  x1  p1 , x2  p2
x1 x2

We can always shift singular point to origin:

 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
It is reasonable to expect trajectories of nonlinear system in a small
neighborhood of an equilibrium point to be close to trajectories of its
linearization about that point.
If origin of linearized state equation is:
• A stable (respectively, unstable) node with distinct eigenvalues
• A stable (respectively, unstable) focus
• A saddle point
Then, in a small neighborhood of equilibrium point, trajectories of
nonlinear state equation will behave like linearized system.

E.g.: Tunnel-diode circuit (1958, 1973’s Nobel prize).

dvC di
iR  h(vR ) , iC  C , vL  L L
dt dt
h( x1 )  17.76 x1  103.79 x12  229.62 x13
 226.31x14  83.72 x15
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system

 1
 x  [h( x1 )  x2 ]
iC  iR  iL  0 x1  vC , x2 iL 1
C  x1  0.5[h( x1 )  x2 ]
    
vC  E  RiL  vL  0 1
 x  [ x  Rx  u ]  x2  0.2[ x1  1.5x2  1.2]
 2
L
1 2

u  1.2V , R  1.5k , C  2 pF , L  5 H

Three equilibrium points: Q1 : (0.063,0.758) , Q2 : (0.285,0.61) , Q3 : (0.884,0.21)

f  0.5h( x1 ) 0.5 
A  Bistable
x  0.2 0.3 

  3.598 0.5 
 1 1 
Q : A   , eigenvalue : ( 3.57, 0.33)
   0.2 0.3 
  1.82 0.5 
Q2 : A2    , eigenvalue : (1.77, 0.25)
   0.2 0.3 
  1.427 0.5 
Q3 : A3    , eigenvalue : ( 1.33, 0.4)
  0.2 0.3 
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
E.g.: pendulum equation with friction.

 x1  x2
x1  , x2 

ml  mg sin   kl   g k
 x   sin x  x2  10sin x1  x2

2 1
l m

Evaluating Jacobian matrix at equilibrium points (0,0) and ( ,0) :

  0 1
 1
Q (0,0) : A    , eigenvalue : 0.5  j3.12
 10 1
1
f  0 1 
A   
x  10cos x1 1 Q ( ,0) : A   0 1 
 10 1 , eigenvalue : 3.7, 2.7
 2 2
 


Note: If Jacobian has eigenvalues on imaginary

axis, behavior of nonlinear system near
equilibrium point is distinct from that of
linearized system.
Only available for hyperbolic equilibrium point.
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
E.g.: system has an equilibrium point at origin. Linearized state equation at
origin has eigenvalues  j .

 x1   x2   x1 ( x1  x2 )
 2 2 x1  r cos  , x2  r sin 
  r  r3 ,   1

 2
x  x1   x2 ( x1
2
 x 2
2)

Trajectories of nonlinear system will resemble a stable focus when   0

and an unstable focus when   0 .
Qualitative behavior describing a center in linearized state equation is not
preserved in nonlinear state equation.

2) Limit cycles
In phase plane, a limit cycle is defined as an isolated closed curve.
Trajectory has to be both closed, indicating periodic nature of motion, and
isolated, indicating limit nature of cycle.
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
Depending on motion patterns of trajectories in vicinity of limit cycle, one
can distinguish three kinds of limit cycles:

Stable Unstable Semi-stable

E.g.: stable limit cycle.

x1  x2  x1 ( x12  x22  1) , x2   x1  x2 ( x12  x22  1)
dr d
r  ( x12  x22 )1/2 ,   tan 1 ( x2 / x1 )   r( r 2  1) ,  1
dt dt
1 1
r (t )  , c0   1 ;  (t )  0  t
(1  c0e 2t )1/2 r02
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
Existence of limit cycles: (classical theorems)
Theorem 1 (Poincare): If a limit cycle exists in second-order autonomous
system, then N  S  1 .
N : number of nodes, centers, and foci enclosed by a limit cycle.
S : number of enclosed saddle points.
Lemma: (index theorem)
• Index of a node, a focus, or a center is +1;
• Index of a saddle is -1;
• Index of a closed orbit is +1;
• Index of a closed curve not encircling any equilibrium points is 0;
• Index of a closed curve is equal to sum of indices of equilibrium points
within it.

Theorem reveals a simple relationship between existence of a limit cycle

and number of singular points it encloses. A limit cycle must enclose at
least one equilibrium point.
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
E.g.: system x1   x1  x1 x2 , x2  x1  x2  2 x1x2

 f   1 0   f  0 1
Two equilibrium points:
 x    1 1 ,  x    1 1
(0,0)   (1,1)  
saddle stable focus

Only combination of equilibrium points that can be encircled by a

periodic orbit is a single focus. Other possibilities of periodic orbits, like a
periodic orbit encircling both equilibria, are ruled out.

Theorem 2 (Bendixon): For second-order nonlinear system, no limit cycle

can exist in a region  of phase plane in which f1 / x1  f 2 / x2 does not
vanish and does not change sign. (sufficient condition )
dx2 f 2 Green theorem
f1 f 2

dx1 f1
  f 2dx1  f1dx2  0   (
S

x1 x2
)dx1dx2  0

f1 f 2
E.g.: x1  x2 , x2  ax1  bx2  2 x12 x2  x13    b  2 x12
x1 x2
 2. Limit cycle analysis
2.2 Phase plane analysis – Nonlinear system
Theorem 3 (Poincare-Bendixon): If a trajectory of second-order autonomous
system remains in a finite region  , then one of following is true:
• Trajectory goes to an equilibrium point;
• Trajectory tends to an asymptotically stable limit cycle;
• Trajectory is itself a limit cycle.