Progress In Electromagnetics Research, PIER 79, 179193, 2008

NONLINEAR STABILITY ANALYSIS OF MICROWAVE

OSCILLATORS USING THE PERIODIC AVERAGING
METHOD

H. Vahdati and A. Abdipour

Microwave/mm-Wave and Wireless Communication Research Lab
Radiocommunication Center of Excellence
Electrical Engineering Department
Amirkabir University of Technology (Tehran Polytechnic)
Hafez Ave., Tehran, 15914, Iran

AbstractIn this paper an approach for stability analysis of

microwave oscillators is proposed. Using the perturbation theory and
averaging method, a theorem which relates the oscillation stability
to the stability of the periodic average of the circuits Jacobian is
mentioned. Using this theorem, a criterion for oscillation stability is
devised. The proposed criterion is applied to the stability analysis of
a negative resistance diode oscillators and a Colpitts oscillator. This
method is readily applicable to microwave CAD routines.

1. INTRODUCTION

Oscillators are autonomous nonlinear circuits with dierent stability

phases. Both stability and instability are present at dierent phases of
oscillator operation. Instability of bias point, causes the oscillation
start up while the stability of steady state oscillation is necessary
for a physical and measurable oscillation. Oscillation stability is
the sustainability of the oscillation amplitude and frequency in the
presence of small perturbations [1, 2]. The instability of DC bias point
can be studied by linearization of the circuit around its DC operating
point. The stability of this linearized circuit can be studied using
the stability criteria of linear circuits e.g., Nyquist diagram. But for
stability analysis of oscillation, the criteria of linear system stability
can not be used [3]. For oscillation stability, it is ubiquitous to use the
stability analysis method of nonlinear circuits [4].
Most oscillators are designed by means of linear concepts and their
performances are satisfactory at least for basic application. But many
180 Vahdati and Abdipour

oscillatory behaviors have intrinsically nonlinear nature and they are

becoming increasingly important in microwave applications e.g., using
the chaotic oscillators for secure spread spectrum communication and
radar [5]. Also a fully nonlinear analysis can give a more accurate
evaluation of the actual performance of the oscillator.
On the other hand, the microwave oscillators have some special
problems which do not exist at lower frequency ones. For example at
microwave and millimeter wave ranges, the device models change and
some nonlinear dynamic elements like junction capacitances e.g., Cjc
and Cje in HBT transistors, could not be neglected. Nonetheless the
harmonic balance HB method which is widely used for the analysis
of nonlinear microwave circuits, has deciency in stability analysis of
its computed solutions. As shown in Fig. 1, in HB, the circuit is
divided into two distinct linear and nonlinear networks. The nonlinear
network is analyzed in time domain while the linear network is analyzed
in frequency domain. The Fourier transform and its inverse are used for
transforming these two domains. Expanding the voltage at the ports of
nonlinear networks in the Fourier series, the current components of the

I N +1 I1 I1
+ ZZs s(( )) +
VS (t ) VN +1 N +1 1 VN NL 1

Linear Nonlinear
Subcircuit Subcircuit

I N +2 IN IN
+ Z L () +
VL (t ) VN +2 N +2 N V1 NLN

Figure 1. Separation of the linear and nonlinear networks in harmonic

balance method.
Progress In Electromagnetics Research, PIER 79, 2008 181

linear network can be calculated using its admittance matrix. Using

the ports voltage, the current waveforms and hence the components of
the current of the nonlinear networks can be calculated. In fact, the
main idea of harmonic balance approach is the KCL for the components
of current carrying between the linear and nonlinear networks [6].

I + I = 0 (1)

where I and I are the current vectors of the linear and nonlinear
networks. The current vector of the linear network is:
I = YS VS + YN N V (2)
where VS is the voltage sources and YS and YN N are the admittance
matrixes which can be obtained using systematic approaches e.g. loop
set matrix. The current vector of nonlinear networks at the common
terminals is [6].
I = IG + jQ (3)
where IG is the vector of nonlinear current source and Q is the vector
of the charge of nonlinear capacitors. The general HB equation can
be expressed as (4).
F (V ) = IS + YN N V + IG (V ) + jQ(V ) = 0 (4)
Equation (4) represents an equation that should be solved to obtain
the voltage V . There are various techniques to solve the harmonic-
balance equation (4). Among them, the Newton technique is simpler,
faster and has high gradient convergence [6]. HB method considers
only the steady state solution and ignores its time evolution, hence the
computed solution could be unstable [2]. The commercial microwave
CADs have no tools for stability analysis of oscillation. These facts
along with the modern expensive MMIC technology which requires
comprehensive simulation and ensuring the rst pass design, clarify
the importance and the need for the nonlinear stability analysis of the
modern microwave oscillators. It should be also noted the dierences
between the stability of numerical method for analyzing the circuit and
the stability of the circuit itself [7, 8].
There is a large amounts of studies in oscillation stability [2].
Using the characteristic matrix, Fluquet [9] derived a criterion for
oscillation stability. This method is based on the characteristic
matrix, but derivation of the characteristic matrix even for the simple
circuits is cumbersome. Assuming the negative resistance model for
the microwave oscillator and using the describing function method,
Kurokawa derived a criterion for oscillator stability [10]. While
182 Vahdati and Abdipour

Kurokawas criterion is widely used in texts, it has the inherent

restriction of the describing function method i.e. it considers only
the fundamental harmonic. Rizzoli and Lipparini proposed a general
method for stability analysis of microwave circuits, but their method is
very complex and dicult to implement [1113]. Bifurcation is another
tool for stability analysis of nonlinear circuits [14]. Based on the
bifurcation theory, Suarez proposed a method for stability analysis.
But this method is suitable only for the forced oscillators [15, 16].
Mons used the normalized determinant function for stability analysis of
microwave circuits. He used auxiliary source in commercial microwave
CAD to calculate it [17]. Mons method is suitable for forced circuits
too.
In the following sections, the theoretical basis for an oscillator
stability criterion has been reviewed and by stating some theorems,
this criterion has been proposed. The main idea of this method is
mapping the oscillator stability problem to the stability analysis of the
equilibrium point of a linear time invariant system which is derived
by averaging the Jacobian matrix of the oscillator state equations over
one period of oscillation.

2. THEORETICAL BASIS

Periodic averaging is a suitable method for studying the stability of

(5):
x = f (t, x, ) (5)
where f ([0, ) D [0, 0 ]) n is a vector of bounded functions
with continuous rst and second order derivatives, D is the domain
of x. f is also periodic in time, i.e.:
n N f (t, x, ) = f (t + nT, x, ) (6)
in which is a small positive parameter.
If xs (t, ) and xp (t, ) are the steady state and perturbed responses
of (5), the dynamic system (5) is stable if and only if [18]:
lim xs (t, ) xp (t, ) 0 (7)
t
0

Averaging is a suitable method for verifying (7). The averaged system

of (5) is:
x = fav (x) (8)
where  T
1
fav (x) = f (t, x, = 0)dt (9)
T 0
Progress In Electromagnetics Research, PIER 79, 2008 183

By these preliminary denitions, it has been proved [19, 20] that:

Theorem 1 : Suppose that f (t, x, ) is a function on [0, )
D [0, 0 ] with continuous and bounded partial derivatives. Also
f (t, x, ) is periodic in time with period of T and is a small positive
number. If x(t, ) and xav (t, ) be the solutions of (10) and (11)
respectively
x = f (t, x, ) (10)
and
x = fav (x) (11)
and if p D be the asymptotic stable equilibrium point of the
averaged system (11), then the positive constant exists so that if
0 < < , then the equation x = f (x, t, ) has a unique asymptotic
stable periodic solution with period of T .
Corollary 1 : If both (10) and (11), have equilibrium point at
x = 0 and if it be the asymptotically stable equilibrium point of (11),
then the origin is asymptotically stable equilibrium point of (10).
The theorem 1 and its corollary express a strong relation between
the stability of the periodic response of a system with the stability of
its averaged system over one period. Now we apply the above results
to the special case (12):

x = f (t, x, ) = A(t)x (12)

in this case fav is the termwise average of A(t) i.e.,

fav (x) = Ax

1 T
A = A(t)dt (13)
T 0
The averaged system is a time invariant system which is expressed
by matrix A and according to the above theorem (12) is asymptotic
stable if all the eigenvalues of A have negative real parts. Here
another corollary has been stated and proved which is the base of
the proceeding discussions.
Corollary 2 : The linear time varying system (14)

x = A(t)x (14)

where A(t) is periodic with period of T , is stable if all the eigenvalues

of A, have negative real parts. The proof of this corrollary has been
given in Appendix A.
184 Vahdati and Abdipour

3. FORMULATION OF OSCILLATOR STABILITY

The corollary 2 can be used for stability analysis of free running

oscillators. Free running oscillator is an autonomous dynamic system
and using the state variables formulation, it can be expressed as (15):

x = f (x) (15)

Here it is assumed that (15) has an oscillatory solution. Even though

nding the oscillatory response of the circuits, is a complex problem
[21], but in this paper it is assumed that an oscillatory solution of (15)
has been found by HB and solely the stability of this solution should be
studied. For stability analysis, a small perturbation should be exerted
to the steady state response of (15). Suppose that xs (t) is the steady
state solution and (t) is a small perturbation, the perturbed circuit
is (16):
f
xs + = f (xs + ) = (xs (t))(t) (16)
x
where fx is the Jacobian matrix of (16) and henceforth we name it
Jacobf (t). It is important to note that (16) is a linear time varying
ordinary dierential equation. Oscillation stability of (15) is equivalent
to the stability of the equilibrium point (t) = 0 of (17):

= Jacobf (t)(t) (17)

As xs (t) is periodic with period of T , the Jacobian in 21 is periodic

too i.e.:
Jacobf (t) = Jacobf (t + T ) (18)
Using the previous discussions, the following method for stability
analysis of microwave oscillators can be proposed.
(i) Derive the state equations of the oscillator 15.
(ii) Calculate the steady state response of (15) by the harmonic
balance method.
(iii) Perturb the system and derive the perturbed circuit equations of
(17).
(iv) Compute the averaged Jacobian matrix of oscillator over one
period.
(v) Check the average matrix Jacobf for stability e.g., test its
eigenvalues or use the Nyquist diagram.
Progress In Electromagnetics Research, PIER 79, 2008 185

R 0 .75 

L 0.224 H C 37.26nf D

Figure 2. Diode oscillator.

Briey it can be said that in the proposed method, the oscillation

stability can be deduced from the stability analysis of averaged
Jacobian matrix which describes a linear time invariant system. In
the next section the proposed method has been applied to a diode and
a Colpitts oscillator.

4. APPLICATION OF THE PROPOSED METHOD TO

STABILITY ANALYSIS OF OSCILLATORS

As the rst example, the proposed criterion is applied to stability

analysis of a diode oscillator. Although its frequency is not in the
microwave range, it is widely studied in literature [2, 10] and it is
added here only for the comparison of results with the literature. A
simple model of this oscillator has been shown in Fig. 2. The I-V
characteristic of the diode is:
Id = avd + dvd3 (19)
where vd is the diode voltage, a is negative and equals to 0.2 and
d = 0.0375. Using the capacitor voltage and inductance current as the
state variables, the state equations are:
diL R 1
= iL + vC
dt L L
dvC 1 1
= iL 3
(a.vC + d.vC ) (20)
dt C C
The Jacobian matrix of this system is:

R 1
L
J =

L

(21)
1 1
(a + 3d.vC
2
)
C C
186 Vahdati and Abdipour

To study the stability of this oscillator, the average of (21) should be

calculated over one period of its steady state response. The steady
state response of this circuit can be calculated using the harmonic
balance. If vd (t) and id (t) be the diode voltage and current and i(t) be
the linear parts current, the Fourier series expansion of these quantities
are:

k=N
vd (t) = Vk ejk
k=N

k=N
id (t) = IDk ejk
k=N

k=N
i(t) = Ik ejk (22)
k=N

The admittance of the linear part is:

1
Yl (j) = jC + (23)
R + jL
Using these expansions, the harmonic balance equations of this circuit
is:

yl (j0) 0 ... 0 V0 ID0 0
V1 ID1 0
0 yl (j) . . . 0
. + .
.. ..
= ..
.
0 ... 0 yl (jN ) VN IDN 0
(24)
Using the oscillation frequency as the optimization parameter and by
5 harmonics and Newton-Raphson method, the steady state solution
has been derived. This solution has been shown in Fig. 3. To check the
oscillation stability, the proposed approach of the previous section has
been used. The averaged Jacobian matrix over one period of steady
state solution is:


0.3348 0.4464
Jav = 107 (25)
2.6817 0.1330

The eigenvalues of (25) are:

1 = 107 (0.1009 + i 1.0689)

(26)
2 = 107 (0.1009 i 1.0689)
Progress In Electromagnetics Research, PIER 79, 2008 187

As the real parts of both eigenvalues are negative, the averaged system
is asymptotically stable and hence the oscillator is stable. The stability
of this oscillator has been shown in [2] using the time domain analysis.
In the next example, the stability of a Colpitts oscillator at 5.2
GHz has been studied. The instability of this circuit can cause chaotic
behavior [22]. Using the Ebers Moll II model [24] for the HBT
transistor, the oscillator and its equivalent circuit has been shown in
Fig. 4 and Fig. 5.
Although the circuit has 4 capacitors and 1 inductor, as they
construct two capacitive loops, there are only 3 independent state
equations. Using v1 , v2 and iL as the state variables, the state
equations is (27):
diL
L = v1 + v2 + iL
dt
dv1 

dv
2
C1 + C2 + Cbe (v2 ) (v2 )Cbe (v2 )
dt dt
 v2  v1 +v2 v2 +VDD
= IS e VT 1 R IS e V T R E 1


dv (27)
1
C1 + Cbc (v1 v2 ) (v1 + v2 )Cbc (v1 v2 )
dt


dv
2
+ Cbc (v1 v2 ) (v1 + v2 )Cbc (v1 v2 )
dt
v1 +v2  v2  v1 +v2
= iL F IS e VT
1 IS e VT
1
RL
To apply the proposed method for the stability analysis of this
oscillator, the steady state response should be calculated using the
HB method. The procedure for applying HB method to this circuit
is shown in Fig. 6. The linear part is determined by the admittance
matrix:

jc1 + R1C + jL1 jc1 R1C jL
1
0

jc1 jc1 +jc2 + R1E 0 R1E

Yn =
R1C 1
0 1
+ jL1 0
jL RE
0 R1E 0 1
RE
(28)
The nonlinear elements have similar congurations. They consist of a
diode, a junction capacitor and a dependent current source. Then if
the voltage at the nonlinear ports be v1 and v2 , the current at each
188 Vahdati and Abdipour
2.0

1.5

1.0

0.5
ts(VC), V

0.0

-0.5

-1.0

-1.5

-2.0

800

600

400

200
ts(I_L.i),, mA

-200

-400

-600
0

-800
0.0 0.1 0 .2 0..3 0.4 0.5 0 .6 0.
0.7 0.8 0.9 1 .0 1.1 1 .2 1.3

time, usec

Figure 3. Steady state response of diode oscillator.

v1
IL

RL L
C1

v2

RE C2 VCC
z

VEE

Figure 4. Colpitts oscillator.

Progress In Electromagnetics Research, PIER 79, 2008 189

v1
IL

Cbc D1 RID2

RL L
C1
C be D2 F ID 1

v2

RE VCC

C2

VEE

Figure 5. Equivalent model of Colpitts oscillator.

port can be expressed as:

 d  v2
C0jc v1 + R Is e VT 1
v1
i1 = Is e VT 1 +
dt 1 v1
0
(29)
 d  v1
C0je v2 + F Is e VT 1
v2
i2 = Is e VT 1 +
dt 1 v2
0
Using the above descriptions of the nonlinear ports, the harmonic
balance equations have been derived and the steady state responses of
v1 , v2 and iL are shown in Fig. 7. Analytic derivation of the Jacobian
matrix of (27) is cumbersome and it has been calculated numerically.
The eigenvalues of the averaged Jacobian matrix are:
1 = 107 (1.67)
2 = 107 (0.5298) (30)
3 = 107 (0.05351)
As all the eigenvalues have negative real parts,the averaged system is
stable. Therefore using the proposed criterion, the oscillation is stable.
190 Vahdati and Abdipour

Nonlinear Part Linear Part

L
v1
iL

2ID 2 RC
VCC
D1
Cbc
C1

v2

RE
 f I D
1

VEE
C2
Cbe D2

Figure 6. Separation of linear and nonlinear part for harmonic

balance analysis.

0.5
IL , mA

0.5
0 1 2 3 4 5 6
Time , Second 10
x 10
20
VC1, Volt

20
0 1 2 3 4 5 6
Time , Second 10
x 10
1
VC2, Volt

1.5
0 1 2 3 4 5 6
Time , Second 10
x 10

Figure 7. Steady state response of Colpitz oscillator.

Progress In Electromagnetics Research, PIER 79, 2008 191

The oscillation stability of this oscillator has been veried using the
time domain analysis [2].

5. CONCLUSION

In this paper a method for stability analysis of microwave oscillator has

been proposed and its theoretical basis has been discussed. It is based
on the perturbation theory and averaging method. By deriving the
perturbed system which is a linear time varying with periodic behavior
and checking the stability of its average, the oscillation stability can
be veried. The feasibility of the proposed method has been shown by
analyzing the stability of a diode and a Colpitts oscillators. Because
of its theoretical basis and its simplicity, it is promising for attaining
a general yet simple stability analysis of microwave oscillator.

ACKNOWLEDGMENT

This work is supported by the Iranian Telecommunication Research

Center ITRC.

APPENDIX A. PROOF OF COROLLARY 2

Proof : It has been shown [24] that every stable system has a Lyapunov
function like V (x) where:
V (x) > 0 and
V (x) < 0 (A1)
Expanding the time derivative of Lyapunov function V (x):
V (x) < 0 xV < 0 (A2)
where is the gradient of V . Now consider the system x = A(t)x
where is a small positive number and use the same Lyapunov function
as used in (A1). To avoid confusing we call it V1 (x) which is obviously
positive denite and
V1 (x) < 0 xV1 < 0
A(t)xV1 < 0 where >0 (A3)
As is a positive number, it follows from (A2) that V1 (t) < 0. This
means that the system is stable in the sense of Lyapunov. The above
corollary states that to show the stability the equilibrium point of a
linear time varying system (14) with periodic A(t) it suces to show
the stability of its averaged system on its period.
192 Vahdati and Abdipour

REFERENCES

1. Giannini, F. and G. Leuzzi, Nonlinear Microwave Circuit Design,

John Wiley, New York, 2004.
2. Suarez, A. and A. Quere, Stability Analysis of Nonlinear
Microwave Circuit, Artech House, Norwood, 2003.
3. Jackson, R. W, Rollett proviso in the stability of linear microwave
circuits A tutorial, IEEE Trans. MTT., Vol. 54, No. 3, 993
1000, 2006.
4. Obregon, J., J. Nallatamby, M. Prigent, M. Camiade, and
D. Rigaud, RF and Microwave Oscillator Design, Artech House,
Norwood, 2002.
5. Shi, Z. G., S. Qiao, and K. S. Chen, Ambiguity function of direct
chaotic radar employing microwave chaotic Colpitz oscillator,
Progress In Electromagnetics Research, PIER 77, 114, 2007.
6. Maas, S. A., Nonlinear Microwave and RF Circuits, Artech House,
Norwood, 2003.
7. Kung, F. and H. T. Chuah, Stability of classical nite diference
time domain (FDTD) formulation with nonlinear element
A new perspective, Progress In Electromagnetics Research,
PIER 42, 4989, 2003.
8. Wu, C. G. and G. X. JiangKung, Stabilization procedure for
time-domain integral equation, J. of Electromagn. Waves and
Appl., Vol. 21, No. 11, 15071512, 2007.
9. Coddington, E. A. and N. Levinston, Theory of Ordinary
Dierential Equations, Mc. Graw-Hill Inc., New York, 1983.
10. Kurokawa, K., Some basic characteristics of broadband negative
resistance oscillator circuits, Bell System Technical Journal ,
19371955, 1969.
11. Rizzoli, V. and A. Lipparini, General stability analysis of periodic
steady state regimes in nonlinear microwave circuits, IEEE
Trans. MTT., Vol. MTT 33, 3037, 1985.
12. Rizzolli, V., A. Lapparini, A. Cotanzo, F. Mastri, C. Cecchetti,
A. Neri, and D. Masotti, State of the art harmonic balance
simulation of forced nonlinear microwave circuits by piecewise
technique, IEEE Trans. MTT., Vol. 40, No. 1, 1227, 1992.
13. Rizzoli, V., A. Neri, and D. Masott,Local stability analysis of
microwave oscillators based on Nyquists theorem, IEEE Trans.
Microwave and Guided Wave Letters, Vol. 7, No. 10, 341343,
1997.
Progress In Electromagnetics Research, PIER 79, 2008 193

14. Makeeva, G. S., O. A. Golovanov, and M. Pardavi-Horvath,

Mathematical modeling of nonlinear waves and oscillations in
gryomagnetic structures by bifurcation theory methods, J. of
Electromagn. Waves and Appl., Vol. 20, No. 11, 15031510, 2006.
15. Suarez, A., S. Jeon, and D. Rutledge, Stability analysis and
stabilization of power ampliers, IEEE Microwave Magazine,
Vol. 7, No. 5, 145151, 2006.
16. Suares, A., V. Iglesias, J. M. Collantes, J. Jugo, and J. L. Garcia,
Nonlinear stability of microwave circuits using commercial
software, IEE Elec. Letter , 13331335, 1998.
17. Mons, S., J. C. Nallatamby, R. Quere, P. Savary, and J. Obregon,
Unied approach for the linear and nonlinear stability analysis
of microwave circuits using available tools, IEEE Trans. MTT.,
Vol. 47, No. 12, 24032410, 1999.
18. Sanchez, D. A., Ordinary Dierential Equation and Stability
Theory, An Introduction, Dover, New York, 1960.
19. Khalil, H., Nonlinear Systems, Printice Hall, N.J., 2002.
20. Sagar, V., Nonlinear Control Systems, Academic Press, New York,
1999.
21. Migulin, V. V., Basic Theory of Oscillation, Mir Publisher,
Moscow, 1983.
22. Shi, Z. G. and L. X. Ran, Microwave chaotic Colpitts oscillator:
design, implementation and applications, J. of Electromagn.
Waves and Appl., Vol. 20, No. 10, 13351349, 2006.
23. Mokari, M. E., S. Ganesan, and B. Blumgold, Systematic non-
linear model parameter extraction for microwave HBT device,
IEE, 10311036, 1993.
24. Slotine, J. E. and W. Li, Applied Nonlinear Control , Printice Hall,
New Jersey, Englewood Clis, 1991.