Automatica, Vol. 13, pp. 153-159. Pergamon Press, 1977.

Printed in Great Britain

Synchronization Characteristics of
Controllable Oscillators*
JACOB TALt

The conditions for synchronizing a controllable oscillator, such as an electric motor or

an electronic circuit, with an external clock are related to the limit cycles which occur
during loss of lock.

Key Word Index-Synchronization; phase-locked loops; stepping motors; limit cycles; nonlinear control
systems; frequency control; speed control.

Summary-This paper considers the problem of synchronizing a itions. It is driven by a periodic voltage signal and,
controllable oscillator with some external frequency and de- under desired conditions, the shaft position should
termines the conditions for synchronization. The requirements
are determined analytically and the results are applied to several synchronize with the input voltage. Further details
examples in various fields. about this application can be found in [8-12].
In all the above cases, if no control signal is
1. INTRODUCTION applied, the oscillator will run at its natural
THE TERM controllable oscillator refers, in this frequency, which may be zero or constant. When
paper, to a device which operates periodically and synchronization is desired, it appears to depend on
whose period is controlled by some external the system parameters, as well as on the input
variable. It is often required to synchronize such frequency. However, no general solution is available
oscillators with an external periodic signal with for this problem. The objective of this paper is to
different frequency; however, the synchronization provide a unified model for those systems and
depends on the system parameters. The problem of determine the conditions for synchronization.
determining the conditions under which synchroni-
zation occurs is important and has applications in 2. SYSTEM DESCRIPTION AND
various fields. Typical examples are the following MATHEMATICAL MODEL
(a) Phase-Locked Loops (PLL)--in this appli- The synchronization system under discussion is
cation, a voltage controlled oscillator (VCO) is illustrated by Fig. 1. The phase detector is a
connected in a closed-loop form and is forced to nonlinear element which compares the two signals
synchronize with an incoming periodic signal. The sl and s2 and produces an error signal, u, that is a
resulting control voltage which drives the VCO is function of their phase difference
proportional to the input frequency, and therefore a
common use ofPLL is as an FM demodulator. This u=g(¢) (1a)
application is described widely in the literature. For where
example, see [1--4].
(b) Phase-Locked Servo Systems (PLS)--the (1 b)
oscillator in this case is a de motor with an
The phase detector may be a separate component,
incremental encoder on its shaft. The encoder emits
as in PLL and PLS, or an integral part of the system
a pulse for a given degree of rotation, and therefore
as in a step motor. The most common function for a
the frequency increases with the motor velocity, or
phase detector is sinusoidal where g(¢) =A sin¢;
the driving voltage. The main application of this
however, with the development of digital phase
system is to synchronize the encoder frequency with
detectors, one can find triangular as well as
a fixed frequency, thereby achieving constant
sawtooth functions. In any case, since g(¢) is
velocity over long terms. This is a recent develop-
periodic, it can be represented by a Fourier series
ment which is described in [5-7].
w
(c) Step Motor Drive-a step motor is an
electromechanical actuator with discrete shaft pos-
g(¢)=Asin¢+ L Aksin(k¢+ek). (2)
k=2

Note that the phase difference,¢, is the integral of

*Received 30 December 1975; revised 29 April1976; revised 7 the frequency difference between the input and the
September 1976. The original version of this paper was not output signals. When the two signals are not
presented at any IF AC meeting. It was recommended for synchronized, ¢ includes a dominant term which
publication in revised form by associate editor P. Dorato.
tUniversity of Utah, Department of Electrical Engineering, increases linearly. In this case, it is shown in
Salt Lake City. UT 84112. U.S.A. Appendix B that the components of u, generated by
153
D
154 JACOB TAL

Position
sensor

FIG. 1. Block diagram representation of synchronization system.

the Ak, k ~ 2, terms, have high frequencies; therefore, Thus, 0 1 and (} are relative to the free-running
they are attenuated by the following low-pass filter phase, w 0 t. Next, it is desirable to combine the phase
and they may be ignored. This leads to a simplified detector gain, A, with the low-pass filter. This
expression for g(<f.l ). requires some new definitions

g(<f.l) =A sin¢. (3) F(s)=A.H(s)}

x=u/A
(10)
T =;v/A
The control signal, y, is the difference between the
error signal, u, and the input v, which represents the z=r!A
constant friction and load torque. Therefore,
These lead to the following simplified model
y=u-v. (4)
(}1 =W1t
The low-pass filter, H (s ), represents the transfer
¢=& 1 -e
functions of motors, filters, and all the other
dynamic components that stand between the con- x =sin <P
trol signal and the controlled frequency w. Using the z=x-T (11)
Laplace transformations, the relationship between y w(s)=F(s)z(s)
and w is given by equation (5) 1
O(s)= -w(s)
s
w(s) = H (s )y(s ). (5)
The simplified model, equation (1 1-), is called the
The actual oscillator frequency, wout, equals the free- baseband model and is represented by Fig. 2.
running frequency, w 0 , and the controlled fre-
quency, w.

Wout = Wo + W. (6)

Finally, the output phase, (}out• is the integral of (J)OUI'

Therefore,
FIG. 2. Baseband model for synchronization systems.

(7)
Note that in this model, w 1 represents the
difference between the desired frequency and the
The position sensor is a modulator which produces free-running frequency.
a periodic function S2 (00 u 1) of the phase. In most
cases the function S2 is sinusoidal or a square wave. 3. SYNCHRONIZATION REQUIREMENTS
The above model may be simplified by consider- The synchronization of the loop is defined as the
ing the phases ein and (}out as the signals instead of s1 condition where the phase difference, ¢, remains in
and S2 .1t can be further simplified by introducing the range- n/2 < <P < n/2. The requirements for
synchronization are actually global stability, or at
(8) least some regional stability. However, since this is
difficult to determine, the same requirements are
and stated as follows
(a) Existence of an Equilibrium Point-in order
(9) for the system to operate, the input frequency, w 1 ,
 Synchronization characteristics of controllable oscillators 155

must be in the range for which an equilibrium point In view of this, we may investigate the limit cycles
exists. This frequency range is called the lock range, of the system and the conditions for their existence.
and since x is bounded by 1, the lock range, wv is If limit cycles cannot exist, we conclude that the
given by phase difference is constant and the loop must lock.
Although this argument is not mathematically
rigorous, it is justified by experimental results, and it
is accepted by all researchers in this area [ 1-4].
This range is also called the speed range for PLS and
the slew range for step motors. 4. LIMIT CYCLE ANALYSIS
(b) Local Stability-the system must be locally Suppose that the system is out of lock and the
stable at the equilibrium point. However, this is output frequency, w, is different from w 1 • In the
easily determined from the linearized model of the steady state, the output of the phase detector, x, is a
loop. periodic function of the time and includes a de term,
(c) Pull-in Range-once it is established that a a fundamental frequency and its harmonics. In the
stable equilibrium point exists, it remains to following analysis it is assumed that the harmonics
guarantee that the system response will converge to are attenuated by F(s) and, therefore, we will
it from the initial conditions. The set of initial consider x of the form
conditions and inputs for which this holds is called
the pull-in range, the capture range, and the start- x=asin(yt+b)+b. (13)
stop range. Although the pull-in range is a function
of the input signals and the initial conditions, it may Consequently, the phase difference, c/J, will include a
have a subset which depends only on the input sinusoidal component and a linearly growing term
signals. This subset, denoted by A, represents the due to the difference between wand w 1 .
cases where pull-in will occur for all possible initial
conditions. The complement of this subset in the cjJ = nt + £ sin (')' t + 1/t ). (14)
pull-in range is denoted by B, and it represents cases
where pull-in is a function of both initial conditions
and input signals. The pull-in range and the subsets Next, it is desired to determine the equations of
A and B are illustrated graphically in Fig. 3, where the limit cycle. The approach taken here is to
evaluate the Dual-Input-Describing-Function of sin
cjJ for c/J, as given by equation (14). This is done by
Input
variables evaluating sin cjJ and equating it to (13). More about
I

~·
this method can be found in [13]. The details of the
work are given in Appendix A, and the results are

~--- a=1 (15)

. y=Q (16)

'/;'// //;?//I Initial conditions

b=O
b=tt:sinljt.
(17)
(18)
FIG. 3. The pull-in range.

Noting that the term t: sin (Qt + 1/t) is the response of

the shaded area represents the pull-in range. Since the linear section - F(s)/s to sin Ot, we can write
the pull-in range which depends on initial con-
ditions is not very useful, section B of the pull-in £=---
IFUO)I (19)
range is ignored by engineers who concentrate on n
section A. Furthermore, the pull-in frequency range and
is defined as the range of those frequencies which,
along with the input T, result in section A; this (20)
indicates locking regardless of initial conditions.
In order to determine the pull-in frequency, we
argue that if the system is out of lock, the oscillator In view of these, (18) becomes
will run at some frequency, w, which is different from
the input frequency, w 1 . The frequency difference b= R(Q) (21)
results in an increasing phase error, c/J, which forces 2Q
the phase detector to vary periodically, producing a where
limit cycle in the system. R(Q)=Re[FUQ)]. (22)
!56 JACOB TAL

Also note that the average value of the output When w 1 is larger than the pull-in frequency,
frequency is given by w 1 - n. Since this is the point din Fig. 4, (24) will have two solutions. 0 1 and
response of F(s) to the constant b- T, we can write it Q2 , indicating that two limit cycles are possible. The
as conditions for locking requirements can be sum-
marized now as the lock range, equation 12, local
R(Q) ) stability, and the pull-in condition, equation (27).
w 1 -Q=F(O) ( ~- T (23)
The following section shows how these results are
applied to various examples.
or
5. APPLICATION OF RESULTS
w 1 +F(O)T=f(Q) (24) Case [~Phase-Locked Loop (PLL)
Consider a second order PLL with a filter
where
s+0.5
F(s)= ---
j(Q)=O+ F(O)R(Q). (25) s+O.l
2Q
with no constant load, T = 0. The filter functions are
Equation (24) is the key to determining the pull-in
range, since it relates the inpute frequency and F(O) = 5
torque to the limit cycle frequency. In order to
illustrate its use, consider its graphical repre- and
sentation shown in Fig. 4.

The resulting lock range, as given by (12), is

and the pull-infrequency is

-_I-
I
wp= min I 0+5(Qzz+O.~~-J.
!1>0 L 2n(n +0.01)

This minimum is found to be wP = 3.16 at Q = 1.6.

If the input frequency is increased above wP to 3.5,
two limit cycles will occur at frequencies nl = 1.06
and 11 2 =2.5.
Simulation of this system was performed by
Viterbi[1, p. 60, Figs 3.13-3.17] where it was found
that the pull-in frequency is between 3.0 and 3.5.
FIG. 4. Graphical representation of equation (24). Furthermore, it was shown that when wP = 3.5,
simulation results in two limit cycles at frequencies
of 1.15 and 2.6. This indicates a good agreement
Assuming that j(Q) has a minimum, as shown in with analytical results.
the figure, it is clear that as long as (w 1 + F(O)T) is
between c and -c, equation (24) has no solution Case 2-Phase-Locked Servo (PLS)
and, therefore, no limit cycle exists, and convergence Consider a phase-locked servo system for speed
is guaranteed. Thus, the pull-in frequency range is control where the mechanical time constant of the
given by the condition motor is the dominant factor in system dynamic.

w P = mm
n>o
. [n + --'---'---
u
F(O)R(Q)J
2Q
(26)
The corresponding filter is given by

k
H(s)=---- (28)
sr + l
and
where r=0.05sec and k= 100 when the motor
(27) velocity is measured in radjs. Also suppose that the
 Synchronization characteristics of controllable oscillators 157

amplifier is biased to drive the motor at an angular Case 3--Step Motor

velocity of w 0 = 200 rad/sec. Therefore, any input Consider a permanent magnet step motor driven
frequency discussed in this example is in reference to by a sinusoidal current. The mechanical torque
the free-running speed. generated by it is given by
Suppose that a constant opposing torque of load
and friction equals 0.06 Nm and the torque pro- (29)
duced by the motor under largest phase error is
0.5 Nm. The normalized load and filter are where ¢ is the phase difference between the desired
position and actual shaft position. This result can be
50 derived from step motor modds given in [7-11].
T =0.12 and F(s)=--. The dynamic equation of the stepper is given by
sr + 1

The filter functions in this case are (30)

F(0)=50 where J is the moment of inertia, D is the viscous

friction, and TL is the constant friction and load
and torque.
The equivalent model of (29) and (30) which fits
the format of Fig. 2 is

(31)
The resulting lock range is found from equation (12)
to be and

wL =50 radjsec (32)

and the input frequency is limited to
Since step motors' speed cannot be biased, the free-
-56<w 1 <44. running frequency, w 0 , is zero.
In order to illustrate the results, consider a motor
The absolute speed range of the motor is found by with the parameters
adding w 1 to w 0 . This results in the range between
Tm=0.3Nm
144 and 244 radjsec.
The pull-infrequency range is found from (27). TL=0.06Nm
J =4 X 10- 4 kg X m 2
D = 2.5 x 10- 3 Nm/(rad/sec)

For the given parameters, the minimum is found to The lock range is found in this case to be
be wp = 42.5 at Q = 30. Thus, the input frequency is
limited by Tm
wL =n= 120rad/sec.

-48.5 < w1 < 36.5 rad/sec

However, due to the constant load, the actual
and the absolute pull-in speed range is between velocity is further limited, according to (12).
151.5 and 236.5 radjsec.
w 1 < 96 radjsec.
The physical interpretation of this result is that as
long as the motor is required to run within the pull-
The pull-in range is derived from equation (27).
in range, it will acquire that velocity from any initial
Here we have
condition. However, if a velocity of 240 radjsec is
required, the acquisition depends on the initial Tm
conditions and a gradual speed increase is recom- F(0)=-=120
D
mended.
In order to verify the results on pull-in range, a and
simulation of this system was performed, resulting
in a lock range identical to the one predicted and a 120
R(Q)
pull-in range between 155 and 233 rad/sec.
15R JACOB TAL

The corresponding pull-in frequency is given by [11] C. K. TAFT and R. C. GAUTHIER: Stepping motor failure
model. IEEE Trans. IECI IECI-22, August (1975).
[12] G. SINGH: Mathematical modeling of step motors. B. C.

wP= min [
!1>0
7200
Q+ Q(l +0.010 2 )
J .
Kuo, Ed., Symposium Proceedings: Incremental Motion
Control Systems and Devtces. University of Illinois, March
(1972).
[13] A.GELBand W. E. VANDERVELDE: Multiple-Input Describ-
ing Functions and Non linear System Design. McGraw HilL
This is equal to wP = 50.2 at Q = 37. The actual pull- New York (1968).
in velocity due to the constant load is found from
(27) APPENDIX A
DIDF FOR A SINUSOIDAL DETECTOR
w 1 < 26.2 radjsec. Given a phase detector with a sinusoidal nonlinearity,

g(</>)=sin </> (AI)

Thus, the pull-in velocity is 26.2 and the slew, or lock
range, is 96 radjsec. This means that a gradual and an input function of the form

acceleration is required for velocities between the <P=nt +~:sin (yt+l/1 ). (A2l
two boundaries.
Computer simulation of this system resulted in a The objective is to determine the de and the fundamental
harmonic components of the detector output x =sin</>. The
pull-in velocity of wP = 48 radjsec, and the actual output signal is given by
velocity pull-in range is w 1 < 24 radjsec. Again,
good agreement with analytical results is found. x =sin [nt +~:sin (yt + 1/J)]
=sin nt cos [e sin (yt + 1/1)]
6. CONCLUSIONS +cos nt sin[~: sin (yt + 1/J )]. (A3)
The conditions for the synchronization of a
controllable oscillator with some external frequency Since 1; is the response of the low-pass filter, F(s ). and the
integrator to the sinusoidal component of x, we assume that
were determined. The first requirement is the
existence of a stable equilibrium point. This can be <.~21! (A4l
determined easily, as a linearized model may be
and we apply the standard approximations to the sin and cos
used. The second requirement is that the system terms in (A3).
response must converge to the equilibrium point.
The analysis of this condition is much more difficult, X =Sin nt +cosntesin (yt + 1/J ). (A5)

due to the nonlinear nature of the system, and Equating (A5) to the expression
therefore it is solved indirectly by considering the
limit cycles which result from the loss of synchroni- x =a sin (yt+b)+ b (A6)

zation. The main contribution of this paper is the we observe that

analytical conditions for synchronization, which
can be applied to different systems, as illustrated by (A7)

the examples. a=1 (A8)

(A9)
REFERENCES
Furthermore, in view of (A 7), the de ten~ of (A5) equals
[1] A. J. VITERBI: Principles of Coherent Communication.
McGraw-Hill, New York (1966).
b=issinl/J.
[2] W. C. LINDSEY: Synchronization Systems in Com-
munication and Control. Prentice Hall, New York (1972).
[3] J. TAL: On the capture range of phase-locked loops. IEEE APPENDIX B
Trans. Communications COM-23, March (1975). THE EFFECT OF HARMONICS OF g(</>)
[4] L. J. GREENSTEIN: Phase-locked loop pull-in frequency. The greatest accuracy in the derivation is required when the
IEEE Trans. Communications COM-22, (8) p. 100 (1974). output frequency is neither converging nor diverging. In that case
[5] DC Motors, Speed Controls, Servo Systems-an Engineer- </>(t) has the form
ing Handbook. Electro-Craft Corporation, Hopkins, Min-
nesota (1975). e~2rr. (81)
[6] J. TAL: Phase-locked servo-systems-a new approach to
speed control. Proc. Fourth Annual Symposium on In- Let
cremental Motion Control Systems and Devices. University
~
of Illinois, April (1975).
[7] D. H. SMITHGALL: A phase-locked loop motor control g(</>)=Asin<f>+ L B.sink<f>. (82)
k=2
system. IEEE Trans. IECI IECI-22, November (1975).
[8] D. J. ROBINSON and C. K. TAFT: Dynamic analysis of The resulting phase comparator output is
magnetic stepping motors. IEEE Trans. IECI IECI-16,
111-125, September (1969). u=g(</>)=Asin<f>
[9] K. VENKATARATNAM, S. C. SARKER, and S. PALANI:
Synchronizing characteristics of a step motor. IEEE Trans.
Aut. Control AC-14, October (1969). + L B.sin[knr+k~:sin(nt+l/1)]
k=2
[10] K. VENKATARATNAM and M. C. MOULI: Stability of a
stepping motor. Proc.IEE 118, (6) June (1971). =Asin<f>(t)+Stft)+S 2 (t) (83)
 Synchronization characteristics of controllable oscillators 159

where
S2(t)= I Bkcosk!lt[2Jdke)sin(!lt+r/t)
k=2

Sdt)= I Bksin(k!lt)cos[kesin(!lt+r/t)] (B4) + 2J 3 (ke) sin (3!lt + 3rjt + ... ] (B9)

k= 2

and Note that the contribution to the bias term of u comes only from
S1 , whereas S2 contributes to the fundamental harmonic. Our
objective is to show that these contributions are negligible,
s2 = I Bk cos (k!lt) sin [ke sin (!lt +rjt)]. (B5) compared with those of the first term, A sin l/J, which are
k= 1

bias: u0 =!e sin rjt }

Since (BIO)
fundamental: u 1 =A sin !lt
cos (/J sin 8) = J 0 (/3) Note that J 0 (ke) does not produce a bias term, and the
+ 2J 2 ({.J)cos 28 + 2J 4 ({.J)cos 48 + ... (B6) contributions to the bias are proportional to J 2 (ke), J 4 (ke), efc.
Similarly, the contribution to the fundamental harmonic are
and proportional to J 1 (ke) etc.
Noting the approximation:
sin ({.J sin 8) = 2J 2 (fJ) sin 8 + 2J 3({.J) sin 38 + ... (B7)

s1 and s2 become
I -
J.(fJ)""--
n! 2
(P)" . (Bll)

It can be seen that the bias additions are proportional to e2 or

Sdt)= I Bksink!lt[J 0 (ke) higher power. The fundamental harmonic terms are proportional
k=2
to e1 or higher power. In any case, they are very small compared
+ 2k 2 (kecos (2!lt+ 2rjt)+ ...] (88) with terms of(BlO).