THE ANALYTICAL DYNAMICS OF
THE WOODPECKER PROBLEM
By
RUSSELL A. ROY
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1974
ACKNOWLEDGMENTS
The author wishes to thank many people who have helped
him over the years. His parents provided help and support in
spite of many setbacks. Several of his friends helped him
considerably during his work in the Physics Department on
Project Sisyphus.
Special thanks are extended to the chairman of the
author's committee, Dr. Knox Millsaps, for his suggestion of
a topic and his encouragement in completing the work. The
other members of his committee also offered many helpful
suggestions.
Appreciation is extended to many people at the Orlando
Division of the Martin Marietta Corporation, especially Max
Farrow, Avery Owen, and Al Roy, who were generous with time
and equipment.
The author's wife, Eileen, has been unfailingly patient
and supportive for many years, so much so that it would have
been impossible to complete the work without her.
Finally, it is necessary to thank Terence (185159 B.C.)
who said:
Ita vita est hominum, quasi cum ludas tesseris;
si illud quod maxime opus est jactu non cadit, illud,
quod cecidit forte, id arte ut corrigas.
TABLE OF CONTENTS
ACKNOWLEDGMENTS .
LIST OF SYMBOLS .
ABSTRACT . . .
CHAPTER
I INTRODUCTION
II THEORY . .
III EXPERIMENTAL
IV COMMENTS .
APPENDIX A ..
APPENDIX B . .
APPENDIX C . .
LIST OF REFERENCES
BIOGRAPHICAL SKETCH
Page
ii
iv
vi
RESULTS
. . .
. . .
. . .
. . .
. . .
iii
LIST OF SYMBOLS
MB mass of the bird, grams
MSL mass of the sleeve, grams
M mass of the spring, grams
a dimensionless factor describing an effective
spring mass
k spring constant, dynes/cm.
F frictional force opposing the motion of the spring,
0 dynes
1 frequency, radians/sec.
4a arithmetic decrement of the spring per cycle, cm.
xI displacement of the bird mass with respect to the
sleeve, cm.
I represents an idealized impulse delivered to the
bird mass twice each cycle, dynes
x the distance of the bird mass from equilibrium,
o caused by the rotation of the sleeve on the shaft,
also where the impulses occur (xI = x ), cm.
T dimensionless variable related to time (T = wt)
x dimensionless variable related to displacement
(x = x1/x0)
f dimensionless variable related to friction
S (fo = F /kx )
I dimensionless variable related to the impulses
S (I = I/kx )
O O
b dimensionless variable related to those positions
o of the bird mass with respect to the sleeve where
it locks and unlocks and where the impulses take
Dlace.
LIST OF SYMBOLS Continued
R radius of circular arc in phase space
v dimensionless speed in phase space
2
h dimensionless energy associated with energy given
to the bird mass during an impulse
A the real amplitude of the motion of the bird mass, cm.
y velocity of bird mass in phase space
Ad the amount the sleeve slips down the shaft each
half cycle as a result of the unlocking of the
sleeve, cm.
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
THE ANALYTICAL DYNAMICS OF
THE WOODPECKER PROBLEM
By
Russell A. Roy
December, 1974
Chairman: Knox A. Millsaps
Major Department: Aerospace Engineering
A toy, usually in the form of a small woodpecker, is
shown to be a problem in the area of selfexcited oscilla
tions. An analysis of the problem by J. P. Den Hartog is
discussed and shown to be without solution. An alternative
theory is proposed and its results compared with those of
experiment.
CHAPTER I
INTRODUCTION
In the celebrated course at M.I.T. on nonlinear mechanics
J. P. Den Hartog demonstrates a toy "woodpecker" to his class.
The toy can be idealized as shown in Figure I. The solution
is given as a problem on the course's final exam and Den Har
tog's analysis is found in the Study Guide [1] for the course.
The problem and his analysis are included in all pertinent de
tail in Appendix A. His analysis concludes with the equation
fA() = 2 + B 2 62 = 0 (11)
where
a = A2(cosh Apl) (12)
B = A(A(sinhA) (13)
y = (A2+2)(cosh Apl) A( sinh A( (14)
6 = A sinh AM + A2 cosh A( (15)
with A and ) defined in Appendix A. The solution of (11)
would mean the solution to the analysis, but this has not been
done until this point. In fact, Den Hartog remarks [1] that,
"This I have not done, and if anyone ever does it, I shall be
grateful if he or she will send me the solution."
Letting
x = cosh A(
(a) (b)
Figure I
3
and
y = sinh A(
and substituting these into (12), (13), (14), and (15),
and thence into (11) we have, after canceling and gathering
terms:
fA() = A4 2 A4 2 x 4A3 y + 4A3 xy
+ A2(4x +8x4 y 2) + A(40xy40y) 4x2 + 8x 4.
The righthand side can be factored and then both sides
divided by x1, giving:
S( A 2 22
Ax A42(x1) + 4A3 y A2[4(x1) + x ]
x1 x1
+ 4A(y 4(x1).
This expression can be regrouped and divided by 4:
_fA A44 2 3
4x1) = x(  4+ A2+1) + y(A 3 +Ao)
A4~2 2 2
 (1 + A2) + A + 1.
(x1)A
2 2 2
Noting that x y = 1 or y = (x+l)(xl) and dividing by
A2+1 we get
fA() A4 2 x+1
fA) A [x+1 + 2 ] + A(y (x1).
4(x1)(A +1) 4(A +1) A
Rearranging this results in
fA() A 2 6)
=  (x+1) + Acy (x1) fA((). (16)
4(x1)(A +1)
Clearly fA( ) and fA(j) have the same nontrivial roots
so we need only examine f*(j), the interesting factor of
fA(]). By letting z = Aq, fA*() can be written as
fA(q) = z sinh z ( 2(cosh z+l) (cosh z1). (17)
Equation (17) can be explored by expanding sinhz and cosh z
in their familiar power series,
3 5 2n+1
z z z
sinh z = z + + + + + + z
3! 5! (2n+1)! +
2 4 2n
cosh z = 1 + + + + +
2! 4! 2n!
and then substituting them into (17), obtaining
f2 2 z2 z4 z4 z4
fA ) = (z 2 + 42! 4
6 6 6
z z z
+ T5! T 60 +
2n+2 2n+2 2(n+l)
Z z z
(2n+l) 4(2n!) 2(n+l) ) +
The first two groupings are equal to zero and the third
grouping is negative for all values of z except for z = 0,
th
when it is zero. The n grouping can be written as
2n+2
z 1 1 1
2n! 2n+ 4 (2n+2) (2n+l)]
and will be negative for all n > 2 for all nonzero z. Thus
f*(O) is negative for all z with the exception of the physic
ally trivial case of z = 0. One must conclude that the Den
Hartog analysis is not satisfactory and the basic physical
model must be reexamined.
Several models of the toy woodpecker were built and
examined. The first models built were like that shown in
Figure Ia. These were not stable and could not be con
structed in such a way as to produce periodic motion. While
this may not be impossible, it seems quite difficult, with
several parameters being critical. After several hundred
variations were attempted unsuccessfully, this model was
abandoned, and one like that in Figure Ib was built. It was
made using a spring from a ball point pen, a piece of broken
capacitor, a scrap of wooden dowel, and a shaft of aluminum
rod, and delighted the author by working perfectly the first
time it was tried. By observing the motion for various con
figurations several things became apparent. To make sure of
what was happening, the motion was videotaped and then the
tape was played one "frame" at a time, allowing intervals of
onesixtieth of a second to be observed. The following
features of the motion were clear:
1. The "bird" mass oscillates up and down, above and
below the sleeve. When this mass is above or below the
sleeve by a certain amount, the sleeve is locked on the shaft
due to friction. In between these points the sleeve is un
locked and drops a short distance.
2. "Hard" excitation is necessary to start the motion.
That is, a certain threshold energy input was necessary to
start the characteristic slipstick motion.
3. The motion is periodic with no change in frequency
from start to finish.
4. The sleeve locks in both possible orientations.
5. The frequency of the system is not the same as the
natural frequency of the bird mass on the spring with the
sleeve locked, but somewhat less.
6. When the sleeve is locked and the natural motion
observed the decay in the amplitude is not geometric but
appears to be arithmetic.
7. The sleeve moves a short distance down (about 0.15
centimeter) twice each cycle.
8. Running the woodpecker in a vacuum rather than in a
normal atmosphere resulted in no change in frequency.
9. The steady state motion is independent of the initial
conditions.
10. None of the physical parameters appear to be critical.
11. When the system is started with a very large initial
displacement the sleeve is observed to lock and unlock, but
not to drop for several cycles.
The above observations indicate that this problem falls
in the area of selfexcited oscillations. More specifically,
it falls in the area of mechanical clocks. The theory
describing mechanical clocks has been worked out, most
notably by A. A. Andronov and coworkers [2,3] although other
people [4,5] have discussed it as well. As Andronov points
out [2, p. 168], "A clock is an oscillating system which
maintains oscillations whose amplitude is independent of the
initial conditions. To start the clock a large initial im
pulse is usually needed. If the initial impulse is too
small, then the clock comes to rest again." He also notes
[2, p. 168] that
Any clock mechanism can be broadly divided into
three parts: (1) an oscillating system, for exam
ple, a pendulum, a balancewheel, etc., (2) a source
of energy such as a weight or spring, (3) a trigger
mechanism, connecting the oscillating structure
with the energy source. For fixed positions of the
oscillating system (referred to, for the sake of
brevity, as the pendulum) the trigger mechanism
acts and energy is given to the pendulum during a
short impulse. In a good clock the impulse is of
very short duration. The trigger mechanism usually
acts twice during a period close to the position
of equilibrium where the velocity is greatest. It
is important to note that the instant of time when
the trigger mechanism begins to act is entirely
determined by the position of the pendulum. In
addition, the manner of its action and the magni
tude of the impulse also depend on the state of
the pendulum. Consequently, all forces which
arise in the mechanism depend only on the positions
and velocities of the separate parts of the system,
and not on the time. Thus the clock is an
autonomous system.
There can be no question that the toy woodpecker is in
fact a mechanical clock, and can be analyzed with the aid of
clock theory. In Chapter II, the theory of the woodpecker,
now revealed as a cuckoo, will be discussed, and in Chapter
III a comparison of the theory with some experimental results
will be given. Chapter IV includes a summary of this work
and its conclusions.
CHAPTER II
THEORY
Selfexcited oscillations form an important class of
phenomena including some electronic oscillators, buzzers,
electric bells, wind and string musical instruments, brake
"squeal," some kinds of aerodynamic flutter, and so forth.
These phenomena are characterized by definite oscillations
which are properties of a specific system and not dependent
on the initial conditions. One of the most interesting
aspects of these systems is that they can generate a periodic
process from a nonperiodic source.
Clocks are a distinct class of mechanical selfexcited
oscillations, and consist of three parts: (1) an oscillating
system, (2) an energy source, and (3) a trigger mechanism or
escapement connecting the two. Clock theory has been worked
out [25] and, although the woodpecker is different in
several ways from the clocks discussed in the literature, the
theory can be suitably extended to describe it.
The physical features of the motion mentioned in Chapter
I should now be more carefully examined in the context of
developing the most descriptive mathematical model. Hard
excitation is necessary to start the motion. Soft excitation
is defined as when an arbitrarily small initial displacement
is needed to start the motion and hard excitation as when an
8
initial displacement (or other appropriate energy input) must
be larger than a certain amount to start the motion. The
hard excitation observed implies that the friction present
is either of the Coulomb type (dry friction), or dominated
by Coulomb friction. A summary of an argument by Andronov
indicating that Coulomb friction (compared to linear or
velocity dependent friction) is the more appropriate type for
clocks is given in Appendix B.
The actual motion of the system is not complex and can
be analyzed in a series of several steps. Beginning with the
bird mass in the maximum upward position with respect to the
sleeve, and with the sleeve locked the features of the motion
are as follows:
1. The bird begins to descend and continues until it is
a short distance above the horizontal.
2. At this point the sleeve unlocks and the whole sys
tem begins to descend.
3. The bird continues to descend with respect to the
sleeve and when the bird is a short distance below the hori
zontal the sleeve locks. At this point from the frame of
reference of the bird it would appear that its velocity with
respect to the sleeve suddenly increases. The idealization
of this is an impulse applied to the bird at the instant the
sleeve locks.
4. The bird continues to its downward extreme and then
starts back up.
5. When the bird is just below the horizontal the sleeve
unlocks and as the bird continues up with respect to the
sleeve, the system again begins to descend.
6. The bird continues up with respect to the sleeve
and when it is just above the horizontal the sleeve locks
and again this can be idealized as an impulse applied to the
bird.
7. The bird continues to its upward maximum and the
cycle begins again.
The general form of the differential equation describ
ing the motion can now be written as
MX1 + kxI = +F + I (21)
where each of the terms will be discussed below.
The mass, M, is equal to MB + aMs for that part of the
motion when the sleeve is locked. The mass of the bird is
MB and Ms is the mass of the spring. The quantity a is a
correction factor used to introduce an effective spring mass.
It would be measured by locking the sleeve and then measuring
the frequency of the bird for various values of MB. These
values would then be fitted by the curve
f + kM (22)
27 M S I
using least squares to get the best value for a.
During that part of the motion when the system is falling
M would be given by
(MB + aMS)MSL
M B + S + MSL (23)
B S SL
where MSL is the mass of the sleeve. For this situation the
natural frequency would be
2 k
k = = (24)
with M given by Equation (23).
The spring constant, k, has the usual definition and
would be measured by the normal static means. The Hookes
Law approximation is made and must be substantiated by exper
imental results. Although the present theory is based on
phase plane analysis with linear equations pieced together
on the phase plane, it would be possible to extend it to in
clude a nonlinear term such as kx if it were necessary. To
handle such a term an approach such as the "phaseplane delta"
method [6, p. 244] would have to be used.
The term Fo is a quantity describing the frictional
losses within the spring. The sign is chosen as opposite to
that of xl. It would be measured by locking the sleeve and
measuring the arithmetic decrement of the amplitude per cycle,
4a, and noting [2, p. 153] that
F = amw2 = ak. (25)
0
As far as the impulse term, I, is concerned, one of two
assumptions is usually made in clock theory: that either
(1) momentum or (2) energy is conserved during the impulse.
Since in the woodpecker a uniform amount of gravitational
potential energy enters the system during each cycle, the
second assumption seems to be the more reasonable one. It
is also the one made more frequently in clock theory [6,
p. 168].
There is a slight amount of rotational motion present
due to the sleeve locking in an "up" and a "down" position.
It is possible to consider a treatment with two variables,
one linear and one angular, to describe this motion. However,
because the angular motion is relatively slight, it is pos
sible to do an analysis with one variable by approximating
the effect of the rotational motion. To do this, the motion
is divided into three regions:
MX1 + kx1 = +F + I for x0 < x1 <_ X (26)
MXk + kx' = +F x' = x + x for x > x (27)
1 1 0 1 1 0 1 0
M3' + kx' = +F, x' = x x for x1 < x0 (28)
Where the impulses are pictured as occurring at xl = x0, and
at this value of xl, the position of the bird is changed by
x0, an amount which corrects xl for the slight rotational
motion of the sleeve.
The equations (26), (27), and (28) can be simplified
by a change of variable. First the equations are divided by
kx0, and then let
k 1/2
and
T = wt (29)
13
so that
M x1
x x (210)
0
where now we have a dimensionless time, T. Also let
f (211)
and
x0
b0 0 1.00, (212)
and
10 k (213)
so that now (26), (27), and (28) take the form
S+ x = f0 + I0 for 1 < x < +1 (214)
R + x = f0 + b0 for x > 1 (215)
x + x = +f0 b0 for x < 1 (216)
This motion can be shown with a phase plane diagram. The
velocity, y, of the bird with respect to the sleeve is repre
sented by the vertical axis with upwards (positive) y corre
sponding to downward motion of the bird. The x axis repre
sents the displacement of the bird with respect to the sleeve
which is also positive for downward motion of the bird.
There are two impulses per cycle which in this model are con
sidered to be of equal magnitude, occur symmetrically in each
halfcycle and take no time. The phase diagram is shown in
Figure II. The convention used in clock theory is to follow
Figure II
,v I b',&
the phase path in a clockwise sense. At point S the bird is
at its maximum negative (upwards) displacement and its
velocity is zero. The bird then begins to speed up as it
falls. At x equal to minus one the sleeve unlocks and the
system falls, the bird going slower with respect to the
sleeve than before. At x equal to plus one the sleeve locks,
an impulse is delivered to the bird and it continues to its
positive (downward) extremum at f. This process repeats it
self in the other direction ending at S. The phase paths in
the various parts ot the motion are portions of circles
centered at (0,f0) or (0,f0+b0). This can be seen by
letting y = k; then equation (214), for example, can be
written as
y + x = f0 (x i 1)
or
y 1+ x = f
Rearranging, we have
ydy = (f0x)dx
1 R2 2
and letting (R f0) be a constant of integration, we get
after integration
2 2 2 2
y = 2f0x x + R f0
or
y2 + (xf)2= R2 (217)
which is a circle in the phase plane centered at (0,f0).
For periodic motion we require that the speed at point A
be the same as that at point B:
vA = B (218)
This condition will allow the frequency and amplitude of the
motion to be calculated. As shown in Figure III, we have
2 2 2
R2 = 2 + (1b +f 2 (219)
1 A o (219)
R + 2fo = R (220)
and
R2 (1f 2 = R (b +f 1)2 (221)
3 o 2 o o
Also, we see that
(v)2 + (f+1)2 = R (222)
and
(vA)2 + h = v2 (223)
where
A(y2) = h2 (224)
is the change in the square of the speed in dimensionless
units associated with each impulse. The equations above must
be solved to obtain a relation between vA and vB. Substituting
first (222) and then (221) into (223) we obtain
2 = h2 (l+f)2 R2 + (1f) (b +f 1)2
B o 2 o o o
and then substituting (220) into this, we have
Figure III
2 2 2 _2 2
Sh (1+f ) + (f ) ( (b+fl) + R 4fR + 4f1.
B o o o o 1 o
2
Using (219) for R1 one gets
2 = V + (+b+f )2 + h (+f)2 + (1f)2
B A 0 0 0 0
(b +f )2 + 4(b +f ) 4(b +f )R .
Setting
2 2 2
VB = VA = v
and canceling, we get
4f R = 4f2 + h2 4b f (225)
ol 0 0 0oo
Dividing by 4fo, squaring both sides, and again using (220)
results in
7 h2 b h 4
2 2 2 b0 h 2 ho
v + (1+b +f ) = f 2f b + + + + .
o o o oo 2 o 2f 1f
S 16f
Rearranging terms gives
S 4 2 b h2
v= + 2 o + 2b 1 2f (226)
2 2 2T_ o o
16f o
It is still necessary to determine h to be able to cal
culate v. By noting that the system drops a specific dis
tance, Ad, once each half cycle,the energy input during the
impulse can be calculated by assuming it is due to this
change in gravitational potential energy:
g(Ad) = (A2 (227)
glad2
where
2 2 2
(v)2 (x1)2 x) 1
is the change in the square of the physical speed as a result
of the impulse. This must still be related to the dimension
2
less quantity h2. By definition
2 d x 1 2 d x1 2
h = T () 2 UT x1
and using equation (29) we have
h2 _1 d X 2 1 d x1 2
O O
Then substituting in the definition of w, we get
2 M 2 *2
h = 2 [(x2? (x1)]
kx
or
2 M 2
h 2  [(Av) ]
kx
and using (226), we finally get
2 M
h = [2g(Ad)]. (228)
kx
o
The mass, M, in this equation is MB + aMs and does not in
clude the sleeve mass since the energy the sleeve picks up
will be lost when it is stopped by the friction between it
and the shaft.
Now that h has been obtained it is possible to solve for
v and then to solve for R1, R2, and R3 using the equations
developed above. The frequency of the motion can be calcu
lated by noting "that the representative points move along
the phase paths with an angular velocity equal to unity"
[2, p. 197]. Thus the time of transit in T units along one
of the semicircular arcs is equal to the angle subtended at
its center by the arc. As shown in Figure III, the relevant
angles are a, 8, 6, and y, and the period in T units would
be given by
T = 2(a + B + 6 + y + ). (229)
By inspection of Figure III the angles above can be easily
obtained in terms of known quantities:
v (230)
tan a = lb +f
O O
b +f 1
sin B o (231)
2
1f
sin 6 o (232)
3
l+f
sin y (233)
In physical units the period would be
M+ M MMSL 1
T 2(a++ ( B s 1/2 2(61)(k(ML /2
T = 2(++ )( ) + 2(6+y)(k(M+MSL)")
and the frequency would be 1/T.
The energy balance is an important aspect of this motion.
For regular, periodic motion the energy input per cycle must
equal the losses per cycle. The energy input for the bird
and spring is on the order of 2(MB+ Ms)g(Ad) per cycle and
the energy losses per cycle are on the order of 4f A or 4akA
o
where A is the amplitude of the motion measured from the
equilibrium point. Equating these two energies and solving
for the amplitude, we have
A Mg(Ad)3
2ak (234)
To the extent that the energy balance is correct, this rela
tion should predict the amplitude. This balance is basic to
the theory and (233) can be derived another way by noting
that the normalized amplitude, Ao, where
A = A/x ,
can be derived from the phase diagram, Figure III, as
A R + b f
o 1 o o
Using (225) for R1 we have
A h2
A T
o 0
Substituting (25), (211), and (228) into this results in
A Mg(Ad)
x 2akx
O O
which is the same as (234).
22
Before turning to an experimental exploration of this
theory, some general points about it can be made. The theory
predicts that the frequency will be less than the natural
frequency of the bird on the spring with the sleeve locked.
We can understand why this is so physically because the peak
to peak amplitude of the woodpecker will be 4b0 longer than
the amplitude of the natural system, other things being equal,
and this will take longer.
We also note the necessity for at least one dynamical
measurement, Ad being the most appropriate. In this it is
like the Den Hartog theory where f, the coefficient of fric
tion between the sleeve and the shaft, is idealized as having
the same value for motion as for rest. In fact, this would
be unlikely and some sort of dynamic measurement would be
necessary to establish some appropriate average value for f.
In the present theory it is possible to get around this
necessity for a dynamic measurement, at least theoretically,
although some approximations will be necessary. First
assume that when the sleeve unlocks it falls freely with an
acceleration of g. Since it starts from rest we have
Ad = gt2 (235)
where t will correspond to that part of the period when the
system is unlocked:
S= (6+ M MSL 1/2
t = +y) k(M+M (236)
Now
l+f
1 o
6 = sin
R3
and
I 1fo
y = sin R
R3
and these can be expanded in a power series:
3 5
sin x =x x+ .
3! 5!
By taking only the first two terms of the series (this and
other approximations remain to be justified in Chapter III)
we have
1fo l+fo (1fo) 3 (+f) 3
S+y R + R 3 3 
3 3 6R3 6R
3 3
and
2+6f2
2 o
6 + Y = R 3 
3 6R3
3
It is further assumed that f is small and terms of the order
2
f and higher can be neglected. So we have
0
6 + y = 2 (237)
3 3R
By substituting (220) and (219) and then (226) into (221)
we obtain,after some algebra
2 8f 16f 16f 32f 1/2
R3 = 0o + + (238)
o h o h h h
2
Again, terms of the order f and above are neglected, result
ing in
2 8f /
h o 1/2
R3 h 1 2)
o h
(239)
Expanding this in a series and keeping only the first two
terms gives approximately
2 4f
R3 = 4T (1 2)
o h
or
h2
R3 =
Substituting this into (237) yields
8f f
So 16 o
h 4f (h 24f )
O O
The second term is neglected and substituting this for
6 + y in (236) gives
Ad g SL o 2
2 k(M+Ms) 11 2 'f
Letting
L M M S 1/2
2 k(M+MSL)
and
2 B(Ad)
h = B(Ad)
(240)
(241)
where
B (2gM
kx
and putting these into (241) gives
64Lf2
2 o
(BAd4fo) = (242)
This relation can be solved numerically to give a Ad for a
particular MB. In Chapter III this equation will be explored
with specific data to determine its validity.
Although only the steady state motion has been explored
here it would not be difficult to extend the phase plane
analysis to cover various types of transients present when
the motion begins. Several references (see, for example
[6, p. 274,196], [2, p.178], and [3, p. 13]) discuss various
types of transient analysis.
CHAPTER III
EXPERIMENTAL RESULTS
The present work was originally intended to be entirely
theoretical in nature, consisting of completing the analysis
begun by Den Hartog and then exploring various physical situ
ations with the completed theory. Because it was necessary to
redo the theory, it was felt to be prudent to include experi
mental results with the new theory so that a complete and con
sistent picture would be present for the first time.
Because the original thrust of this work was not experi
mental, the measurements were made in most cases on the basis
of rather limited resources. Fortunately, most of the physi
cal quantities required are not difficult to measure and the
results form a consistent pattern which unquestionably sub
stantiates the theory.
The frequencies of the woodpecker model for various bird
masses were measured first. They were measured by counting
forty or fifty cycles and timing them with a stopwatch. Many
runs were made with the same bird mass with different numbers
of cycles (for example, 10, 20, 40 and 50 cycles) to see if
the woodpecker slowed down or speeded up as it moved down the
shaft. No such effect was observed. These measurements as
well as many others were also made through the use of a video
tape arrangement made available for this work from the Orlando
Division of the MartinMarietta Corporation. The woodpecker
was taped in front of a numbered centimenter grid with a
digital clock reading in hundredths of a second also in the
picture, because the video tape could be played back and
"framed" to show a still picture of the motion every sixtieth
of a second. Appendix C contains some examples of data col
lected from the video tape.
Table I shows the average frequency, fe, obtained for
various bird masses, MB. The masses were measured to an
accuracy of about 0.05 gram. The individual frequency measure
ments were only precise to about 0.1 Hertz, but because large
numbers of measurements were made the error estimate was cut
in half.
Table I
M + 0.05 (grams) f 0.05 (Hertz)
0.00 4.32
10.15 2.95
11.50 2.83
12.00 2.77
13.50 2.70
16.20 2.59
18.10 2.49
20.70 2.42
21.20 2.32
24.55 2.21
The spring constant, k, was measured in the usual way,
measuring deflection for various loads. There was some scat
ter and a slight indication of spring "hardening" at large
deflections. Table II gives the values of the loads in grams
and the respective deflections.
Load 0.05
(grams)
2.00
3.00
4.00
5.00
5.93
6.00
7.00
8.00
8.43
9.00
9.68
10.00
10.15
10.90
11.00
11.37
12.19
13.00
15.00
17.71
18.96
20.18
21.45
Table II
Deflection 0.05
(centimeters)
0.40
0.60
0.75
0.85
1.10
1.10
1.25
1.45
1.50
1.50
1.68
1.70
1.90
1.87
1.75
2.10
2.04
2.10
2.30
3.01
3.22
3.44
3.54
The average k for these values is 5.738 grams per centimeter
or 5,623 dynes per centimeter. A graph of the loading versus
deflection is shown in Figure IV.
The effective spring mass coefficient, a, could now be
determined using the value for k obtained above. This quan
tity was determined by locking the sleeve in place and measur
ing the frequency, f, for various MB. Then a was determined
from the equation
c)
14
I 
7 2
II
J
7
4
5 
I I
I 2
DEFLECTION IN CM.
Figure IV
f 1 ( k 1/2
2 77 IMB +aMS
or, solving for a:
k B
a = (31)
4 2 2M MS
The mass of the spring, MS, is 5.1 grams and the following
values for a were obtained:
MB + 0.05 f + 0.01 a
(grams) (Hertz)
9.70 3.19 0.84
11.30 3.03 0.83
12.75 2.86 0.91
17.75 2.54 0.85
18.99 2.46 0.90
The average value of a is 0.87 and this was the value used.
The quantity Ad was measured for each of a variety of
bird masses by looking at the total distance the system
dropped for a certain number of cycles and then averaging to
find the distance per half cycle. By inspection of the video
tape it was determined that the drops were uniform although
the drops when the bird mass was going down were slightly
different from those when it was going up. Table III summar
izes the values of Ad for the respective bird masses.
The most difficult quantity to measure accurately was 4a,
the arithmetic decrement. This quantity was measured by lock
ing the sleeve in place, starting the bird mass oscillating,
and then measuring the amplitude from the equilibrium point
Table III
MB 0.05
(grams)
0.00
10.15
11.50
12.00
13.50
16.20
18.10
20.70
21.20
24.55
Ad 0.005
(centimeters)
0.231
0.182
0.175
0.173
0.166
0.152
0.143
0.130
0.128
0.112
for each half cycle. During these measurements the bird mass
would begin processing which made the vertical amplitude
appear to decay more quickly on the video tape than it actu
ally was. To avoid as much error as possible, only the first
five or six extrema were considered. For a bird mass of 12.75
grams, for example, the following values were observed:
Run I
Amplitude
(centimeters)
5.179
5.083
4.986
4.839
4.451
4.562
4.104
3.939
Run II
6.12
5.93
5.98
5.83
5.75
5.32
5.59
2a
(centimeters)
0.096
0.097
0.147
0.388
+0.111
0.458
0.165
0.19
+0.05
0.15
0.08
0.43
+0.29
Extrema
up
down
up
down
up
down
up
down
down
up
down
up
down
up
down
Of course, many more runs were made and analyzed but these
data are shown as examples of the data obtained, showing the
experimental scatter and the precessional effects. It was
not possible to make a defensible statement with regard to
the possible variation of a with MB. Since it was not
desired to fit a to theoretical data, a value of 0.044 centi
meters was assigned to all values of MB.
It is possible that the damping could be linear (propor
tional to velocity in some way) rather than constant. The
woodpecker system was run in a partial vacuum (<0.1 atm) to
see if there was any velocity dependent damping due to air
resistance. No effect was observed and the system ran exactly
the same (frequency and amplitude) as it did in a normal
environment. It is still possible that the losses in the
spring could be due to linear friction and this should be
explored.
In the case of linear friction the ratio of two succes
sive maxima would be given [6, p. 201] by
n+1l 2Tv
 e (32)
x
n
or
v Ln n (33)
27r n+l
where v is called the damping ratio, and where the damping is
small. In the case of constant friction we have
xn xn+/2 = 2a = constant.
(34)
Substituting some experimental values into (33) we have
1 5.18
v Ln .
27T 4.99
or v = 0.0059 to the correct order of magnitude. With this
value of v, e2 7is 0.964. If we expand e2~v in a power
series, consider only the first two terms, and substitute
in the argument we have
e 27 = 10.037 = 0.963.
Hence, to a good approximation
n+l
l v (35)
n
an d
xn xn+ = vx (36)
for a full cycle and
x x+/2 n (37)
for a half cycle. In the woodpecker we note that at the end
of each half cycle the impulse restores the amplitude. In
fact, it is this exact relationship of energy balance which
gives the woodpecker its clocklike behavior. We note the
resemblance of (37) and (34) for small v and the wood
pecker's physical situation and conclude that whatever linear
friction is present, if any, can be treated as constant to a
good approximation.
The final experimental quantity necessary to be measured
is x the displacement of the bird mass with respect to the
34
horizontal at which the sleeve locks and unlocks and at which
the impulses take place. This was measured statically by
rotating the sleeve up and down to its two locked positions
without letting the sleeve move vertically and measuring the
resulting displacement of the bird mass. These measurements
give an x0 of 0.75 0.01 centimeter.
The final results of the frequency calculations, a sum
mary of data values used to get them, and the experimental
frequencies are shown in Table IV.
Tab le
k = 5,623 dynes/cm
a = 0.044 cm
MB (grams) d (cm)
0.00 0.231
10.15 0.182
11.50 0.175
12.00 0.173
13.50 0.166
16.20 0.152
18.10 0.143
20.70 0.130
21.20 0.128
24.55 0.112
IV
xo = 0.75 cm
aMSL = 4.43 grams
f f % error
e
4.50 4.32 4.1
2.90 2.95 +1.6
2.85 2.83 0.7
2.81 2.77 1.4
2.70 2.70 0.0
2.53 2.59 +2.4
2.43 2.49 +2.4
2.30 2.42 +4.9
2.28 2.32 +1.7
2.15 2.21 +2.8
The agreement seems to be very good. By considering
the values of the amplitude predicted from (234),
A Mg(Ad)
2 ak
we can understand better one source of error in the frequency
calculations. The quantity a is the most inaccurate in (234)
and is most responsible for error. Table V shows the calcu
lated amplitudes compared with experimentally observed ones.
The experimental amplitudes here are taken from the equilibrium
position and obtained from a least squares fit to five values.
The agreement is not bad, although it could be improved, most
probably by using more accurate values of a.
Table V
MB (grams)
10.15
11.50
12.00
13.50
16.20
18.10
20.70
21.20
24.55
Theoretical
A (cm)
5.26
5.52
5.63
5.89
6.21
6.38
6.47
6.50
6.43
Experimental
A (cm)
5.40
5.49
5.52
5.62
5.80
5.93
6.11
6.14
6.36
The last step in the calculations will be to examine the
validity of the approximate formula for Ad derived at the end
of Chapter II:
64cf2
(BAd4f )2 = (242)
(Bdlo) ad
where
B= 2M
kx
and
g2 M MSL 1/2
2 k(MSL+M)
% error
2.6
0.5
2.0
4.9
7.1
7.6
5.9
5.8
1.1
~
Calculating Ad for an MB of 20.70 grams one gets a value
of 0.23 centimeters, which is higher than the observed value,
and with this value of 0.23 a frequency of 2.34 Hertz. This
is higher than the original calculated frequency but is in
better agreement with the observed frequency of 2.42 Hertz.
Each step of the derivation of (242) can be examined to
see how appropriate the approximations are. With the original
data for an MB of 20.70 grams, the actual value of R3 is
found to be 7.626. The approximate formula (240),
h2
R = 1
3
0
gives a value of 7.622 which is in excellent agreement. The
actual value of 6+y is found to be 0.263, whereas the approxi
mate formula
8f
6 + y = O
h 2_4f
gives a value of 0.262 which is also in very good agreement.
The main source of error responsible for the two values of Ad
not agreeing any better is probably in
1 2
Ad = 2 gt2
in that the system is not in free fall but accelerating at
less than g due to the presence of some friction between the
sleeve and the shaft even when the sleeve is unlocked. A
smaller value of the acceleration would make the calculated
value of Ad smaller which is the correct direction. An
accurate measurement of the actual friction between the shaft
and the sleeve would call for rather sophisticated dynamic
measurements and is probably not worth doing in the sense that
Ad can be measured directly much more easily. In fact, it is
a characteristic of this type of problem that the friction
present is idiosyncratic and complex. It must usually be
idealized to a large extent for a class of problems rather
than dealt with for a specific case. In this sense the analy
sis here seems to be one of the few cases where a real situa
tion involving impact maintained oscillations is studied
completely.
CHAPTER IV
COMMENTS
This work begins with an analysis by J. P. Den Hartog
and shows that this approach to solving the problem of the
"toy woodpecker" is not satisfactory. Several possible reasons
for the failure of this analysis are discussed although a com
plete "negative" analysis indicating the exact reasons for
the failure of the model and proposed solution has not been
done. This work proposes a different physical model for the
woodpecker than that chosen by Den Hartog and one which corre
sponds more closely to the toys actually observed. This
physical model is then analyzed mathematically and the fre
quency and amplitude of the motion predicted for various
physical situations.
A major step in the theoretical development is to recog
nize the toy woodpecker as belonging to a class of phenomena
known as mechanical clocks. Clock theory has been worked out,
most notably by the Russian, A. A. Andronov, and his coworkers.
This theory was extended and applied to the toy woodpecker.
The method involves piecewise linear differential equations
matched together on the phase plane and is very suitable for
handling the impacts assumed in the model.
Actual working models of the toy were built and measure
ments of several variables of the models were taken and com
pared with the theoretical results. The agreement is very
good and it is clear that the theory successfully describes
the physical motion of the toy woodpecker. The calculations
and measurements are not difficult to do for the most part
and it is suggested that the toy woodpecker model, appropri
ately constructed, would make a useful mechanical analogue
for dealing empirically with various types of selfexcited,
impact maintained oscillations described by equations like
(21).
Although the analysis is necessarily more complex, the
theory can be extended to cover various other situations with
out too much difficulty. If, for example, we consider a situ
ation with a nonlinear spring we could have
MK + klx + k2x3 = F (41)
This could be analyzed in the same fashion as that in Chapter
II if some changes are made. The arithmetic decrement must
be recomputed for example. This can easily be done. The
potential energy for a spring with the nonlinear spring con
stant is
2 4
kx kx
V + 2 (42)
2 4
The work, W, spent in overcoming friction will not depend on
the velocity (for Coulomb friction) but only on the path.
40
If xol is the amplitude, measured from equilibrium, of the
first maxima and xo2 the amplitude of the first minima, the
work done in overcoming friction will be
w = Ix0ll+X02 1)F.
(43)
The potential energies at x01 and x02 are
2
klx01
1 2
k 2 ^
k2x01
4
and
2
klx02
2 2
k2x02
2
from (42). By conservation of energy we know that
V1 V2 = V
or, substituting in we have
2 4 2
klx1 k3x02 k02
2 4 2
4
k3xx02x02)F
4 (IX011+I x021)F
For k2 equal to zero we have the situation for the linear case,
IX01 Ix02 = 2aL
^Ol1 ~ k~ ^L
(44)
with
2aL I1l 1x021
(45)
as noted in Chapter II.
For k2 not equal to zero we have
S 2 2 2 4
 (x01x02 (x01x02) = (1x011+1x011)F
41
Canceling like factors results in:
k k
21 (Ix01 IX02 I + l4x 1 0 I2)( x01 Ix02) = F.(46)
This can be factored and written as
2 X1 2+ 2 2F
(Iol 1 Ix021)[1 + 2k (Ixol l +x02 l)] k
Substituting in (45) and defining
a. F (47)
N. L.
1
as the nonlinear value for the arithmetic decrement, we get
2 2+ 12
N.L. = aL[l + 2k (lx012x02 2 (48)
1
as the formula relating a to the observed amplitudes.
It is possible to solve (41) on the phase plane
although a more involved process of geometrical construction
is required. First, the equation is rewritten in the form
S k2 3 F
2+ M x + k x + = 0
and letting
2 k1
P2 M1 (48)
we have
2 k2 3 F 2 2
2 +[p x px] + p x = 0
or
x + p2(x+6) = 0
(49)
where
6 2 [M x 3+ ] (410)
P
Equation (49) is now in the standard form for solution by
the phaseplane delta method [6, p. 244]. This method in
volves approximating 6 as a constant over short intervals of
the motion and constructing the phase paths as short arcs
approximating the true path.
Linear friction can also be included in a very straight
forward way if it is proportional to the simple velocity
[2, p. 170]. For linear damping of a type proportional to
xn where n is not equal to one and may be fractional, more
sophisticated methods may be useful although the phaseplane
delta method can still be used. For small nonlinearities a
variety of methods are available [3,4] and can be applied as
necessary. On the whole, it is remarkable that the toy wood
pecker can be dealt with as effectively as it is without
requiring more advanced methods.
APPENDIX A
The problem is stated by Den Hartog as follows: The toy
woodpecker demonstrated at the end of the last lecture can be
idealized as shown in Figure AI. The sleeve m of length Z
fits loosely around the fixed stem. The birdie (mass M)
tilts the sleeve slightly within the clearance, so that point
contact occurs at A and B. When M stands still, this is a
selflocking frictional system. When M moves upwards, the
torque of the sleeve is released and the sleeve slips down a
bit. Assume the sleeve mass m (including the arm to the left)
to have its center of gravity in the center of the vertical
stem. Let the friction coefficient be f, the same value for
motion as for rest. Let x be the downward displacement of
the sleeve and let y be the downward displacement of M rela
tive to the sleeve m, measured from the position of static
equilibrium. The system is selflocking at rest (y = 0) and
x then is constant in time. For y negative (upwards) there
may be slip.
(a) Discuss the static equilibrium problem and write an
expression for the net downward force on m. This expression
is used only when positive; for negative values the friction
reverses sign. Here is the nonlinearity!
(b) Set up the two differential equations of motion of
the system. There will be two such pairs: one good while m
43
K
(a)
Figure AI
is slipping (variables xl,Y1); another good when m is at
rest (variables x2,y2).
(c) As you have seen in the model demonstration at the
end of Lecture 23, there exists a steady, periodic, motion
which is shown in Figure AII in its general characteristics.
The period is w t = 6 divided into two parts. First the
sleeve slips from A (ant = 0) until B(wnt = o0), when is
seizes by friction. Between o < wnt < 6 the sleeve m is
locked on the stem and only the bird M moves (the y shown
above is positive, i.e., the bird is down, binding the sleeve
on the stem). Since in general 60 is not equal to 2r (but
somewhat greater probably) the natural frequency is not a .
Study Figure AII until you clearly understand its physical
implications.
(d) Now, by exact, piecewiselinear method, write the
general solution of the four equations for x1,Y1 (good from
A to B above) and for x2,Y2 (good from B to C). Since all
individual equations are linear, this can be done without
great difficulty; the solution contains many (up to 6 or 7)
integration constants C1, C2, ....
(e) Set up the boundary (continuity) equations at the
points A, B, and C (wnt = 0, o o ). These will involve the
unknowns C1, C2, ..., o 0. Check and polish until you
have as many condition equations as there are unknowns.
. = nt
Figure AII
Solution
(a) When the sleeve is just ready to slip downwards, but
does not move yet, it must be in static equilibrium under
forces as shown in Figure AIII.
Moment equilibrium about the center requires:
(Mg + ky)a = Nk
or
N = (Mg + ky)a/k.
The vertically downward force on the sleeve then is
(M+m)g + ky 2fN = (M+m)g + ky 2 (Mg+ky)
( m 2fa 2fa
= Mg + m 2a + ky 1 (A0)
M a 1
The toy's dimensions are approximately m = 10; = 4; f 4'
and it is clearly selflocking when not vibrating, i.e., for
y = 0. Then the downward force must be negative or at least
not positive;
m 2fa 1 2 9
1+  1 + 2 1 < 0.
With these values, the downward driving force (a) becomes
9_ Mg ky, which means that it drives downwards only for
negative, i.e., upward values of y. This is physically clear,
because for upward y, the normal force N is diminished and
the friction lock is broken.
(b) and (c) Now consider expression (A0) to be the down
ward driving force on the sleeve during slipping motion

N
tTVJ
Figure AIII
Mg + Ky
N
'I
49
(x > 0). Whenever x = 0, the expression (a) is meaningless,
because the friction is no longer fN, but smaller than that.
The differential equations of motion are:
mo = Mg(l + fa + kyl( 2a) (Al)
for x > 0
(from A to B)
M(II + 1) + kyl = 0 (A2)
for 0 = 0 (A3)
for x = 0
(from B to C)
My2 + ky2 = 0 (A4)
(d) To solve these differential equations (all linear),
2 k
use the notation W = 2 This wn is the natural frequency
of the bird on its spring when the sleeve is fixed and not
slipping; it is not the frequency of the periodic motion
including slip. Substitute x1 from equation (Al) into
equation (A2), divided by M:
2 M m 2fa k 2fa
S+ wn Y1 + g( + M ) + m l 1(1) = 0
or
2 M 2fa M 2fa
l + wn Y[I + m (1 )] = g[1 + m (  )].
Now let
M 2fa 2
1 + ( 2fa A
M a 1
a constant of the system. (For = 10; = 4; f = we have
A2 = 9 or A = +3). Then:
2 2
Y (A n) Y1 = Ag
with the solution:
Y1= CleAnt + C2eA2nt g2 (A5)
n
To find xl, we return to equation (A2):
xl = Yl Wn Y
R = A2w2 [C eAwnt + C2eAwnt] 2 [ g
X1 = A1Wn n "
l = (A2 + 1)wn2[CleALnt + C2eAnt] + g
A2 1 [CeAwnt + CeAnt] gt2 + Ct + C. (A6)
X 2 3 t 4
Now we pass to equations (A3) and (A4) for the x2,Y2 func
tions for the noslip domain. We leave equation (A3)
without further work out, and integrate equation (A4) to:
Y2 = C5 sin w t + C6 cos n t. (A7)
The equations (A5), (A6), (A3), and (A7) are the answer
to part (d).
(e) Now the boundary conditions. Referring to the figure
on page 46, we see that the conditions for the y motion are:
(Yl) t= = (Y2) t=6 (A8)
n n o
(Yl)w t=O = (Y2w t=6 (A9)
(l) t = (Y2) =o (A10)
n o n o
(y t = o (2) t= (A11)
1 cm t=c = b2)w t=4
n o n o
The geometric conditions for the x motion are:
(Xl) t=o = 0 (A12)
n
(xl)w t=0 0 (A13)
n o
There is one more (and very important) condition for the x
motion of a mechanical nature. Just before A (before wnt = 0)
the sleeve was selflocking, but at point A it just ceases to
lock. The net driving force at A then just breaks away from
zero, so that
(:l)1 t=0 = 0. (A14)
n
The acceleration just before B ( nt = 0 ) is not necessarily
zero. Summarizing, we have seven condition equations (A8)
to (A14) and apparently we have eight unknowns:
C1,C2 'C3,C4 C5 ,C6 ,o ,6 .
But we notice that the constant C4 only occurs in equation
(A6) for the xl motion. The seven condition equations (A8)
to (A14) do not mention xl itself, but only x1 and xl, and
we see from equation (A6) that the constant C4 drops out in
the differentiation. Thus, we have seven unknowns
C1,C2 ,C3,C5 ,C6,' o,6
in the seven equations (A8) to (A14). In what follows the
letters 4o and 6 occur many times; from now on we will
just write and 6 without subscripts, for simplicity. Sub
stitution of equations (A5), (A6), (A3), and (A7) into
the condition equations (A8) to (A14) gives:
C1 + C2  = C sin 6 + C6 cos 6
1 2 2 5 6
n
A(C1 C2) = C5 cos C6 sin O
CleA + C2eA  = C5 sin 4 + C6 cos 4
n
A CleA
 A C2eA = C5 cos q C6 sin
(A2 + 1) (C1 C2) + C3 = 0
0 3
(A2 + 1) (Cle C2eA 3 = 0
n
(A2 + 1) 2(C1 + C2) + g = 0
Eliminate C3 between (A12) and (A13), leading to:
A +1 [C1(leA) C2 (eA)] +  = 0.
n
Solve for C5 and C6 from (A8) and (A0):
(A8)
(A9)
(A10)
(A11)
(A12)
(A13)
(A14)
(A12*)
C5 = (Cl + C2 )sin 6 + A(C1
n
C6 = (C1 + C2
 )cos 6 A(C1
n
 C2)cos 6
 C2) sin 6
and substitute these first into (A10), and later into (A1l)
CleAQ+C2eAQ = (C1+C2 g) (sin sin 4 + cos 6 cos 4)
n n
+ A(C1C2)(cos 0 sin 4 sin 6 sin P).
CleA1+C 2eA 2 = (C+C 2)cos(e)A(C C2)sin(e0).
n n (A10*)
A CleA_A C2 e = (C1+C2 ) (sin 0 cos 4 cos a sinp)
W
n
+ A(C1C2)(cos 0 cos p + sin 9 sin q)
A C2eAA C2eAO = (C1+C2 2)sin(6))+A(C1C2)cos(ec).
n (A11*)
By this time the set of seven equations (A8) to (A14) have
been whittled down to four equations (A10*), (A11*),
(A12*) and (A14), involving the four variables Cl, C2',
and (60). We now solve for C1 and C2 from (A14) and (A12*):
C g 1AceA
1 W 2 (A2 +1) 2eAO eA
1 2 2 A_ A4
2 (A2+1) 2e Ae
n
gC +A2A+ e
21 2 2eAe AO
P (A +1)
Then we substitute C1 and C2 and the two combinations
C1 +C2) and (C1C2) first into equation (A10*) and later
into (AllA):
into (A11*):
eAAce A eA+Ae Al 2(A21
P A A + P A A p(A+)
2e e 2e AeAc
2eA_ eeA
where we have used the abbreviation:
S= e p.
Dividing all terms by p and multiplying them with the denomi
nator of most of the terms gives:
eA AqeA 1 + eA1 + AqeAP = 1 = (A2+1)(2eAeAP)
= A2(2eAeA )cos 4 2 2A2+AeAAeAeA)sin n
or
2 cosh Ap 2Ap sinh A 2 A2(1cosh Ap) 2(1cosh A4)
= 2A(1cosh A))cos 4 + 2A(Apsinh Ac)sin 4
or, divided by 2:
[2 A2 + (A2+2)cosh AO Ap sinh Ap]
= [A2(cosh Apl)]cos 4 + [A(Apsinh Ap)]sin 4. (A10**)
In the same manner, we find equation (All**):
[A sinh A1 + A2p cosh AO]
= [A2(cosh Al)]sin 4 + [A(Apsinh Ap)]cos 4. (All**)
These are two equations in the unknowns p and 4, and they
have the form:
y = a cos + + B sin i
6= a sin I + 6 cos '
where a, 6, Y, and 6 are complicated functions of '. We now
eliminate P from the above pair by squaring each one of them
and then adding:
2 + 62 = (a2 + 2)(cos2' + sin2i).
Our final result in the analytical part of the investigation
thus is:
2 + B2 y2 62 = 0,
where
a = A2(cosh A'l)
6 = A(A'sinh A')
y = (A2+2)(cosh A'1) A' sinh A'
6 = A sinh A' + A2 cosh A'.
Further analytical work does not seem to yield any simplifi
cation, so that we now turn to numerical computation.
The quantity A, which must be positive, is a character
istic parameter of the system, expressing the amount of dry
friction locking. For A = 0 [see equation (A0)] the fric
tion f is so small that the sleeve is just ready to slide
down with zero acceleration. For A > 0, there is a self
locking at rest. The range of practical values of A appears
to be from 0 to 3. For A = 0, all four expressions a to 6 go
to zero, so that no direct conclusions can be drawn (we have
to employ a limit process). Hence, take
A = 0.1, 1, 2, 3.
Then put in P in steps of one degree from 0 to 360 degrees
and find the zeros of a2 + 2 Y2 It is possible
that several such zeros will be found. Since a squaring
operation has occurred at the end of our analysis, not all
of the Psolutions may be valid. For each q, the motion has
to be computed and inspected. This I have not done, and if
anyone ever does it, I shall be grateful if he or she will
send me the solution.
This concludes Den Hartog's remarks. There are probably
several reasons why the analysis is not more fruitful. With
out considering any of the experimental results as to the
relative stability of the two models in Figure AI, we can
see in the light of Chapter II that several things should have
been included in the above analysis:
1) Losses in the spring. These are assuredly present
and do influence the motion. When a larger than necessary
displacement is used to start the motion, most of the excess
energy seems to be dissipated in the spring as the sleeve is
not observed to start dropping for several cycles.
2) The impulsive nature of the motion when the sleeve
locks. In this light boundary condition (All) is incorrect
and probably should read
(1)w t= + = (Y2 t=
n o n o
57
where v is related to the drop of the system as in Chapter
II.
3) Energy balance considerations. These would relate
directly to whether stability is theoretically possible since
if energy input per cycle is not equal to energy loss per
cycle the motion will either delay and cease, or grow uncon
trollably.
APPENDIX B
Let us consider a situation where there are two constant
energy impulses delivered each period and let the losses be
due to linear (velocity dependent) friction. Let the loga
rithmic decrement per half cycle be denoted by d, and let the
phase velocity be y. Just after an impulse let the phase
velocity be y1. A half cycle later it will be
y = d2 (B1)
Just after this an impulse occurs which changes the velocity:
2 2 = h2 (B2
y2 Yo = h(B2)
The quantity h2 is a constant representing a constant energy
input during the impulses. Just after the impulse the
velocity is thus
Y2 = [Y2 + h2]1/2
or
y2 = [12ed + h2]1/2 (B3)
using (Bl). We require that yl equal y2 for periodic
motion so that we have
h (B4)
yl = 2 = Y d 1/2 (B4)
[1 e ]
This defines a closed phase path corresponding to periodic
oscillations. The motion can be further investigated by con
structing the graph of the sequence function (B3) [4, p.
433], sometimes called "Lamerery's diagram" [2, p. 161]. This
is shown in Figure BI. The vertical axis plots values of
Y2 and the horizontal axis, yl. For a stationary oscilla
tion yl equals y2 and this is represented by a straight line
through the origin. If the oscillations are not stationary
the sequence of amplitudes can be determined by (B3) and
Figure BI. If the initial amplitude is y', then the subse
quent amplitude will be y2, and the corrected amplitude will
be y2 on the yl axis. The next amplitude will be yl on the
Y2 axis and the sequence of amplitudes will continue until y
is reached, at which point a stable limit cycle exists.
Physically this corresponds to a situation where initially
the energy introduced is greater than that dissipated and the
oscillations grow in amplitude until a stable energy balance
is reached.
The important point to note is that y' can be made
arbitrarily small and the oscillations will still grow until
a stable limit cycle is reached. This implies soft excitation
is possible, a phenomenon which is not observed physically.
In fact, Andronov concludes his analysis [2, p. 172] with the
statement, "Again, a model with linear friction does not
explain the need for an initial finite impulse to start the
clock and must therefore be rejected. It is necessary to
YZ =l t ed t IJ
Figure BI
assume that the clock is a selfoscillating system with dry
friction."
The fact that Coulomb friction is appropriate can be
seen by considering a phase trajectory of the actual wood
pecker. This is shown in Figure BII. This trajectory
terminates on the x axis between the points (b f ,O) and
(b +f ,O) which characterizes the zone of static friction.
When a trajectory does not emerge from this zone the motion
ceases. Physically, this corresponds to the situation where
the amplitude of the bird mass is not sufficient to cause the
sleeve to unlock initially. The quantity b0 used in Chapter
II can now be seen as a parameter describing the friction
lock of the sleeve as far as its effect on the motion is con
cerned.
4.
1%(
III
A
b~i** hb
I II
I'.
Figure BII
" L 
APPENDIX C
The following are examples of data collected from the
videotape made of the woodpecker motion. The times listed
were read from an electronic timer with a digital readout
and are accurate to 0.005 second. The hundredths position
was sometimes difficult to read. The elapsed time is more
definite with each reading occurring onesixtieth of a second
after the last. The position readings refer to positions in
centimeters 0.15 read from a labeled grid behind the wood
pecker. Vertical displacements are labeled y and horizontal
ones, x. The first series of readings are for a bird mass of
11.30 grams.
Sleeve y
0.2
0.2
0.2
0.2
0.2
0.2
0.3
0.4
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.6
0.6
Bird x
7.2
7.9
8.9
10.0
10.6
11.1
11.3
11.3
11.2
10.0
11.0
11.0
11.2
11.3
11.1
10.7
9.8
Bird y Time
8.8 3.80
8.2 3.82
7.4 3.84
5.9 3.86
4.8 3.87
3.2 3.89
3.0 3.91
1.0 3.92
0.1 3.94
0.7 3.96
0.7 3.97
0.2 3.99
0.2 4.01
1.6 4.02
3.1 4.04
4.9 4.06
6.3 4.07
(continued)
Sleeve v Bird x Bird y Time
0.6 8.7 7.8 4.09
0.6 8.0 8.5 4.11
0.6 6.9 9.1 4.12
0.6 6.8 9.2 4.14
0.6 6.1 9.1 4.16
0.7 8.0 8.5 4.17
0.7 9.0 7.6 4.19
0.7 9.9 6.1 4.21
0.7 10.8 4.9 4.22
0.7 11.0 3.5 4.24
0.7 11.2 2.3 4.26
0.8 11.2 1.2 4.27
0.85 11.0 0.3 4.29
0.85 10.9 0.1 4.31
0.85 10.8 0.2 4.32
0.85 11.0 0.1 4.34
0.85 11.2 0.9 4.36
0.85 11.4 2.0 4.37
0.90 10.1 3.5 4.39
1.00 10.9 5.2 4.41
1.00 10.1 7.0 4.42
1.00 8.9 8.2 4.44
1.00 7.9 9.0 4.46
1.00 7.1 9.8 4.47
1.00 6.8 9.8 4.49
1.00 7.1 9.7 4.51
1.00 7.9 9.0 4.52
1.00 8.8 8.2 4.54
1.00 10.0 6.9 4.56
1.00 10.8 5.2 4.57
1.00 11.1 4.0 4.59
1.00 11.6 2.9 4.61
1.10 11.5 1.8 4.63
The following data were taken for a bird mass of 11.30
grams with the sleeve locked in place.
Bird x Bird y
12.1
12.0
12.0
11.9
12.0
12.0
12.0
7.3
6.5
6.0
5.9
6.2
6.8
7.7
Time
2.08
2.10
2.11
2.13
2.15
2.17
2.18
(continued)
Bird x Bird y Time
11.7 9.0 2.20
11.1 10.1 2.21
10.5 11.1 2.23
9.8 12.0 2.25
9.0 12.9 2.27
8.7 13.2 2.28
8.6 13.2 2.30
8.9 13.0 2.31
9.6 12.4 2.33
10.2 11.4 2.35
11.0 10.4 2.37
11.4 9.4 2.38
11.9 8.3 2.40
12.0 7.3 2.42
12.0 6.9 2.43
12.0 6.4 2.45
11.9 6.1 2.47
12.0 6.3 2.48
12.0 6.9 2.50
Figure CI shows a graph of some of the data points for
the position of the sleeve versus time and Figure CII shows
a graph of the speed of the sleeve versus time obtained by
calculating the slope of the line in Figure CI.
66
a
IS
F
LI 'I 5 I I < I 1 I 0I 0 I V SW 20 10 o
cS<3mI3l33 A A3as
67
I t
oII 01 b g L 7
C)Nc/ (;N21?W'uaaLt tqDa7
LIST OF REFERENCES
1. Den Hartog, Jacob P., Nonlinear Vibrations Study Guide,
Massachusetts Institute of Technology, Cambridge, Mass.,
1972.
2. Andronov, A. A., Vitt, A. A., and Khaikin, S. E., Theory
of Oscillators, AddisonWesley Publishing Company, Inc.,
Reading, Mass., 1966.
3. Butenin, N. V., Elements of the Theory of Nonlinear
Oscillations, Blaisdell Publishing Company, New York,
1965.
4. Minorsky, N., Introduction to NonLinear Mechanics,
Edwards Brothers, Inc., Ann Arbor, Michigan, 1947.
5. Den Hartog, Jacob P., Mechanical Vibrations, McGrawHill
Book Company, Inc., New York, 1956.
6. Jacobsen, Lydik S., and Ayre, Robert S., Engineering
Vibrations, McGrawHill Book Company, Inc., New York,
1958.
BIOGRAPHICAL SKETCH
Russell A. Roy was born February 7, 1942, in Norristown,
Pennsylvania. He completed most of his secondary education
in Pennsylvania and was graduated from Winter Park High School
in Winter Park, Florida, in June, 1960. He received a B.A.
in physics from Swarthmore College in 1964 and an M.S. in
physics from the University of Florida in 1967. He has a
wife, Eileen, and two children, Kevin and Katy.
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequa in scope and quality,
as a dissertation for the de re D Philosophy.
K6 t T. Rll 11 ran
Pro essor of ering Science,
Mech nics A space
Engine g
I certify that I have read this study and that in my
opinion it conforms to acceptable stan yards of scholarly
presentation and is fully adequate, i scope and quality,
as a dissertation for the degree of cctor of Philosophy.
awrence E. Malvern
Professor of Engineering Science,
Mechanics, and Aerospace
Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Robert L. Sierakowski
Professor of Engineering Science,
Mechanics, and Aerospace
Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Richard L. Fearn
Assistant Professor of Engineer
ing Science, Mechanics, and
Aerospace Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Vasile M. Pop6v
Professor of Mathematics
This dissertation was submitted to the Graduate Faculty of
the College of Engineering and to the Graduate Council, and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
December, 1974
e College of Engineering
Dean, Graduate School
