Content uploaded by Mark Nagurka
Author content
All content in this area was uploaded by Mark Nagurka
Content may be subject to copyright.
Aerodynamic Effects in a Dropped
Ping-Pong Ball Experiment*
MARK NAGURKA
Dept. of Mechanical & Industrial Eng., Marquette University, Milwaukee, WI, USA.
E-mail: mark.nagurka@marquette.edu
This paper addresses aerodynamic modeling issues related to a simple experiment in which a ping-
pong ball is dropped from rest onto a table surface. From the times between the ball-table impacts,
the initial drop height and the coefficient of restitution can be determined using a model that
neglects aerodynamic drag. The experiment prompts questions about modeling the dynamics of a
simple impact problem, including the importance of accounting for aerodynamic effects. Two
nonlinear aerodynamic models are discussed in the context of experimental results.
INTRODUCTION
ALTHOUGH OFTEN INCIDENTAL, sound
signatures that occur in dynamic events can serve
as useful clues in understanding the dynamic
behavior of a physical system. This paper describes
a simple experiment in which physical information
can be extracted from the time intervals (of silence)
between bounce sounds, after a ping-pong ball is
dropped onto a hard table surface. (Musician
Arthur Shnabel once said, `The notes I handle no
better than many pianists. But the pauses between
notesÐah, that is where the art resides.') In
particular, the time lapses in the audio signal
resulting from ball-table collisions can be exploited
to determine the height of the initial drop and the
coefficient of restitution at the impacts. The experi-
ment is simple to conduct, and affords students
the opportunity to compare their results with
established theoretical concepts and increasingly
complex models.
The experiment, shown in the photograph of
Fig. 1, involves dropping a ping-pong ball from
rest and recording the acoustic signals generated
by the ball-table impacts using a microphone (with
an integrated pre-amplifier) attached to the sound
card of a PC. Available software (shareware
package, GoldWave [4] ) can be configured to
display the temporal history of the bounce
sounds of successive impacts, from which the
times between bounces, or `flight times,' can be
determined. These flight times are used to calculate
the height of the initial drop and the coefficient of
restitution of the ball-table impacts, under several
assumptions, including negligible aerodynamic
drag.
The bouncing ping-pong ball in the experiment
can be modeled classically using the theory of
direct, central impact of particles, which follows
from Newton's laws of motion. This theory is
described in traditional textbooks [5, 6] that
present the fundamentals of dynamics. More
advanced treatments of planar and three-
dimensional impact for both particles and
rigid bodies are developed in specialized textbooks
[2, 7, 9]. A method for the measurement of the
coefficient of restitution for collisions between a
bouncing ball and a horizontal surface is given by
Bernstein [1], and has been adopted as the basis for
this experiment, as described below, and reported
previously by Nagurka [8]. This paper extends the
development and considers the applicability of
aerodynamic models in light of the experimental
results.
BACKGROUND: TABLE TENNIS
The official rules of table tennis specify the
characteristics and properties of a ping-pong ball:
`The ball shall be spherical, with a diameter of 38 mm.
The ball shall weigh 2.5 g. The standard bounce
required shall be not less than 23.5 cm nor more
than 25.5 cm when dropped from a height of 30.5 cm
on a specially designed steel block. The standard
bounce required shall not be less than 22 cm nor
more than 25 cm when dropped from a height of
30.5 cm on an approved table.'
The specifications for the rebound height can be
used to determine the coefficient of restitution,
which reflects properties of the ball and the surface
of impact.
In the interest of slowing down the game and
making it more `viewer friendly' the official rules
were changed by the International Table Tennis
Federation in September 2000 and now mandate a
larger (40 mm) and heavier (2.7 g) ball. This paper
follows the earlier rules (presented above). More
information about table tennis can be obtained
from the ITTF web site, http://www.ittf.org.
* Accepted 10 January 2003.
623
Int. J. Engng Ed. Vol. 19, No. 4, pp. 623±630, 2003 0949-149X/91 $3.00+0.00
Printed in Great Britain. #2003 TEMPUS Publications.
THEORETICAL FOUNDATION
The development assumes that a ball is dropped
from height h
0
onto a tabletop, taken as a massive
(i.e., immobile), smooth, horizontal surface. The
ball is modeled as a particle, and hence rotation is
neglected. Vertical motion only is considered, as
depicted schematically in Fig. 2, which shows the
ball height from the surface as a function of time
for the first few collisions. Due to the inelastic
nature of the ball-table collisions, the maximum
height of the ball decreases successively with each
impact.
The free-body diagram of a falling ball, drawn in
Fig. 3, accounts for three external forces:
.the body force Fbody mg (the weight)
.the buoyancy force Fbuoy gV
.the aerodynamic drag force Faero 1
2ACDv2
where m,v,A1
4D2, and V1
6D3are the ball
mass, velocity, cross-sectional area, and volume,
respectively, for a ball of diameter D, density of air
, drag coefficient C
D
, and gravity acceleration g.
(For a rising ball, the aerodynamic force in Fig. 3
acts in the opposite direction.)
For a ping-pong ball in air at room temperature,
it can be shown (see Results section) that the
buoyancy force is approximately one percent of
the weight. Hence, buoyancy is presumed to have a
negligible effect and not treated in the analysis
below.
Neglecting aerodynamic drag
For the simplest model, aerodynamic resistance
is neglected. The acceleration of the ball during
flight is then constant, due to gravity only. The
equations of motion are gdv=dt for a falling ball
and gdv=dt for a rising ball, and are readily
integrated.
Under the assumption of no air resistance, the
flight time between bounces is comprised of equal
and symmetrical time segments for the rise and fall
Fig. 1. Photograph of experimental setup (ball to be dropped in front of microphone).
Fig. 2. Ball height vs. time for first few bounces (assuming no aerodynamic drag).
M. Nagurka624
phases, i.e., the time from the surface to the top of
the trajectory is the same as the time from the top
back to the surface, each being half of the total
flight time, T
n
, between the nth and (n1)th
bounce. The post-impact or `takeoff ' velocity, v
n
,
which is the speed of the ball associated with its nth
bounce is then:
vngTn
2
n1;2;3;...1
and is equal to the velocity prior to impact at the
(n1)th bounce, as shown in Fig. 2.
The coefficient of restitution is a composite
index that accounts for impacting body geome-
tries, material properties, and approach velocities.
Assuming a constant coefficient of restitution, e,
for all impacts,
vnevn1env0n1;2;3;...2
where v
0
is the velocity prior to the first bounce.
An expression for the flight time, T
n
, can be
obtained by equating Equations (1) and (2) giving:
Tnen2v0
g
n1;2;3;...3
which, after taking logarithms of both sides, can be
written as
logTnnlog elog 2v0
g
n1;2;3;...
4
Plotting Equation (4) in a graph of log(T
n
)vs.n
yields a straight line with slope alog eand
ordinate intercept blog(2v
0
/g). The slope can be
used to determine e, and the intercept can be used
to determine v
0
from which the initial height h
0
can
be found. In general, the height h
n
reached after the
nth bounce can be written as:
hnv2
n
2gn0;1;2;...5
from conservation of energy. The coefficient of
restitution and the drop height can then be
expressed from the slope and intercept, respectively,
as:
e10a;h01
8g10b6
These simple relations provide an easy means to
predict the coefficient of restitution and the initial
drop height solely from knowledge of the times
between bounces. They are predicated on an
analysis that neglects aerodynamic drag, an
assumption that is addressed later in the paper.
Total time and distance
It can be shown (Nagurka, 2002) that under the
assumption of no air resistance the total time
required for the ball to come to rest, T
total
, is:
Ttotal 1e
1e
2h0
g
s
! 7
and the total distance traveled along the path,
S
total
, is:
Stotal 12X
1
n1
e2n
"#
h0;8
both of which are finite, although the number of
bounces in theory is infinite.
Accounting for aerodynamic drag
Models that consider the retarding effect of
aerodynamic drag during the flight phases can be
developed. With aerodynamic resistance included,
the flight time, T
n
, between the nth and (n1)th
bounce is comprised of unequal and asymmetrical
time segments for rise and fall. Since aerodynamic
losses occur in both the rise and fall portions, the
time from the surface to the top of the trajectory is
not equal to the time from the top back to the
surface, and the take-off speed v
n
of the ball
associated with its nth bounce is not equal to the
speed prior to impact at the (n1)th bounce.
The equation of motion for a falling ball can be
written in a form that shows the aerodynamic force
normalized with respect to the body force, i.e.
g1Faero
Fbody
dv
dt
Similarly, the equation of motion for a rising ball
can be written as:
g1Faero
Fbody
dv
dt
The magnitude of the normalized force term
determines the importance of aerodynamic effects
in the model.
In this paper two aerodynamic models, both of
the form:
Faero 1
2ACDv2
Fig. 3. Free-body diagram of falling ball showing body,
buoyancy, and aerodynamic forces.
Aerodynamic Effects in a Dropped Ping-Pong Ball Experiment 625
are considered. One model assumes a constant
drag coefficient C
D
, whereas the alternative
model assumes a velocity-dependent coefficient
based on an empirically determined relation. The
two models are developed below, and then
compared in the Results section.
Constant drag coefficient
For the case in which the drag coefficient, C
D
,is
constant, the normalized aerodynamic force can be
written as a quadratic function of velocity, namely:
Faero
Fbody v2where ACD
2mg
The equations of motion can then be expressed as:
g1v2dv
dt 9
for the ball falling, and:
g1v2dv
dt 10
for the ball rising. These equations can be solved
analytically, and give:
vvttanh gt
vt
11
for the ball falling from rest, and:
vvttan tan1voff
vt
gt
vt
12
for the ball rising with initial take-off velocity, v
off
,
at time t0, where vt1=
pis the terminal
velocity, or the fastest speed the ball can attain.
The expression for v
t
follows from setting the
derivative in Equation (9) to zero. Equation (12)
is valid only as long as vis positive.
Additional expressions can be derived from the
equations of motion, Equations (9) and (10). For
example, if the ball falls from rest, the displace-
ment, y(measured vertically down from the height
at release), can be written as a function of velocity
as:
yv2
t
2g
ln v2
t
v2
tv2
13
or, as a function of time, as:
yv2
t
g
ln cosh gt
vt
14
Rearranging Equation (13):
v
1e
pvtwhere 2gy
v2
t
15
A ball imparted with initial take-off velocity, v
off
,
will rise to height, h, given by:
hv2
t
2g
ln 1 v2
off
v2
t
! 16
in time, t,
tt
g
tan1off
vt
17
The equations derived here rely on an aerodynamic
force expression involving the velocity squared
and a constant drag coefficient. The problem of
selecting a proper value for this constant is
addressed in the Results section, as is the validity
of the underlying model.
Velocity-dependent drag coefficient
Classic fluid dynamics experiments show that the
drag coefficient of flow over a smooth sphere is not a
constant, but a function of the velocity. (The drag
coefficient is usually plotted as a function of the
Reynolds number, which itself depends on velocity.)
For the case of a drag coefficient, C
D
, that is velocity-
dependent, the normalized aerodynamic force can
be written as a nonlinear function of velocity,
Faero=Fbody vwhere vA=2mgCDvv2.
The equations of motion can then be expressed as:
g1v dv
dt 18
for the ball falling, and
g1v dv
dt 19
for the ball rising. These equations can not be
solved analytically, but can be integrated
numerically given C
D
(v).
Graphs of the normalized aerodynamic force as
a function of velocity for this model as well as the
constant drag coefficient model are presented in
the Results section.
RESULTS
A ping-pong ball (Harvard, one-star) was
dropped from rest from a measured initial height
of 30.5 cm onto a butcher-block-top lab bench
such that it landed and bounced near the micro-
phone. (See Fig. 1.) From the bounce sounds,
indicated by spike amplitudes in the software
display window of the audio signal (Fig. 4), the
times between the bounces (the flight times) were
found.
The buoyancy force for the ball in air at room
temperature and pressure was calculated as:
Fbuoy 3:46 104N, or 1.41% of the weight
(Fbody 2:45 102N), and its effect was
neglected. The coefficient of restitution, e, and
the height of the initial drop, h
0
, were determined
from the linear regression curve fit of the log(T
n
)
vs. ndata presented in Fig. 5 for the first ten
bounces.
The results, calculated from Equation (6), are:
e0.9375 and h
0
26.7 cm, representing a 12.3%
error in drop height. The predicted coefficient of
M. Nagurka626
restitution is higher than the ranges based on the
rules of table tennis: 0:878 e0:914 for a `steel
block' and 0:849 e0:905 for an `approved
table'. The differences in the predicted vs. actual
initial heights and in the coefficients of restitution
may be attributable to several factors, including
neglecting aerodynamic drag in the analysis.
Under the assumption of no drag, the total time
required for the ball to come to rest is a function of
the coefficient of restitution, as indicated in Equa-
tion (7). In the experiment, the total time before
the ball comes to rest is 7.50 s corresponding to a
coefficient of restitution of 0.938 assuming an
initial drop height of 30.5 cm. Similarly, the total
distance traveled by the ball before it comes to rest
is a function of the coefficient of restitution for a
given initial height, as indicated in Equation (8).
For a drop height of 30.5 cm and a coefficient of
restitution of 0.94 the total distance traversed is
4.94 m, or over 16 times the initial height.
The first aerodynamic model, for which the drag
force is a quadratic function of velocity, requires a
Fig. 4. Bounce history audio signal with ball-table impacts indicated by spikes. (For ball dropped from 30.5 cm, bounce sounds end
in 7.5 s.)
Fig. 5. Log of flight time vs. bounce number.
Aerodynamic Effects in a Dropped Ping-Pong Ball Experiment 627
value of the constant drag coefficient, C
D
. One
approach is to select a value of C
D
corresponding
to the maximum Reynolds number in the flow.
This occurs at the maximum velocity, v
0
, which
arises prior to the first impact. As a first approx-
imation, the equation, v0
2gh0
p, from Equation
(5), which assumes no aerodynamic drag, can be
used to predict a maximum approach speed of
2.45 m/s for a 30.5 cm drop. For a ping-pong ball
traveling in air at room temperature at this velo-
city, the Reynolds number is Re 6200, indicating
turbulent flow. Experiments in fluid mechanics
show that at this value and more generally in the
range 103<Re <3105the drag coefficient for a
smooth sphere is relatively constant at C
D
0.4.
(See, for example, Fig. 9.11 of [3].) Separation of
the boundary layer occurs behind the sphere, and
the pressure is essentially constant and lower than
pressure at the forward portion of the sphere. This
pressure difference is the main contributor to the
drag.
A constant value of C
D
0.4 corresponds to a
ball terminal speed of v
t
9.38 m/s. At the terminal
speed, which is not attained in the experiment
(since the initial drop height is too small), the
aerodynamic force is in equilibrium with the
weight. Accounting for aerodynamics (with the
constant value of C
D
), the velocity prior to the
first impact can be calculated from Equation (15)
as 2.405 m/s, or approximately 2% less than the
non-aerodynamic prediction.
The more advanced aerodynamic model consid-
ers the fact that the drag coefficient corresponding
to flow over a smooth sphere is a nonlinear
function of the Reynolds number, and hence is
velocity dependent (Fig. 9.11 of [3]). Using this
(experimentally determined) function, the normal-
ized aerodynamic force:
v A
2mg
CDvv2
Fig. 6. Normalized aerodynamic force vs. velocity for nonlinear drag coefficient model.
Fig. 7. Comparison of normalized aerodynamic force vs. velocity for two models.
M. Nagurka628
can be constructed, as shown in Fig. 6. At the
maximum approach speed of 2.4 m/s for a 30.5 cm
drop, the aerodynamic force is 6.4% of the body
force. Although not achieved in the experiment, a
ball terminal speed of 8.8 m/s is predicted using the
nonlinear drag coefficient model.
A comparison of the two models, i.e., the
constant vs. nonlinear drag coefficient cases,
shows that the differences are minor. When plotted
in the log-log graph of Fig. 7, the nonlinear drag
coefficient model can be seen to closely match the
straight line representing the constant coefficient
model (with slope 2) except for slight deviations.
The differences occur at low velocities where the
aerodynamic force levels are insignificantly small,
and at higher velocities that are not achieved by
the ball in the experiment. The figure implies that
the two models are virtually equivalent (for speeds
below 0.5 m/s the force differences are inconse-
quential), and suggests indistinguishable effects
on the predicted behavior of the ball. This has
been confirmed by simulation studies, which show
no differences when comparing the height histories
of the two models. Figure 8 gives the height vs.
time predicted by both models for the case of a
coefficient of restitution of 0.935. There appears to
be no additional predictive value using the more
advanced model that accounts for the velocity
dependent drag coefficient.
CONCLUSION
This paper has addressed aerodynamic modeling
issues related to a simple experiment that engages
students in creative thinking about the dynamics of
a simple impact problem. Several findings emerge:
1. By adopting a model which neglects aerodynamic
resistance,the time intervals between bouncescan
be used to determine with reasonable accuracy
the initial height of a dropped ping-pong ball
and the coefficient of restitution (assumed
constant).
2. Although aerodynamic effects do not dominate
the trajectory dynamics, they influence the
bounce history and are needed for more
accurate predictions.
3. The constant drag coefficient model is a very
reasonable approximation of the nonlinear
coefficient model, and can be used to determine
the aerodynamic force and the predicted height
history of the ball.
This experiment has been incorporated success-
fully into a junior-level mechanical engineering
course, `MEEN 120: Mechanical Measurements
and Instrumentation' at Marquette University.
The experiment has minimal requirements for
hardware (PC with sound card, microphone,
ball), software (GoldWave shareware), and time
(taking approximately 5±10 minutes to conduct).
Since it does not require any special laboratory
facilities, the experiment can be demonstrated in a
classroom using a notebook computer with a
sound card and microphone.
The `open-ended' nature of the experiment, as
well as its amazing simplicity, has been attractive
to students. Many questions beyond those ad-
dressed here can be posed to trigger discussion
and prompt student thinking about the physics of
impact and assumptions of appropriate models.
AcknowledgementÐThe author wishes to thank several profes-
sional colleagues who have been supportive of this work and
provided technical contributions: Dr. Shugang Huang and Prof.
G. E. O. Widera (Marquette University), Dr. Allen Duncan
(EPA), Prof. Imtiaz Haque (Clemson University), Prof. Thomas
Kurfess (Georgia Tech), Prof. Uri Shavit (Technion, Israel) and
Prof. Burton Paul (Emeritus, University of Pennsylvania). The
author is grateful for a Fulbright Scholarship (2001±2002) and
the opportunity to work with Prof. Tamar Flash of the
Department of Computer Science & Applied Mathematics at
the Weizmann Institute (Israel).
Fig. 8. Ball height vs. time for entire bounce history accounting for aerodynamic effects.
Aerodynamic Effects in a Dropped Ping-Pong Ball Experiment 629
REFERENCES
1. A. D. Bernstein, Listening to the coefficient of restitution, American Journal of Physics,45(1), 1977,
pp. 41±44.
2. R. M. Brach, Mechanical Impact Dynamics: Rigid Body Collisions, John Wiley & Sons, New York,
NY (1991).
3. R. W. Fox, and A. T. McDonald, Introduction to Fluid Mechanics (5th ed) Wiley, New York (1998).
4. GoldWave software can be downloaded from http://www.goldwave.com
5. D. T. Greenwood, Principles of Dynamics (2nd ed) Prentice-Hall, Upper Saddle River, NJ (1988).
6. R. C. Hibbeler, Engineering Mechanics: Dynamics (9th ed) Prentice-Hall, Upper Saddle River, NJ
(2001).
7. R. F. Nagaev, Mechanical Processes with Repeated Attenuated Impacts, World Scientific, River
Edge, NJ (1999).
8. M. L. Nagurka, A simple dynamics experiment based on acoustic emission, Mechatronics,12(2),
2002, pp. 229±239.
9. W. J. Stronge, Impact Mechanics, Cambridge University Press, Cambridge, UK (2000).
Mark Nagurka received his BS and MS in Mechanical Engineering and Applied Mechanics
from the University of Pennsylvania (Philadelphia, PA) in 1978 and 1979, respectively, and
a Ph.D. in Mechanical Engineering from MIT (Cambridge, MA) in 1983. He taught at
Carnegie Mellon University (Pittsburgh, PA) from 1983±1994, and was a Senior Research
Engineer at the Carnegie Mellon Research Institute (Pittsburgh, PA) from 1994±1996. In
1996 Dr. Nagurka joined Marquette University (Milwaukee, WI) where he is an Associate
Professor of Mechanical and Biomedical Engineering. Dr. Nagurka has expertise in
mechatronics, mechanical system dynamics, controls, and biomechanics. He has published
extensively, is the holder of one U.S. patent, and is a registered Professional Engineer in the
State of Wisconsin and Commonwealth of Pennsylvania. He is a Fellow of the American
Society of Mechanical Engineers (ASME). During the 2001±2002 academic year, he served
as a Fulbright Scholar in the Department of Computer Science and Applied Mathematics
at The Weizmann Institute (Rehovot, Israel).
M. Nagurka630
