Modeling and Analysis of A Weight Driven Mechanical Tower Clock
Modeling and Analysis of A Weight Driven Mechanical Tower Clock
3
Marriott Waterfront, Baltimore, MD, USA
June 30-July 02, 2010
Abstract: The tower clock has served European communities weights, escapement, pendulum, and eventually, dial(s) with
beginning in the 14th century and American towns since the 18th hands. As discussed by Leinhard [5], the evolvement of
century. Although the early clocks acted as audible alarms, sophisticated mechanical clocks with artistic features served
they progressed to feature time dials and synchronized as a catalyst for societal progression from the renaissance
melodies from struck bells. The mechanical tower clock
(15th century) to the industrial revolution (19th century).
consists of gear trains, escapement, pendulum, driving weights,
large dials with hour/minute hands, and bell(s) with striking Simply put, clock technology helped to drive human
mechanism. These clock components offer excellent examples innovation and critical thinking. Landes [2] has written that
of fundamental engineering concepts and control system the clock should be considered as “… one of the great
principles. In this paper, a Seth Thomas Graham deadbeat inventions in the history of mankind – not in a class with fire
escapement tower clock has been investigated. A brief survey of and the wheel, but comparable to movable type in its
the societal impact of horology technology has been offered to
revolutionary implications for cultural values, technological
motivate the study. The general operation of a weight driven
tower clock has been reviewed and the governing mathematical change, social and political organizations, and personality.”
equations discussed. Representative numerical and The tower clock mechanism embodies fundamental
experimental results have been presented and discussed to
validate the model.
concepts from the mechanical and control engineering
disciplines. The analysis of the mechanical drive with
I. INTRODUCTION integrated escapement and pendulum clearly illustrates the
precise conversion of potential energy into periodic motion
T HE design of early mechanical clocks featured weight
driven escapements that controlled the periodic gear
motion which measured time in equal intervals. By 1280
for time keeping. Further, the Graham deadbeat escapement
[6] interfaces with the escape wheel to impulse the
pendulum which determines the minute hand speed [7]. In
AD, the available clock construction components included
flywheels, cranks, differential gears, and a regulator [1]. In this paper, the dynamic behavior of a time and strike tower
general, the main clock elements are the power source clock will be analyzed using Newtonian mechanics. In
(weights, main spring), regulator (escapement – verge and Section II, the operation of a Seth Thomas tower clock will
foliot, anchor with pendulum), and display or bell. The be discussed with supporting illustrations. A mathematical
escapement (or mechanical controller) maintains the power model will be presented in Section III for the weight driven
source by carefully releasing the stored energy, introduces wheel, gear train, escapement, pendulum, and hand motion
oscillations to divide time into equal parts, and counts beats for the clock’s time side. In Section IV, numerical and
with metal pallets [2]. In other words, the escapement experimental results are offered to validate the mathematical
controls the periodic motion of mechanical clocks through model. The summary is contained in Section V. Finally, a
the rhythmic behavior of an oscillating mechanism, which Nomenclature List is provided in the Appendix.
advances the gears in small increments to conserve energy.
Lastly, the pendulum with bob offers a consistent swing II. SETH THOMAS TOWER CLOCK OPERATION
(period) due to its length and center of mass.
The Seth Thomas & Company (1853-1942) of
Tower clocks appeared in European monasteries to ring Thomaston, CT at the turn of the 20th century offered a well
bells in the 14th century [3]. The word “clock” originates engineered and factory manufactured tower clock [4]. In
from the Latin word “clocca” (bell) which monasteries used 1905, a Seth Thomas Model 16B Graham deadbeat
to announce morning prayer time [4]. Tower clocks escapement tower clock (refer to Figure 1) with eight day
subsequently appeared in bell towers with their construction operation was purchased for Tillman Hall by Clemson
evolving over time with gear trains, lantern pinions, driving University President Mell. The clock featured four dials that
were 248.9cm in diameter. The setting and strike weights
J. Wagner, Mechanical Engineering Department, Clemson University, were 45.4kg and 124.7kg, respectively, with specified drops
Clemson, SC 29634 (e-mail: [email protected]) of 15.2m and 50.3m. The pendulum rod length was 243.8cm
C. Huey, Mechanical Engineering Department, Clemson University, with a 61.2kg bob and period of approximately 3.0s. The
Clemson, SC 29634
K. Knaub, National Watch & Clock Museum, Columbia, PA 17512 tower platform was cast iron with an A-frame construction
E. Volk, Western Carolinas Chapter 126, National Association of Watch supporting an open bed. Finally, a stationary McShane Bell
and Clock Collectors, Lake Toxaway, Asheville, NC 28747 Foundry (Baltimore, MD) single bell was struck on the hour
A. Jagarwal, Mechanical Engineering Department, Clemson University,
Clemson, SC 29634 which has been long since removed.
635
Authorized licensed use limited to: University of Split. Downloaded on August 07,2023 at 10:17:37 UTC from IEEE Xplore. Restrictions apply.
Lepschy et al. [10], a verge and foliot escapement was To calculate the impulse torque, T '(θ ) , the escape wheel and
modeled by assuming inelastic collisions between the two verge (pallet) geometry must be considered. Let the
blades (escapement pallets) and crown (escape) wheel. Two parameters rD 2 and p denote the escape wheel radius and
common clock escapements, dead beat and gravity, were the distance from the verge pivot to the pallet’s impulse face
mathematically modeled by Kesteven [11] to investigate midpoint. Further, the parameter δ represents the angle
pendulum period error. In our study, the motion of the tower
between the escape wheel’s radius point of contact with the
clock dead beat escapement will be described using the
given pallet, and a tangent line to that pallet’s impulse face.
underlying framework suggested by Kesteven.
The escape wheel’s operation reflects a repeated ordered
Weight Driven Wheel: The torque acting on the time sequence of impulse, hold, and free spins events. Let the
train great wheel arbor, TA1 , has been created by the barrel spacial locations where impulse events occur to the
and attached weight suspended by the steel cable so that momentarily connected (through the verge) escapement and
TA1 = m A grA (1) pendulum assembly be defined by the angles ζ1 , ζ 2 and ζ 3 .
where m A and rA denote the external mass in the time train For a left-to-right (right-to-left) pendulum swing, the
barrel and radius of this corresponding drum. impulse event begins and ends at θ = −ζ 3 ( ζ 3 ) and
Gear Train: A series of three gear interfaces reduce the θ = ζ1 ( ζ 2 ). Specifically, an escape wheel tooth impacts a
clock’s input torque, TA1 , while amplifying the angular pallet’s locking face to transfer the drive torque to the verge
rotation of the great wheel, ω A1 , for application to the (crutch to pendulum) assembly. In contrast, the hold event
corresponds to the sliding motion of the tooth on the pallet
escape wheel [12]. The gears have been assumed to be ideal
which does not permit significant wheel rotation. Lastly,
given the non-flexure of the teeth and low operating speeds
free spin occurs when the pallets are not in contact with the
so that
teeth beginning at θ = ζ1 ( ζ 2 ) and terminating when
TD 2 = (1 Rt ) TA1 , ω D 2 = ( Rt ) ω A1 (2)
θ = ζ 2 ( ζ1 ) for the previously defined pendulum swing. In an
⎛ D ⎞⎛ D ⎞⎛ D ⎞
where the train ratio becomes Rt = ⎜ A1 ⎟ ⎜ B 2 ⎟ ⎜ C 2 ⎟ . ideal situation, the deadbeat escapement features angles so
⎝ DB1 ⎠ ⎝ DC1 ⎠ ⎝ DD1 ⎠
that ζ 2 ≅ −ζ1 with ζ1 < 0 and a symmetrical ζ 3 impulse limit.
636
Authorized licensed use limited to: University of Split. Downloaded on August 07,2023 at 10:17:37 UTC from IEEE Xplore. Restrictions apply.
( )
V is J PV = m1 r12 / 2 + m1 ( r1 + R )2 + m2 ( L2 / 3) . The pendulum In Table 1, the wheel and pinion database for the time
side of a Seth Thomas tower clock are displayed. The arbor,
period becomes T = 2π Mgr J PV by neglecting losses gear, symbol, description, number of teeth, and outer
(assuming Bu = 0 ) and without applying torque. diameter are listed for each component. For completeness,
the strike side arbor, gear, symbol, description, and
The swing of the pendulum was approximately ±0.175
numerical identifier are summarized in Table 2. These
(rads) about the vertical axis to maintain clock timing. The
mechanical element labels are used in Figures 2 and 3 to
pendulum receives impulsive torques, T ''(θ ) , through the
describe the clock’s operation. Finally, the clock parameter
crutch assembly about positions −ζ 3 to ζ1 and ζ 2 to ζ 3 so values are listed in Table 3.
⎧ ⎧ dθ
⎪ ⎛ p ⎞⎛ 1 ⎞ ⎪⎪−ζ 3 < θ < ζ1, dt < 0 Table 1: Tower clock time side nomenclature and values
⎪⎪T ⎜ ⎟ ⎜ ⎟ tan δ for ⎨
T ''(θ ) = ⎨ D 2 ⎝ rD 2 ⎠⎝ t0 ⎠ ⎪ ζ < θ < ζ , dθ > 0 (6) Arbor Gear Sym Description N D
⎪ ⎪⎩ 2 3
dt A 1 A1 great wheel 120 314.9
⎪
⎪⎩ 0 otherwise B 1 B1 input pinion 20 55.35
B 2 B2 arbor C wheel 112 222.7
Hand Motion: The hand motion on each clock face
B 3 B3 cam - strike side - -
depends on the rotational speed of arbor B and transmission
B 4 B4 bevel gear 40 106.7
through bevel gear sets B4-F1 and F2-G2, as well as the
B 5 B5 pinion to J1 13 24.04
motion works (shown on arbors H and I). The angular speed
C 1 C1 arbor C pinion 14 28.96
of the minute, ωm, and hour, ωh, hands becomes
C 2 C2 drive wheel - D1 90 129.0
ω m = (1 Rs )ω B 4 , ωh = (1 Rh ) ωm (7)
D 1 D1 arbor D pinion 12 42.33
⎛D D ⎞ ⎛D D ⎞ ⎛ D D ⎞ D 2 D2 escape wheel 20 95.25
for Rs = ⎜ F1 G 2 ⎟ , Rm = ⎜ B 2 C 2 ⎟ , Rh = ⎜ H 1 I1 ⎟ , and
⎝ DB 4 DF 2 ⎠ ⎝ DC1DD1 ⎠ ⎝ DG1DH 2 ⎠ E 1 E1 verge - -
ω B 4 = Rmω D 2 . E 2 E2 crutch assembly - -
F 1 F1 bevel gear 40 106.7
F 2 F2 bevel gear 40 106.7
G leading off rod - -
G 1 G1 minute arbor pin. 15 26.62
G 2 G2 bevel gear 40 106.7
H intermediate arb. - -
H 1 H1 motion wheel 45 14.06
H 2 H2 motion wheel 12 21.33
I hour pipe - -
I 1 I1 hour hand pinion 48 78.90
U pendulum
Figure 4: Clock deadbeat escapement with dual pallets V pivot point
IV. NUMERICAL AND EXPERIMENTAL RESULTS The simulated and experimental escape wheel angular
A computer simulation has been created in Matlab/ displacement for three beats is displayed in Figure 5. As
Simulink with Δt =1.0e-03s to describe the tower clock’s expected, the wheel progresses through 54º during a 9s
operational behavior based on the dynamic model. To window or 6º/s; a full 360º rotation requires 60 seconds. The
validate the numerical results, the motion of the tower two graphs show general agreement with the progression of
clock’s escape wheel, D2, and anchor escapement, E1, were the escape wheel. However, the simulation results do not
experimentally measured. Two miniature gyroscope sensors reflect the small (negative) backward rotation of the wheel
(Analog Devices ADXRS150) were attached to these after each 9º advance. This phenomenon indicates a model
components and interfaced to a 200Hz National Instruments limitation due to the assumption of instantaneous collisions
data acquisition system with 12-bit resolution. The yaw rate rather than an extended contact between the pallet and tooth.
gyros used a third-order filter to prepare the analog velocity Further, the non-symmetry pallet interaction with the escape
signal. An acceleration signal was derived through analog wheel becomes evident when comparing Figures 5a and 5b;
differential with an attached third-order filter. Each filter this phenomenon can be corrected through clock adjustment.
was designed for a 50Hz 3db point. For the escape wheel, The escapement (anchor, pendulum) motion in Figure 6
the acceleration was numerically calculated due to signal reflects the anticipated dominant oscillatory behavior with a
measurement saturation issues. The two angular period of T=3.0s (f=0.33Hz). The estimated pendulum
displacements (Figures 5a and 6a) were analytically motion in Figure 6b agrees favorably with the experiment
computed using a standard numerical integration method. anchor angular position of Figure 6a in terms of frequency
and proportional magnitude. However, a direct comparison
637
Authorized licensed use limited to: University of Split. Downloaded on August 07,2023 at 10:17:37 UTC from IEEE Xplore. Restrictions apply.
of amplitude is difficult given that two different, but ACKNOWLEDGEMENTS
periodically connected, components have been presented. The authors wish to thank the NAWCC Western
Carolinas Chapter (W. Arcuri, E. Kara, B. Schmidt, and G.
Table 2: Tower clock strike side nomenclature Volk) for restoring the clock. We also recognize the efforts
Arbor Gear Sym Description of Clemson staff and students D. Moline, J. Savitsky and L.
J 1 J1 short half shaft interfaces to B5 Srutkowski.
J 2 J2 snail with 12 steps
K 1 K1 pinion in strike train to M1
K 2 K2 short arm to start / stop striking
K 3 K3 fly (governor)
L 1 L1 strike rack
M 1 M1 wheel driven by K1
M 2 M2 pinion driven by strike drum
M 1 M3 gathering pallet (moves rack)
N 1 N1 great wheel interfaced to M2
N 2 N2 Interfaces with winding pinion
N 3 N3 strike actuating cam
O 1 O1 strike side winding pinion
P shaft hosts S and T levers
Q lever driven by N3 to strike bell
R bell / chime
S rack hook (lever) sits on L1
T warning lever raised by cam B3
V. SUMMARY
A tower clock vividly illustrates fundamental engineering
and physics concepts within a historical societal mechanism.
Mumford [15] has stated that “... the clock was the most
influential of machines, mechanically as well as socially”.
The conversion of potential energy by the time and strike
trains into multiple dial displays with accompanying hourly
bell strikes offers an opportunity to motivate student
learning. In this paper, the operation of a weight driven Seth
Thomas tower clock with Graham deadbeat escapement has
been discussed and mathematically modeled. Representative
numerical and animation results have been presented to offer Figure 5: Tower clock escape wheel angular displacement
insight into the dynamic characteristics. The next task is the motion – (a) experimental, and (b) simulation
refinement of the escapement model to better describe the
governing mechanical dynamics.
REFERENCES
[1] J. E. Barnett, Time’s Pendulum, Harcourt Brace & Company: New
Table 3: Summary of clock tower parameter values York, NY, 1998.
Sym Value Units Sym Value Units [2] D. S. Landes, Revolution in Time, Harvard University Press:
BD2 0.25 Ns/rad rA 10 cm Cambridge, MA, 1983.
Bu 56 Ns/rad r1 16.5 cm [3] D. Boorstin, The Discoverers - A History of Man’s Search to Know
dp 2.5 cm Rh 12 - His World and Himself, Random House: New York, NY, 1983.
hp 36.2 cm Rm 60 - [4] F. Shelley, Early American Tower Clocks, NAWCC: Columbia, PA,
JD2 7.75e-04 kg*m2 Rs 1 - 1999.
JPV 293.3 kg*m2 Rt 360 - [5] J. Lienhard, The Engines of Our Ingenuity, Oxford University Press:
New York, NY, 2003.
L 243.8 cm t0 1.0e-03 s
[6] L. Penman, Practical Clock Escapements, ClockWorks Press
M 62.3 kg T 2.96 s
International: West Sacramento, CA, 1998.
mA 45.4 kg TA1 1,127 Nm
[7] D. Bernstein, “Feedback Control: An Invisible Thread in the History
m1 61.2 kg ρwood 900 kg/m3 of Technology”, IEEE Control Systems Magazine, vol. 22, no. 2, pp.
m2 1.1 kg δ 1.134 rad 53-68, April 2002.
p 10.25 cm ζ1 -2.6e-02 rad [8] D. de Carle, Watch & Clock Encyclopedia, Crown Publishers: New
R 201.3 cm ζ2 8.7e-03 rad York, NY, 1977.
r 216.1 cm Δζ3 6.1e-03 rad [9] A. L. Rawlings, Science of Clocks and Watches, Pitman Publishing
Corporation: New York, NY, 1948.
638
Authorized licensed use limited to: University of Split. Downloaded on August 07,2023 at 10:17:37 UTC from IEEE Xplore. Restrictions apply.
[10] A. Lepschy, G. Mian, and U. Viaro, “Feedback Control in Ancient m1 pendulum bob mass (kg)
Water and Mechanical Clocks”, IEEE Transactions on Education, vol.
m2 mass of pendulum rod (kg)
35, no. 1, pp. 3-10, 1992.
[11] M. Kesteven, “On the Mathematical Theory of Clock Escapements”, N number of gear teeth
American Journal of Physics, vol. 46, no. 2, pp. 125-129, Feb 1978. p verge pivot to pallet distance (cm)
[12] J. Wagner, C. Huey, and K. Knaub, “Clock Mechanism Fundamentals r radius (mm)
for Education – Modeling and Analysis”, proceedings of ASME R pendulum pivot P to bob top length (cm)
DSCC, Ann Arbor, MI, October 2008.
[13] M. V. Headrick, “Origin and Evolution of the Anchor Clock r distance from pendulum CG to pivot (cm)
Escapement”, IEEE Control System Magazine, vol. 22, no. 2, pp. 41- rA radius of drum barrel (cm)
52, April 2002. rD2 escape wheel radius (cm)
[14] G. Baker and J. Blackburn, The Pendulum - A Case Study in Physics, r1 pendulum bob radius (cm)
Oxford University Press: New York, NY, 2005.
[15] L. Mumford, Technics and Civilization, Routledge & Kegan Paul: Rh hour hand gear train ratio
London, England, 1934. Rm minute hand gear train ratio
Rs strike gear train ratio
Rt time gear train ratio
t time (s)
t0 impulse time period (s)
T period (s)
T’ applied torque (Nm)
TA1 weight driven torque acting on great wheel (N)
TD2 escape wheel toque (N)
3
ρp density of pendulum rod material (g/cm )
δ escape wheel to verge angle (rad)
ζ i angular limits for impulse, hold, spin events (rad)
θ pendulum rotational angle (rad)
Δt integration time step (s)
ω angular speed (rad/s)
ωA1 great wheel angular speed (rad/s)
ωD2 escape wheel angular speed (rad/s)
ωh hour hand angular speed (rad/s)
ωm minute hand angular speed (rad/s)
639
Authorized licensed use limited to: University of Split. Downloaded on August 07,2023 at 10:17:37 UTC from IEEE Xplore. Restrictions apply.