A Pendulum Escapement is the timekeeping interface between a clock's swinging pendulum and its toothed escape wheel, where two pallets alternately catch and release the wheel once per pendulum swing. It solves the core horological problem of converting a continuous driving torque into discrete, isochronous time intervals. On each swing, one pallet locks the wheel, then releases it to deliver an impulse to the pendulum that replaces lost energy. The result is rate accuracy of a few seconds per day in a domestic longcase, and under 0.1 second per day in a precision regulator like a Riefler.
Pendulum Escapement Interactive Calculator
Vary pendulum length, swing amplitude, gravity, and escape-wheel tooth count to see period, rate drift, and wheel stepping.
Equation Used
The calculator evaluates the article period relation for a pendulum escapement. The amplitude slider is entered in degrees, converted to radians as theta0, and applied as the finite-amplitude correction. Daily drift compares the computed half-period T/2 with an ideal 1.000 second beat.
- theta0 is the pendulum half-amplitude converted from degrees to radians.
- First-order finite-amplitude correction for a simple pendulum is used.
- Escapement impulse sustains motion but does not set the natural rate.
- Escape wheel advances one tooth per full pendulum period for the wheel-speed estimate.
Inside the Pendulum Escapement
The Pendulum Escapement, also called the Pendulum escapement (form) in horological literature, works by gating an escape wheel one tooth at a time using two pallet faces mounted on a crutch-driven anchor. The pendulum swings, the anchor rocks, and on each half-swing one pallet drops off its tooth, the wheel rotates by half a tooth-pitch, and the next pallet catches the following tooth. That catch is what you hear as the tick. During the unlocking phase, the wheel pushes the pallet face along its impulse plane, transferring just enough energy from the going train to overcome pivot friction and air drag — typically 1 to 5 millijoules per swing in a domestic longcase.
The geometry is unforgiving. Pallet span, drop, lock, and impulse angle all have to agree with the wheel's tooth count and the pendulum's amplitude. If lock is too shallow — under about 1.5° on a typical anchor — the wheel can trip and run away under load spikes. Too deep, beyond 3°, and you waste impulse energy unlocking the wheel, which kills amplitude and pushes the pendulum off its isochronous arc. On a deadbeat escapement (the Graham form), the locking face is concentric with the pallet pivot, so once a tooth lands on lock there is zero recoil — the wheel sits motionless until the pallet rocks back through the lock angle.
Get the tolerances wrong and you see specific symptoms. A pendulum running fast in dial-up but slow in dial-down points to an out-of-beat escapement, where one impulse is stronger than the other. A clock that stops after 6 hours but runs cleanly when freshly wound is almost always losing amplitude through pallet wear or thickening oil — the impulse can no longer overcome circular error at low amplitude. And if you hear an uneven tick-tock cadence, the crutch is delivering impulse asymmetrically and the escapement needs putting in beat.
Key Components
- Escape Wheel: The toothed wheel driven by the going train that delivers impulse to the pallets. Tooth count is typically 30 on a longcase 1-second pendulum so the wheel rotates once per minute. Tooth profile must be hardened steel polished to Ra ≤ 0.2 µm or you'll see pallet pitting within 5 years.
- Pallets (Entry and Exit): Two hardened faces — usually polished steel or jewelled — that alternately lock and receive impulse from the escape wheel teeth. The angle between them sets the pendulum's required amplitude. On a Graham deadbeat the locking face is a concentric arc; on a recoil anchor it is a flat plane that pushes the wheel backwards on lock.
- Anchor (Pallet Arbor): Carries the two pallets and rocks on its own pivots in line with the pendulum suspension. Pivot endshake should sit between 0.05 and 0.10 mm — too tight binds, too loose lets the pallets shift sideways and changes the drop.
- Crutch: Forked arm fixed to the anchor that engages the pendulum rod with side-play of around 0.1 mm. The crutch transfers impulse to the pendulum and is the part you bend left or right to put the clock in beat.
- Pendulum: The timekeeping element with a period set by its effective length. A 0.994 m pendulum gives a 1-second beat at standard gravity. The pendulum is what defines the rate; the escapement only sustains and counts.
- Suspension Spring: Thin flat spring (typically 0.10 mm spring steel) that supports the pendulum and acts as its frictionless pivot. Replace it if you see any kink — a buckled suspension changes the effective pendulum length and shifts rate by tens of seconds per day.
Who Uses the Pendulum Escapement
The Pendulum Escapement dominated precision timekeeping from 1657, when Christiaan Huygens applied a pendulum to a verge clock, until quartz took over in the 1970s. You still find it everywhere a mechanical second matters — turret clocks, observatory regulators, longcase clocks, and a surprising amount of new horological work. Different industries use slightly different names: a watchmaker calls the form an anchor or deadbeat, a museum curator may write it up as a Pendulum escapement (form) in the catalogue card, but the mechanism is the same.
- Public Timekeeping: The Great Clock at the Palace of Westminster (Big Ben) uses a double three-legged gravity escapement driving a 4-second pendulum — a derivative of the pendulum escapement family designed by Edmund Beckett Denison in 1854.
- Observatory Regulators: Riefler astronomical regulators built between 1890 and 1965 used a free pendulum escapement form and held rate to within 0.01 seconds per day, used at the US Naval Observatory and Pulkovo.
- Domestic Horology: Howard Miller and Hermle longcase clocks ship today with anchor or deadbeat pendulum escapements running 1-second pendulums on a 30-tooth escape wheel.
- Cathedral and Turret Clocks: Smith of Derby and J.B. Joyce of Whitchurch installations across UK cathedrals — many fitted with deadbeat or gravity pendulum escapements driving compound pendulums up to 2 metres long.
- Scientific Instrument Restoration: Charles Frodsham astronomical regulators rebuilt by conservators in Greenwich and Edinburgh — typically with jewelled deadbeat pallets and Invar pendulum rods to suppress thermal drift below 0.5 ppm per °C.
- Education and Demonstration: School and university physics demonstrators use exposed-movement pendulum clocks (Hermle 451 movements are common) to illustrate isochronism, energy transfer, and damped harmonic motion.
The Formula Behind the Pendulum Escapement
The fundamental relation governing a pendulum escapement is the pendulum period itself — the escapement only counts and sustains the swings, it does not set the rate. What matters in practice is how period varies with amplitude (circular error) and with effective length. At the low end of typical operating amplitudes — around 1.5° half-amplitude on a deadbeat regulator — circular error is negligible, under 0.1 ppm. At a nominal 3° amplitude on a domestic longcase, you pick up roughly 2 ppm of error. Push amplitude past 6° on a worn anchor escapement and circular error climbs above 8 ppm — that's nearly a second a day from amplitude variation alone. The sweet spot for a precision regulator is the smallest amplitude that still reliably overcomes pallet friction, usually 1.5° to 2°.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| T | Period of one full pendulum swing (out and back) | s | s |
| L | Effective pendulum length (suspension to centre of oscillation) | m | ft |
| g | Local gravitational acceleration | m/s² | ft/s² |
| θ0 | Pendulum half-amplitude (peak swing angle from vertical) | rad | rad |
Worked Example: Pendulum Escapement in a marine chronometer maker's reference regulator
A boutique marine chronometer manufacturer in La Chaux-de-Fonds is commissioning a workshop reference regulator with a Graham deadbeat escapement, a 30-tooth escape wheel, and an Invar pendulum cut for a 1-second beat at the local gravity of 9.8055 m/s². The owner wants to know how rate will shift across the typical operating amplitude range as the mainspring runs down between weekly winds, and where to set the nominal amplitude for best long-term stability.
Given
- L = 0.9936 m
- g = 9.8055 m/s²
- θ0 (nominal) = 3.0 °
- θ0 (low end) = 1.5 °
- θ0 (high end) = 6.0 °
Solution
Step 1 — compute the small-angle period (the limit as amplitude approaches zero):
That's the target — a clean 1-second beat with a 2-second full period. Now we need to see what circular error does at three operating points.
Step 2 — at nominal 3° half-amplitude, convert to radians and apply the correction term:
That's an excess of 17.1 ppm over the small-angle period, or about 1.48 seconds slow per day relative to T0. You compensate by shortening L slightly at setup — that's what the rating nut on the bob is for.
Step 3 — at the low end of typical operating amplitude, 1.5°:
That's 4.3 ppm — barely 0.37 seconds per day. Below 1.5° you start risking the pendulum failing to unlock cleanly, especially with cold thickened oil on the pallets.
Step 4 — at the high end, 6°:
That's 68.5 ppm — almost 6 seconds per day slower than the small-angle period. More importantly, the rate change from nominal-to-high is 51 ppm, meaning if your amplitude varies from 3° to 6° between fully wound and run-down, your clock drifts 4.5 seconds across the week from circular error alone.
Result
Nominal period at 3° amplitude is 2. 0000343 seconds, against a small-angle target of 2.0000000 seconds. In practice that means once the rating nut is set for nominal amplitude, the clock keeps time to better than a second a day as long as amplitude stays within roughly ±0.5° of the rated value. At 1.5° amplitude the regulator gains about 1.1 seconds per day relative to the nominal setting; at 6° it loses about 4.5 seconds per day — so the cost of letting amplitude wander is significant and asymmetric. If your measured rate drifts more than ±2 seconds per day across a winding cycle, the most likely causes are: (1) pallet wear opening up the impulse angle and feeding excess amplitude when freshly wound, (2) thickened or migrated oil on the pallet faces, raising friction and dropping amplitude as the week goes on, or (3) suspension spring fatigue introducing a kink that shifts effective length by 0.05 mm or more — easily 50 seconds per day of error.
Pendulum Escapement vs Alternatives
The Pendulum Escapement is one of several form factors for translating a sustained driving torque into isochronous timekeeping. Its main competitors are the recoil verge (its predecessor) and the gravity escapement (its successor for tower clocks). Each has a defensible niche.
| Property | Pendulum Escapement (deadbeat) | Verge Escapement | Gravity Escapement |
|---|---|---|---|
| Typical rate accuracy (per day) | ±1-5 s domestic, ±0.01 s precision regulator | ±60-300 s | ±0.1-1 s on tower clocks |
| Pendulum amplitude required | 1.5°-4° half-amplitude | 30°-50° half-amplitude | Driven by gravity arms, ~2° pendulum |
| Sensitivity to driving torque variation | Moderate — affects amplitude and circular error | Severe — period changes directly with torque | Very low — pendulum is impulsed by independent gravity arms |
| Manufacturing complexity | Moderate — pallet geometry critical to ±0.05° | Low — simple crown wheel and verge | High — multi-arm gravity train and locking system |
| Service interval | 5-10 years for cleaning/oil | 2-5 years (high friction) | 10+ years, but locking faces wear |
| Best application | Domestic longcase, bracket, regulators | Pre-1670 antiques and replicas only | Turret and cathedral clocks with variable wind loading |
| Power consumption | 1-5 mJ per swing | 5-20 mJ per swing | Higher — gravity arms must be lifted each beat |
Frequently Asked Questions About Pendulum Escapement
Position-dependent rate on a pendulum clock is almost always a beat or suspension issue, not the escapement geometry itself. When you tilt the case, the pendulum's effective vertical changes relative to the crutch, and one pallet starts receiving more impulse than the other. The clock goes out of beat, amplitude on one side drops, and circular error becomes asymmetric.
Stand the clock back upright on a level surface, listen for an even tick-tock cadence (the gap between tick and tock should be identical), and bend the crutch a fraction of a millimetre toward the longer interval. A clock that needs to run in a non-vertical orientation should not be using a pendulum escapement at all — that's what balance-wheel watches are for.
Invar's coefficient of thermal expansion is roughly 1.2 ppm/°C versus brass at 19 ppm/°C, so the temperature compensation is doing its job — but the effective length of an Invar rod at room temperature isn't the same as the brass rod you removed, even if the physical lengths match. The bob mass distribution also shifts the centre of oscillation.
Re-rate the clock with the rating nut: 4 seconds per day fast on a 1-second pendulum needs the bob lowered by about 0.046 mm. Use the formula δT/T = ½ × δL/L and adjust accordingly. Don't trust that swapping rods preserves rate — always re-rate.
Deadbeat every time, unless you're explicitly building a period-correct pre-1715 reproduction. The Graham deadbeat eliminates recoil — when a tooth lands on the locking face, the wheel stops dead instead of momentarily reversing. Recoil costs amplitude, drives uneven wear into the wheel teeth, and makes the escapement more sensitive to torque variation across the winding cycle.
The only design penalty of deadbeat is that the locking face must be cut as a true concentric arc to the pallet pivot — within about 0.02 mm of true. Get a competent watchmaker to grind and polish the pallets, or buy them pre-finished from a supplier like Bergeon. The performance gain is worth the extra setup.
If the train is clean and the mainspring or weight is delivering full torque, suspect the suspension spring. A barely visible kink — caused by transport, by the pendulum being lifted off improperly, or by long storage with the bob hung — increases internal damping and raises the effective bending stiffness slightly. You'll see amplitude drop 0.5° to 1° over a few hours and the clock will eventually stop.
Replace the suspension spring as a matter of course on any restoration. They're cheap (a few pounds) and the rate consequence of getting it wrong is large. Use the correct thickness for the pendulum mass — too thick a spring shortens the effective length and runs the clock fast.
For a Graham deadbeat with steel pallets and properly thinned clock oil, around 1.2°-1.5° half-amplitude. Below that, the impulse delivered per swing falls below the energy needed to overcome pallet friction plus suspension damping plus air drag, and the clock will stop — usually overnight when temperature drops and oil viscosity rises.
Riefler and Shortt free-pendulum regulators run lower — under 1° — but they impulse the pendulum mechanically only every 30 seconds and use jewelled pallets with negligible friction. For a conventional one-swing-one-impulse domestic build, design for 2.5°-3° at full wind and you'll have margin to 1.5° at run-down.
Yes — they refer to the same family of mechanisms. "Pendulum escapement" is the functional name (it gates a pendulum), while "anchor escapement" describes the physical shape of the pallet arbor (anchor-shaped). Within that family you'll find the recoil anchor (Robert Hooke / William Clement, c.1670), the Graham deadbeat (1715), the pin-pallet escapement (Brocot), and various jewelled deadbeat variants. All are pendulum escapements; they differ in pallet geometry and how they handle the lock-and-impulse cycle.
Invar suppresses rod expansion, but it doesn't address the other temperature effects: air density changes the buoyancy and drag on the bob, the suspension spring's modulus shifts with temperature, and barometric pressure varies seasonally. Air-density effects alone account for roughly 0.4 seconds per day per °C on an uncompensated bob.
If you need better than ±2 seconds per day across seasons, you either need a barometrically-compensated pendulum (Riefler used a partial vacuum) or the regulator needs to live in a climate-controlled enclosure held to within ±1 °C and ±5 mbar. The escapement itself isn't the problem — the environment is.
References & Further Reading
- Wikipedia contributors. Escapement. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
