A detached pendulum escapement is a clock escapement in which the pendulum swings freely through most of its arc and only touches the gear train for a brief impulse near dead-centre. The best examples — Riefler and Shortt-Synchronome — held rates within 0.01 second per day under controlled conditions. The point of detaching the pendulum is to keep external interference out of the swing so isochronism stays intact. You see this design in observatory regulators, time-service master clocks, and high-grade astronomical pendulums.
Detached Pendulum Escapement Interactive Calculator
Vary the target beat, local gravity, and engagement window to size the seconds pendulum and see where the brief impulse occurs.
Equation Used
The calculator uses the small-angle pendulum relation for a target beat: a 1 second beat has a 2 second full period, so L = gB^2/pi^2. The engagement angle defines the narrow detached escapement contact window around dead-centre.
FIRGELLI Automations - Interactive Mechanism Calculators.
- Small-angle pendulum approximation.
- Beat time is one half of the full pendulum period.
- Impulse engagement is symmetric about dead-centre.
How the Detached Pendulum Escapement Actually Works
The pendulum carries the time, the escape wheel carries the energy, and the detached escapement keeps those two jobs separated for as long as possible. A normal anchor or deadbeat escapement is in mechanical contact with the pendulum across most of the swing, which means every tooth-drop, every pivot friction change, every variation in drive weight bleeds straight into the timekeeping. A detached escapement only engages near the bottom of the arc — typically within ±2° to ±3° of dead-centre — where the pendulum is moving fastest and is least sensitive to small force perturbations. Outside that narrow window the pendulum is, mechanically speaking, alone.
The impulse itself is delivered through an impulse pallet on the pendulum rod or crutch. In a Riefler escapement the impulse comes through the suspension spring — there is no crutch contact at all. In a Shortt-Synchronome system the impulse pallet is released by a gravity arm every 30 seconds and the master pendulum is otherwise completely free. Q factor on these systems sits in the 10,000 to 25,000 range, compared with 2,000 to 6,000 for a good deadbeat regulator, and that Q ratio is exactly why circular error and barometric effects start to dominate the error budget instead of friction.
Get the impulse geometry wrong and the whole point of the design collapses. If the impulse pallet engages too early — say at 4° before dead-centre instead of 1° — the pendulum is being pushed during a part of the swing where the tangential force does not align with the velocity, and you get rate variation with amplitude. If the release timing drifts because of pivot wear in the gravity arm, the pendulum receives an impulse of varying duration and the rate walks. The classic failure on a Riefler is fatigue cracking of the suspension spring at the impulse point, which shows up as a sudden rate jump of 0.1 to 0.5 seconds per day with no other obvious cause.
Key Components
- Pendulum rod and bob: Carries the time. In an observatory regulator the bob is typically 6 to 10 kg of brass or invar with a temperature coefficient under 1 ppm/°C. Rod length is set for a 1-second beat at 0.994 m for a seconds pendulum at 45° latitude.
- Suspension spring: A thin steel ribbon, usually 0.08 to 0.12 mm thick, that flexes at the top of the pendulum. In a Riefler escapement this spring also delivers the impulse, so its dimensions and clamping must be repeatable to within 0.01 mm or rate shifts.
- Impulse pallet: Hardened steel or jewelled face that receives the brief drive impulse from the escape wheel or gravity arm. Contact dwell is typically 4 to 8% of the full period, centred on dead-centre to within ±0.5°.
- Escape wheel: Drives the impulse pallet for the brief contact window. In a detached design it has 30 teeth on a 1-second pendulum running at one tooth per second; tooth-tip geometry is held to ±0.005 mm to keep impulse energy uniform.
- Locking detent or gravity arm: Holds the train still between impulses and releases it cleanly at the right moment. In a Shortt-Synchronome the gravity arm drops every 30 seconds; in a Riefler the locking is integrated with the suspension assembly.
- Remontoire (where fitted): Re-energises the immediate drive every 30 seconds or so, isolating the escapement from main-train torque variation. Drops main-spring or weight-train influence on rate from parts-per-thousand to parts-per-million.
Who Uses the Detached Pendulum Escapement
Detached pendulum escapements live where rate stability matters more than mechanical simplicity. You will not find one in a mantelpiece clock — the cost and adjustment effort only pays back when you need observatory-grade timekeeping or scientific reference output. The named installations below are all real and most are still operable.
- Astronomical observatories: Riefler regulator Type B at the Hamburg-Bergedorf Observatory, used as the time standard from 1908 until quartz replaced it in the 1940s — held rate within 0.01 s/day in a sealed pressure vessel.
- National time services: Shortt-Synchronome free-pendulum clocks at the Royal Greenwich Observatory and the US Naval Observatory, providing the national time signal until the late 1940s.
- Geophysics and seismology: Strasser & Rohde precision regulators used as reference clocks alongside strainmeters and tide gauges at the Potsdam Geodetic Institute.
- Horological education: WOSTEP and BHI advanced courses use working Riefler movements to teach students how a detached escapement preserves Q against drive-train noise.
- Private high-end horology: The Erwin Sattler Astrolabium and similar contemporary regulator clocks use a detached gravity escapement to advertise observatory-grade rate performance to collectors.
- Museum and conservation work: The Deutsches Museum in Munich displays a working Shortt-Synchronome No. 11 with both master and slave pendulums, demonstrating impulse decoupling in real time.
The Formula Behind the Detached Pendulum Escapement
What you actually care about on a detached escapement is how much the rate moves when the drive amplitude changes — because that is what tells you whether detaching the pendulum is buying you anything. Circular error is the dominant amplitude-dependent rate term once you have removed escapement interference, and it sets the floor on how tight the rate can stay. At the low end of the typical operating amplitude (around 1°) the rate is almost amplitude-insensitive but the Q is fragile and a small bump can stop the clock. At the high end (around 4°) circular error swamps the gains from detachment and you may as well run a deadbeat. The sweet spot for a Riefler-type regulator sits at 1.5° to 2°.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| ΔT/T | Fractional rate change relative to small-amplitude period | dimensionless (s/s) | dimensionless (s/s) |
| θ0 | Half-amplitude of pendulum swing | radians | radians |
| T | Small-amplitude period of the pendulum | s | s |
| ΔT | Period change due to circular error | s | s |
Worked Example: Detached Pendulum Escapement in a national metrology lab reference regulator
A national metrology lab in Ottawa is rebuilding a 1925 Riefler Type C regulator as a backup reference for a caesium-comparison rig. The pendulum runs at a 1-second beat (T = 2 s full period) and you need to predict the daily rate sensitivity to amplitude drift across the operating envelope of 1°, 2°, and 3° half-amplitude so you can set the alarm thresholds on the optical amplitude monitor.
Given
- T = 2 s (full period, 1-second pendulum)
- θ0,low = 1 ° (0.01745 rad)
- θ0,nom = 2 ° (0.03491 rad)
- θ0,high = 3 ° (0.05236 rad)
- Seconds per day = 86400 s
Solution
Step 1 — at nominal 2° half-amplitude, convert to radians and apply the circular-error formula:
Step 2 — convert that fractional rate to seconds per day at nominal:
That figure is the absolute offset from infinitesimal-amplitude period. The number that actually matters in the lab is the sensitivity to amplitude drift — i.e. the partial derivative — because the regulator is rated against itself, not against a zero-amplitude ideal. Differentiating gives Δrate/Δθ0 ≈ θ0 / 8 per unit radian.
Step 3 — at the low end of the operating range, 1° half-amplitude:
Sensitivity at 1° is roughly 0.19 s/day per 0.1° of amplitude drift. The pendulum is barely fighting circular error at all here, but Q is precarious and a stray air current can park it.
Step 4 — at the high end of the operating range, 3° half-amplitude:
Sensitivity at 3° climbs to roughly 0.57 s/day per 0.1° of amplitude drift — three times worse than at 1°. At this amplitude the detached-escapement advantage is largely thrown away because circular error now dominates the residual rate noise.
Result
At the nominal 2° half-amplitude the regulator carries a 6. 58 s/day circular-error offset and an amplitude sensitivity of about 0.38 s/day per 0.1° drift. At 1° the sensitivity drops to 0.19 s/day per 0.1° but the pendulum is fragile against air currents and seismic micro-noise; at 3° sensitivity rises to 0.57 s/day per 0.1° and the detached-escapement Q advantage starts to disappear into amplitude noise — the design sweet spot for this Type C is 1.5° to 2°. If your measured rate sensitivity is significantly worse than 0.4 s/day per 0.1° at nominal, the most likely causes are: (1) a barometric-compensation bellows that has lost its seal, letting pressure swings of 20 hPa modulate the effective gravity term, (2) suspension-spring clamp slip allowing the effective rod length to change with amplitude, or (3) impulse-pallet engagement drifting outside ±0.5° of dead-centre because the gravity-arm pivot has worn, which couples amplitude into impulse energy.
Choosing the Detached Pendulum Escapement: Pros and Cons
Detached escapements buy rate stability at the cost of complexity, adjustment difficulty, and parts count. Whether that trade pays off depends entirely on what you are measuring against. Comparing across the realistic alternatives for a precision pendulum clock makes the decision obvious for most builds.
| Property | Detached pendulum escapement (Riefler/Shortt) | Graham deadbeat | Pin-wheel escapement |
|---|---|---|---|
| Rate stability (best case, controlled environment) | ±0.01 s/day | ±0.1 to ±0.5 s/day | ±0.3 to ±1 s/day |
| Pendulum Q factor achievable | 10,000 to 25,000 | 2,000 to 6,000 | 1,500 to 4,000 |
| Impulse contact fraction of period | 4 to 8% | 30 to 50% | 30 to 50% |
| Adjustment skill required | Master horologist, week-long setup | Skilled clockmaker, hours | Skilled clockmaker, hours |
| Typical service interval | 10 to 15 years | 5 to 10 years | 3 to 8 years |
| Parts count in escapement assembly | 20 to 40 (with remontoire) | 6 to 10 | 8 to 12 |
| Sensitivity to drive-train torque variation | Negligible (decoupled) | Direct (1:1 in rate) | Direct (1:1 in rate) |
| Realistic application fit | Observatory, time service, metrology | Regulator clocks, fine longcase | Tower clocks, cathedral clocks |
Frequently Asked Questions About Detached Pendulum Escapement
The bell jar changes the air pressure and density inside the pendulum case, and a Riefler is sensitive to both. Rate change scales with air density at roughly 0.4 s/day per kPa for an uncompensated pendulum. If the gasket is sealing at a different pressure than before — even by 5 kPa — you will see exactly the kind of shift you are describing.
Check the barometric compensation bellows first. If the bellows has gone stiff or developed a pinhole, the rate will track ambient barometer instead of being held constant. The diagnostic is to log rate against a barometer for 48 hours; if the correlation coefficient is above 0.5 the compensation is no longer working.
Riefler delivers tighter raw rate stability — sub-0.05 s/day is realistic — because the impulse comes through the suspension spring with no mechanical crutch contact. Strasser-style gravity escapements peak around 0.1 to 0.2 s/day but are far easier to adjust and do not depend on a bespoke suspension spring that, if it fatigues, takes weeks to replace.
The decision rule we use: if the clock is competing against an atomic reference and the rate must be characterised at the 0.01 s/day level, specify Riefler. If the clock is a museum-grade or high-end private regulator where 0.1 s/day is genuinely good enough, specify Strasser. The maintenance burden of a Riefler is roughly 3× a Strasser over a 20-year horizon.
Slave-pendulum lock does not protect the master from its own environment. The most common cause of unexplained master jumps is suspension-spring creep at the clamp — micro-slip of the steel ribbon under the brass clamp jaw shifts the effective pendulum length by a few microns, which is enough to move rate by 0.05 s/day.
Re-torque the suspension clamp to the original spec (typically 0.4 to 0.6 Nm depending on the model) and check that the clamp jaws have not work-hardened. If the clamp faces show witness marks or polishing, the spring is moving and you need to either re-clamp on fresh ribbon or, in a serious case, replace the suspension assembly.
Below about 10% of the half-amplitude — so for a pendulum swinging at 2°, the impulse should be delivered within ±0.2° of dead-centre. Above that, you are running what amounts to a low-friction deadbeat. The whole point of detachment is that the impulse occurs where the pendulum velocity is maximum and the tangential coupling between drive and timing is at its lowest.
If the engagement window has crept out to ±1° or worse — which happens with worn pivots on the gravity arm — you will measure rate sensitivity to drive weight that should not be there. Hang an extra 5% drive weight and watch the rate. If it shifts by more than 0.05 s/day, the impulse window is too wide.
At that amplitude in a stable vault you are well below the circular-error floor and below the temperature-sensitivity floor for an invar rod. The most likely limit is air-pressure noise. Even in a vault, atmospheric pressure swings 5 to 15 hPa across a weather front, and that translates into roughly 0.2 to 0.6 s/day on an uncompensated pendulum.
Either fit a barometric compensator, or — cheaper for a one-off — house the pendulum in a sealed enclosure pumped down to 30 to 50 hPa. The Riefler Type B did exactly this and dropped its rate noise by an order of magnitude. If pressure is already controlled, look at micro-seismic noise from the building; a metrology vault with a poured-on-bedrock plinth behaves very differently from one on a suspended slab.
You can do better than the originals, but not by as much as you might expect. A modern build with a fused-silica or carbon-fibre rod, a vacuum chamber held below 1 hPa, and active temperature control to ±0.01°C can reach 0.001 to 0.005 s/day — Bert Kelvin's Clock B and the Littlemore Clock are documented examples.
The reason it is not orders of magnitude better is that at that level you are fighting tidal gravity variations, ground tilt, and the gravitational pull of nearby moving masses. The detached escapement was already good enough that the limiting physics moved outside the clock. If you are getting 0.05 s/day on a modern build with a Riefler-style escapement, you are within a factor of 10 of the practical ceiling and further gains require attacking the environment, not the mechanism.
References & Further Reading
- Wikipedia contributors. Riefler escapement. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
