The Harrison winding device is a maintaining-power mechanism that keeps a clock or marine chronometer running at full torque during the few seconds you spend winding it. It solves the problem of the going train losing drive — and therefore time — the moment the mainspring or weight is disconnected. A small auxiliary spring inside the fusee or great wheel takes over driving the train through a sprung pawl-and-ratchet pair while winding. John Harrison patented the principle in 1722, and it became standard in every precision regulator and chronometer that followed.
Harrison Winding Device Interactive Calculator
Vary the normalized train torque, maintaining-spring torque window, and hold-time window to see the torque bridge during winding.
Equation Used
This calculator follows the article diagram rule that the maintaining-power spring should supply about 70 to 90 percent of normal train torque for about 30 to 60 seconds while winding. The train torque input is normalized, so the output torque is shown as percent of rated running torque.
- Train torque is normalized to the rated running torque.
- Maintaining spring torque is treated as a steady percentage of train torque during winding.
- Hold time is the useful interval before the main drive re-engages.
Operating Principle of the Harrison Winding Device
Wind a longcase regulator without maintaining power and the going train reverses for a moment as the weight line lifts off the great wheel. The escapement stops, the pendulum loses its impulse, and you have just put a measurable error into a clock that is meant to hold a few seconds a month. Harrison's fix was elegant — a small spring, pre-loaded against a ratchet, that bridges the brief torque gap and keeps the train moving forward.
The mechanism sits between the fusee (or great wheel) and the centre arbor. During normal running, the mainspring or driving weight loads the fusee, which loads the maintaining-power spring against its detent click. The spring is wound to a fixed pre-tension — typically enough to deliver around 70 to 90 percent of normal train torque for 30 to 60 seconds. When you start winding, the driving torque on the fusee disappears or reverses, but the pre-loaded auxiliary spring releases against the click and continues to drive the centre arbor forward. As soon as you stop winding, the main drive re-engages, the auxiliary spring is silently re-tensioned by the recovering torque, and the click resets.
Get the pre-load wrong and the mechanism fails in obvious ways. Too weak and the train still hesitates during winding — you'll see the seconds hand stutter or the pendulum amplitude drop a degree or two. Too strong and the spring acts as a brake during normal running, robbing the escapement of impulse. The click geometry has to be exact too — if the click face angle drifts past about 12° from radial, the click can disengage under shock and the maintaining spring fires uncontrolled, jamming the fusee chain. That is the most common failure mode in neglected 18th-century examples.
Key Components
- Maintaining-power spring: A flat or coiled steel spring pre-tensioned against the great wheel or fusee. Sized to deliver 70-90% of running torque for 30-60 seconds — long enough to cover a normal winding cycle. Spring rate must match the train's torque requirement at the escapement within ±10%.
- Detent click (pawl): A pivoted hardened steel pawl that locks the maintaining spring against the great wheel during running. Click face is cut between 8° and 12° off radial — too steep and it slips under shock, too shallow and it binds during reset.
- Ratchet wheel: Cut directly into the great wheel rim or as a separate disc on the same arbor. Tooth count is typically 30-60, with tooth pitch chosen so the click engages within one-half tooth of any rest position. Tooth tip hardness 58-62 HRC to resist click impact wear.
- Fusee or great wheel arbor: Carries both the driving torque and the maintaining-power assembly. Bore tolerance against the maintaining-power disc must be a slip fit of about 0.02 mm clearance — looser and the disc cocks under spring load, tighter and you get drag during reset.
- Reset stop pin: A small steel pin that limits the angular travel of the maintaining-power spring during recovery. Sets the pre-load when the train is fully driven. Position is set during assembly to within ±2° of the design angle.
Where the Harrison Winding Device Is Used
Harrison's device is one of those rare horological inventions that solved a single problem so completely that it never needed redesign. You'll find it on every precision pendulum regulator, every marine chronometer, and most fusee-driven bracket clocks where time accuracy mattered. Where you don't find it is in throwaway movements — domestic 8-day clocks of the 19th century often skipped it to save cost, which is why those clocks lose noticeable time on winding day.
- Marine chronometry: Standard fitment on every Kew-rated marine chronometer from 1770 onward, including Earnshaw, Arnold, and later Mercer chronometers used in Royal Navy navigation.
- Astronomical regulators: Fitted to Shelton transit-clock regulators used at the Royal Greenwich Observatory and to Riefler precision regulators in national observatories.
- Longcase precision clocks: Used in Thomas Tompion, George Graham, and John Roger Arnold longcase regulators where 1-second-per-week accuracy was the target.
- Bracket and table clocks: Fitted to high-grade Thwaites & Reed and Vulliamy fusee bracket clocks intended for scientific or naval-officer use.
- Tower and turret clocks: Adapted in larger form for E. Dent & Co. and J.B. Joyce turret movements where winding takes minutes rather than seconds — the auxiliary weight has to carry the train through a much longer disengagement.
- Modern reproduction horology: Specified in Sinclair Harding and Buchanan modern fusee clocks, and in kit-form reproductions of Harrison's H1-H4 sea clocks built by enthusiasts under the British Horological Institute.
The Formula Behind the Harrison Winding Device
What you actually want to know as a clockmaker is how long the maintaining spring will hold the train at running torque. That tells you whether the mechanism will cover your winding time with margin or fail short. The relevant formula relates spring pre-load energy to escapement power consumption — at the low end of the typical range (a quick-wound chronometer fusee) you only need 10-15 seconds of cover; at the high end (a turret clock where winding takes a minute or more) you need 60+ seconds and usually a separate auxiliary weight rather than a spring. The sweet spot for domestic and observatory work sits around 30-45 seconds of hold time at 80% running torque.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| thold | Time the maintaining spring sustains the train at design torque | s | s |
| k | Angular spring rate of the maintaining-power spring | N·m/rad | lbf·in/rad |
| θ | Pre-load angle of the maintaining spring against the click | rad | rad |
| Tesc | Torque required at the great-wheel arbor to drive the escapement | N·m | lbf·in |
| ωtrain | Angular velocity of the great-wheel arbor under normal running | rad/s | rad/s |
Worked Example: Harrison Winding Device in an 1850s Dent marine chronometer overhaul
A specialist horologist in Portsmouth is recommissioning an 1850s E. Dent & Co. two-day marine chronometer with a spring-detent escapement and a chain fusee. The maintaining-power spring needs verification — the previous restorer fitted a replacement spring of unknown rate and the chronometer drops 2 seconds on winding day. The horologist measures spring rate k = 0.018 N·m/rad, pre-load angle θ = 0.42 rad (about 24°), escapement torque demand Tesc = 6.5 × 10-5 N·m at the great-wheel arbor, and great-wheel angular velocity ωtrain = 1.75 × 10-3 rad/s.
Given
- k = 0.018 N·m/rad
- θ = 0.42 rad
- Tesc = 6.5 × 10-5 N·m
- ωtrain = 1.75 × 10-3 rad/s
Solution
Step 1 — compute stored energy in the maintaining spring at nominal pre-load (θ = 0.42 rad):
Step 2 — compute escapement power consumption at the great-wheel arbor:
Step 3 — divide to get nominal hold time:
That raw figure is misleading — only the energy released between full pre-load and the click reset position counts as useful hold. Empirically that working fraction is about 0.3% of stored energy for a typical Earnshaw-style maintaining spring, giving a usable thold,nom ≈ 42 s. That is a comfortable margin for a fusee that takes 12-15 seconds to wind by hand.
Step 4 — at the low end of the typical pre-load range (θ = 0.30 rad, around 17°):
21 seconds is uncomfortably close to winding time. You'd see the chronometer survive a brisk winding but stutter if anything interrupted you mid-wind. At the high end (θ = 0.55 rad, around 32°):
72 seconds gives huge margin but the spring now contributes measurable drag during normal running — you'd see escapement amplitude drop by 1-2° and rate would stiffen up in cold weather as the spring rate climbs.
Result
Nominal hold time is approximately 42 seconds at 80% running torque — comfortable cover for a 12-15 second fusee wind. The low-end 21 seconds at 17° pre-load barely covers a fast wind and would explain the 2-second daily loss the horologist measured; the high-end 72 seconds at 32° pre-load over-tensions the train and would show as reduced escapement amplitude. If the chronometer still drops time on winding day after correcting pre-load, check three failure modes the calculation cannot capture: (1) a worn click pivot letting the click chatter under spring load, releasing energy unevenly, (2) fusee chain stiction at the great-wheel hook causing the maintaining spring to stall against friction rather than driving the train, or (3) a cracked or annealed maintaining spring whose effective k has dropped 20-30% from the measured cold value once it warms in the case.
When to Use a Harrison Winding Device and When Not To
Harrison's spring-based maintaining power is not the only way to keep a precision clock running during winding. The choice depends on winding duration, available space, and how much you trust the operator. Here is how the three practical options compare on the dimensions that actually matter when you specify a movement.
| Property | Harrison maintaining power (spring) | Bolt-and-shutter maintaining power | Endless-chain (Huygens) drive |
|---|---|---|---|
| Hold time at running torque | 30-60 s typical, up to 90 s | 10-30 s typical | Continuous — no interruption |
| Torque delivered vs running | 70-90% of running torque | 60-80% of running torque | 100% (no transition) |
| Cost / build complexity | Moderate — extra spring, click, ratchet on great wheel | Low — just a shutter and bolt over the keyhole | Low — but only works on weight-driven clocks with two weights |
| Failure mode if neglected | Spring fatigues, click face wears, train stutters on wind | Operator forgets to engage shutter, no protection at all | Pulley wears, rope stretches, weights bind |
| Application fit | Marine chronometers, fusee bracket clocks, longcase regulators | Early longcase clocks before 1750, simpler bracket clocks | Weight-driven turret clocks, some longcase regulators |
| Typical accuracy impact on winding day | <0.1 s loss per wind | 0.5-2 s loss per wind | Zero loss |
| Adjustability after assembly | Pre-load angle adjustable via stop pin position | Fixed by shutter geometry | Not adjustable — geometric |
Frequently Asked Questions About Harrison Winding Device
Wind the clock fully and watch the maintaining-power click. The click should sit firmly against the next ratchet tooth with no audible tick or wobble — that confirms the spring is loaded against the stop pin. Then with the train running, gently push the click off its tooth with a peg-wood stick. The train should continue at full speed for the spring's design hold time before any visible slowdown. If the train stutters within 5-10 seconds you have not enough pre-load; if you have to fight the click off the tooth, you have too much. Adjust the stop pin position 1-2° at a time and retest.
Most likely the issue is detent-escapement-specific rather than maintaining-power-related. Spring-detent escapements unlock only on alternate balance swings, so any brief torque dip during winding tends to drop one impulse rather than stutter visibly. Check that your maintaining spring releases before the fusee click disengages during winding — there is meant to be 5-10° of overlap. If the maintaining spring takes over after the fusee has already started moving, you get a torque gap of a fraction of a second that costs you exactly one impulse, which on a half-second beat chronometer is the half-second you're losing.
Technically yes, practically rarely worth it. The going-barrel arbor needs to carry an additional ratchet and click assembly, which means re-machining the great wheel or fitting a sandwich plate. On most domestic 8-day movements there isn't enough axial space between the plates for the extra disc, and you'd need to reduce wheel thickness which compromises the train. Better options are either fitting a bolt-and-shutter (which only requires modifying the dial) or accepting the 1-2 second winding-day loss if the clock isn't a precision regulator.
Measure cold and warm. With the movement at workshop temperature (around 20°C), pull the click off and time the hold. Then warm the movement to about 25-28°C with a hand-held warm-air source for 15 minutes and retest. A healthy spring will show less than 5% change in hold time. A fatigued or annealed spring shows 15-30% change because its rate is now temperature-sensitive in a way the original heat-treated steel was not. If you see that pattern, replace the spring — you cannot reliably re-temper a maintaining spring in situ without distorting the geometry.
No, and this is the most common mistake when restorers fit a replacement of unknown rate. A spring that's 20% over-rated acts as a continuous brake on the going train. You'll see escapement amplitude drop 2-4°, rate becomes position-sensitive because the extra drag amplifies pivot friction differences, and the clock will stiffen further in cold weather as steel spring rate climbs. The original makers sized maintaining springs to a fairly tight window — typically 75-85% of running torque — and going outside that window in either direction degrades performance.
Aim for 36-48 teeth on the maintaining ratchet. Fewer than 30 and the click can rest up to 6° off the spring's loaded position, which means a measurable lost angle when the maintaining spring takes over. More than 60 and the teeth become too fine for reliable click engagement under shock — you'll see the click skip teeth during winding, especially if the fusee chain is even slightly sticky. 40 teeth is the figure used on most surviving Graham and Arnold regulators and it remains the right answer today.
References & Further Reading
- Wikipedia contributors. Maintaining power. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
