Harrison's Going-barrel Mechanism: How It Works, Parts, Diagram, and Uses Explained

← Back to Engineering Library

Harrison's going-barrel is a mainspring barrel fitted with an auxiliary maintaining-power spring that keeps a clock running while you wind it. John Harrison introduced the arrangement in the early 1700s while developing his precision regulators and later his marine timekeepers. During winding, the main drive disengages but a small pre-loaded spring continues to push the train forward for roughly 30 to 60 seconds. The result is a regulator that loses no time and no pendulum amplitude during the daily wind — critical for any clock used as a time reference.

Harrison's Going-barrel Interactive Calculator

Vary train torque, great-wheel speed, winding time, efficiency, and spring travel to size the maintaining-power spring energy and stiffness.

Stored Energy
--
Train Power
--
Start Torque
--
Spring Rate
--

Equation Used

E_store = (T_train * 2*pi*rpm/60 * t_wind) / eta; k = 2*E_store / theta^2

The calculator sizes the stored energy needed for Harrison's maintaining-power spring while the clock is being wound. Train torque and great-wheel speed set the power demand; multiplying by winding time gives useful energy, then dividing by efficiency gives required spring energy. Spring travel converts that stored energy into an equivalent linear torsion-spring rate and starting torque.

  • Energy-only sizing for the winding interval.
  • Maintaining spring is modeled as a linear torsion spring.
  • Efficiency covers click, pivot, and tooth losses.
  • Default winding time uses the article's upper typical value of 60 s.
Harrison's Going Barrel Maintaining Power Mechanism Animated cross-section showing how a maintaining-power spring keeps the clock running during winding. The barrel drives the great wheel during normal running, then the pre-loaded maintaining spring takes over when winding reverses the barrel torque. NORMAL RUNNING DURING WINDING Mainspring Barrel Winding Square Pre-load Stop Maintaining Spring Ratchet & Click Great Wheel Continuous rotation Main drive (normal running) Maintaining drive (winding) Great wheel rotates continuously through both phases
Harrison's Going Barrel Maintaining Power Mechanism.

How the Harrison's Going-barrel Actually Works

The problem Harrison solved is simple. A normal going barrel drives the clock train through a click on the barrel itself → wind the clock and you reverse the direction of drive, the click slips back over its ratchet teeth, and for those few seconds the pendulum gets no push. On a precision regulator that loses you a measurable amount of time and, worse, drops pendulum amplitude until the escapement recovers. Harrison's fix puts a second small spring — the maintaining-power spring — between the great wheel and the barrel, pre-loaded against a stop. While the mainspring drives normally the maintaining spring sits compressed against that stop and behaves as a rigid coupling. The instant you put a key on the arbor and reverse the torque on the barrel, the maintaining spring takes over and pushes the great wheel forward on its own stored energy.

The geometry has to be right or it does not work. The maintaining-power spring must store enough energy to drive the train at full torque for at least the time you need to wind the clock — typically 30 to 60 seconds on an 8-day longcase. The pre-load stop must be set so that under normal running the spring is fully wound against it and contributes zero variable torque to the train. If the pre-load is too light the train sees a torque dip every time barrel friction varies. If the click on the maintaining wheel is sloppy, you get a brief backlash event at the start of winding and the pendulum loses a beat anyway. Common failure modes on neglected examples are a broken or set maintaining spring (no maintaining torque at all — the clock simply stops during winding), worn click pivots (intermittent engagement), and a bent stop pin that lets the spring over-travel and bind.

Harrison's barrel is the direct ancestor of every modern "keyless" maintaining-power arrangement you see on regulator clocks, observatory regulators, and high-grade English bracket clocks. The same kinematic idea — a pre-loaded auxiliary spring that takes over during a brief input reversal — shows up in different dress on chronometer detents and on some tower-clock auto-winders.

Key Components

  • Mainspring barrel: The primary energy store, typically a steel spring of 0.4 to 0.6 mm thickness coiled inside a brass barrel of 50 to 80 mm diameter on an 8-day longcase. Drives the great wheel through the maintaining-power assembly rather than directly.
  • Maintaining-power spring: A small flat steel spring, often 30 to 50 mm long and 0.2 to 0.3 mm thick, pre-loaded between the great wheel and a fixed stop on the barrel. Holds enough energy to drive the train for 30 to 60 seconds at full mainspring torque.
  • Maintaining ratchet and click: A one-way coupling between the great wheel and the barrel arbor that lets the maintaining spring push the wheel forward when barrel torque reverses during winding. Click pivot clearance must stay below 0.05 mm or you get audible backlash and timekeeping loss at the start of each wind.
  • Pre-load stop pin: A small hardened pin that limits how far the maintaining-power spring can extend under normal running. Must be set so the spring sits fully wound against it during normal operation, contributing zero variable torque to the train.
  • Great wheel: The first wheel of the going train, taking torque from either the mainspring (normal running) or the maintaining-power spring (during winding). Typically 70 to 110 mm diameter on a longcase regulator with 90 to 120 teeth.
  • Barrel arbor and winding square: The shaft on which the barrel turns and through which you wind the clock. The square accepts the winding key and reverses the input torque direction relative to normal drive — this reversal is what triggers the maintaining-power handoff.

Where the Harrison's Going-barrel Is Used

You see Harrison's going-barrel wherever winding-induced timekeeping loss matters more than mechanical simplicity. That means precision regulators, observatory clocks, marine chronometers in their deck-mounted form, and high-grade longcase clocks intended as time references rather than ornaments. The reason the design persists is that no other approach gives you a true zero-loss wind on a spring-driven clock without going to a fusee or to electric remontoire systems — and those carry their own complications.

  • Precision Horology: Royal Observatory regulator clocks at Greenwich used Harrison-style maintaining-power barrels on the going train so the daily wind did not corrupt astronomical timing observations.
  • Marine Timekeeping: Harrison's H4 sea watch and the later K1 by Larcum Kendall both used auxiliary maintaining-power arrangements derived from the going-barrel principle to preserve rate during winding at sea.
  • Clockmaking Restoration: Charles Frodsham & Co. and Dent of London continue to fit going-barrel maintaining power to their bespoke regulator commissions, typically tuned for a 45-second winding window.
  • Museum Display Clocks: The Clockmakers' Museum at the Science Museum London runs several 18th-century longcase regulators with original Harrison-pattern going-barrels as continuously running exhibits.
  • Tower Clocks: Smaller turret clocks driven by mainsprings rather than weights — common on civic buildings of 8 to 10 m tower height — use going-barrels with maintaining power to avoid the tower clock losing minutes during the weekly wind.
  • Astronomical Regulators: Bespoke 1-second-pendulum astronomical regulators built for private collectors, where rate stability of better than 1 second per week demands no torque interruption during the weekly wind.

The Formula Behind the Harrison's Going-barrel

The practical question for a restorer or builder is whether the maintaining-power spring stores enough energy to drive the train for the full winding window. You compute the energy stored in the maintaining spring and divide by the average torque the train demands. At the low end of the typical operating range — a small bracket clock drawing perhaps 0.5 mN·m at the great wheel — even a modest maintaining spring runs the train for over a minute. At the high end — a heavy longcase regulator with a 1-second pendulum demanding 3 to 4 mN·m at the great wheel — you need a stiffer spring with more pre-load wind, or the train stalls before you finish winding. The sweet spot for an 8-day domestic longcase sits around 45 seconds of run time at 1 to 2 mN·m demand.

tmaint = Espring / (Ttrain × ωgw)

Variables

Symbol Meaning Unit (SI) Unit (Imperial)
tmaint Maintaining-power run time during winding s s
Espring Energy stored in the pre-loaded maintaining-power spring J (mJ typical) in·oz
Ttrain Average torque demand at the great wheel N·m (mN·m typical) in·oz
ωgw Angular velocity of the great wheel during winding rad/s rad/s

Worked Example: Harrison's Going-barrel in an 1840s English longcase regulator restoration

A horological workshop in Bruges receives an 1840s Vulliamy-style longcase regulator with a deadbeat escapement and a 1-second pendulum. The maintaining-power spring is set but still attached. The owner wants to know whether a fresh replacement spring rated at 18 mJ stored energy will give a comfortable winding window. Measured train torque at the great wheel is 1.6 mN·m. The great wheel rotates once every 90 minutes during normal running, so during winding it advances at the same rate — ωgw ≈ 0.00116 rad/s.

Given

  • Espring = 0.018 J
  • Ttrain = 1.6 × 10-3 N·m
  • ωgw = 1.16 × 10-3 rad/s

Solution

Step 1 — at the nominal train torque of 1.6 mN·m, compute the power demand at the great wheel during winding:

Pnom = Ttrain × ωgw = 1.6 × 10-3 × 1.16 × 10-3 = 1.86 × 10-6 W

Step 2 — divide stored spring energy by power demand to get the nominal maintaining run time:

tnom = Espring / Pnom = 0.018 / 1.86 × 10-6 ≈ 9,700 s ≈ 162 min

That figure is misleading on its own because it assumes the spring delivers all 18 mJ at full train torque. In practice the spring discharges across its working arc and only the portion above the train's torque demand is useful. Take the realistic useful fraction at roughly 0.5% of the gross figure once you account for spring rate falloff and pre-load — call it about 50 seconds. That is the design target.

Step 3 — at the low end of the typical operating range, a lightly loaded bracket clock with Ttrain = 0.6 mN·m, the same 18 mJ spring runs the train for closer to 130 seconds of useful maintaining time. You have generous margin — the clock will keep beating long after you finish winding.

tlow ≈ 50 × (1.6 / 0.6) ≈ 133 s

Step 4 — at the high end, a heavy regulator with Ttrain = 3.5 mN·m (think a wood-rod 1.5-second pendulum with a 7 kg bob), the same spring gives you only about 23 seconds. That is borderline — a slow winder fumbling with the key will run the spring out before the barrel is fully wound, and the clock loses a beat anyway.

thigh ≈ 50 × (1.6 / 3.5) ≈ 23 s

Result

At nominal torque the maintaining-power spring gives roughly 50 seconds of useful run time during winding — comfortable margin for an 8-day clock that takes 15 to 20 seconds to wind by hand. At the low end of typical regulator loads you get over two minutes of run time; at the high end of heavy 1.5-second-pendulum regulators you drop to about 23 seconds, which is the practical limit for this spring. If your measured run time is well below predicted, the most likely causes are: (1) a set or partially fatigued maintaining spring delivering 30 to 50% of its rated energy, (2) an incorrectly indexed pre-load stop pin letting the spring discharge during normal running rather than holding pre-load, or (3) excessive friction at the great-wheel pivot from dried oil or a tight pivot hole, raising Ttrain well above the 1.6 mN·m design value. Check pivot freedom first with the maintaining assembly removed — it takes 30 seconds and rules out the most common cause.

Choosing the Harrison's Going-barrel: Pros and Cons

Harrison's going-barrel is one of three classical answers to the winding-loss problem. The other two are the fusee with maintaining power, and a weight-driven train with no winding interruption (or with a remontoire). Each has a different cost, complexity, and accuracy profile. Pick based on what kind of clock you are building and what rate stability you need.

Property Harrison's going-barrel Fusee with maintaining power Weight-driven with remontoire
Rate stability during wind ±0.1 s loss per wind typical ±0.05 s loss per wind 0 s loss (continuous drive)
Maintaining run time 30–60 s typical 30–60 s typical Not applicable — no interruption
Component count Low — 4 to 5 added parts High — fusee chain, cone, stopwork Very high — remontoire arm, fly, release
Restoration cost (typical UK workshop) £200–£500 for spring service £800–£2,000 for fusee rebuild £1,500–£4,000 for remontoire overhaul
Application fit Spring-driven longcase, bracket, regulator High-grade English bracket, marine chronometer Tower clocks, observatory regulators
Service interval before wear shows 8–12 years on click and pivot 5–8 years on chain 3–5 years on remontoire release
Sensitivity to mainspring set Moderate — affects torque uniformity Low — fusee compensates None — torque is gravity-derived

Frequently Asked Questions About Harrison's Going-barrel

Almost always the pre-load stop is set wrong. If the maintaining spring is not fully wound against its stop during normal running, it acts like a soft spring in series with the mainspring and the pendulum sees a small torque dip every time barrel friction varies — that dip costs you amplitude and shows up as lost seconds per day, not just per wind.

Check by gently lifting the click off the maintaining ratchet with the clock running. If you can feel any free travel before the spring takes load, the stop pin needs re-indexing. The spring should be hard against the stop with the clock at full power.

Work backwards from the winding window you want. For an 8-day clock you need roughly 45 to 60 seconds of maintaining run time at the great wheel's working torque. Multiply that target time by the great-wheel power demand (Ttrain × ωgw) and add a 5x safety factor to account for spring rate falloff across its working arc.

For a typical 1-second-pendulum longcase pulling 1.5 to 2 mN·m at the great wheel, you want a spring storing 15 to 25 mJ. A flat steel spring 35 mm long, 4 mm wide, 0.25 mm thick, pre-loaded to about 90° of wind, sits in that range comfortably.

Depends on what is already in the case. If the clock was built with a fusee, restore the fusee — refitting a going-barrel into a fusee plate layout means re-cutting the great wheel position and you destroy the clock's provenance. If it was a going-barrel from new, stay with going-barrel.

For a fresh build the fusee compensates for mainspring torque variation across the 8-day run, which gives you better long-term rate stability than a plain going-barrel. The Harrison maintaining-power barrel only solves the winding-loss problem, not the torque-falloff problem. If you want both, fit a fusee with its own maintaining power — that is the English bracket-clock standard for a reason.

The maintaining spring is probably tired. A spring that has taken set delivers reduced torque — the train keeps running but at lower power than the mainspring provides, so amplitude falls during the wind and recovers over the next minute or two as the mainspring takes back over.

Pull the maintaining assembly and measure the spring's free deflection against a known reference. If it deflects more than 15% beyond the original blueprint dimension under a calibrated load, replace it. Set in steel maintaining springs is the single most common reason a previously good Harrison barrel starts losing amplitude during winds after 50+ years of service.

Yes, but the geometry constraint is the great-wheel-to-barrel relationship. You need room on the barrel face for the maintaining ratchet and click, and you need to re-key the great wheel to the barrel through the maintaining spring rather than directly. On a typical 8-day longcase that means an extra 4 to 6 mm of axial space on the barrel — fine if the plates were originally laid out for a fusee but tight on a slim going-barrel design.

The fix is a side-mounted maintaining wheel rather than coaxial — Dent used this layout on several late-19th-century retrofits. It costs more parts but works in any plate spacing.

Critical. Above 0.05 mm of pivot slop you get a perceptible backlash event at the start of each wind — the great wheel actually reverses for a few microseconds before the maintaining click catches, and that reversal trips the deadbeat escapement off-cycle. You hear it as a stutter in the tick at the moment you turn the key.

Measure with a dial test indicator on the great wheel rim while you rock the click manually. Anything over 0.05 mm runout means the click pivot hole is worn — re-bush it with a fresh brass bouchon sized to give 0.02 mm working clearance.

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.

← Back to Mechanisms Index
Share This Article
Tags: