Mainspring and Fusee

Cut a 50 mm brass disc and a strip to form a 12 mm deep wall. This is the barrel — the drum the mainspring lives inside.

Drill the disc centrally and fit a 4 mm brass arbor. Cut a slot across the arbor to catch the inner end of the spring.

Cut 600 mm of 10 mm wide spring steel strip. Round both ends and remove every burr along the edges — a nick in a tensioned spring is where it will eventually snap.

Hook the inner end into the arbor slot and wind the spring in, anchoring the outer end to the barrel wall. Wear eye protection — a spring that escapes during winding moves fast.

Before building the fusee, hang a light cord and weight from the barrel and measure the turning force at full wind, half wind and nearly run down. Record all three. This is the problem you are about to solve, in numbers.

Turn or file a brass cone 45 mm long, 15 mm at the small end and 35 mm at the large end. That roughly 1:2.3 radius ratio is what will compensate the spring.

Cut a continuous spiral groove along the cone, 2 mm pitch, deep enough to seat the chain. The groove must run unbroken from small end to large — a break lets the chain jump and the compensation fails at that point.

Mount both on a brass base plate with their axes parallel, 70 mm apart, each free to turn. Misaligned axes make the chain climb out of the groove.

Anchor a fine chain to the barrel and to the large end of the fusee. Wind it so that when the spring is fully wound the chain sits on the small radius of the cone.

Wind and release slowly. The chain must walk from the small radius toward the large radius as the spring unwinds. If it goes the other way the compensation is inverted and the output gets worse, not better — reverse the chain routing.

Repeat the step 5 measurement, now at the fusee arbor, at full, half and low wind. Compare the two curves. A well-cut cone gives a noticeably flatter output than the bare barrel.

If the output still falls off at the end, the large radius is too small — file the cone's large end wider and re-measure. The correct profile is found by iteration; it is not a formula you get right first time.

The problem the spring created. Weight-driven clocks deliver perfectly constant torque — a hanging weight pulls the same whatever height it is at. Springs traded that away for portability. When spring drive arrived in the 15th century, constant force became one of horology's long-running holy grails, and it is still pursued in high-end watchmaking today.

How the cone does it. Torque is force times radius. The mainspring's force falls as it unwinds, so the fusee increases the radius at which that force acts. Fully wound, the chain pulls on the smallest radius: a large force through a small lever arm. Run down, the chain has walked to the largest radius: a small force through a large lever arm. The product stays roughly constant. It is mechanical gearing that varies continuously — the same reasoning as a bicycle's gears, arrived at four centuries before bicycles.

Gut before chain. The first fusees used cord made from animal gut — flexible, cheap, and prone to wear and sudden failure. Metal chains became standard by the mid-1600s and are what the mechanism is now named for. A fusee chain is a miniature roller chain, and making one is a specialist trade in itself; historically much of it was outsourced piecework.

Why it disappeared, mostly. The fusee is bulky, fiddly and expensive. Two later developments made it largely unnecessary: better spring steel with a flatter torque curve, and the going barrel used with an escapement less sensitive to torque variation — plus the practice of only using the middle portion of a spring's travel, where its output is naturally flattest. The fusee survives in marine chronometers, where accuracy justified any complexity, and in a handful of modern luxury watches as a demonstration of craft.

What it bought. A clock is only as steady as its power supply. Before the pendulum, drive regularity was the dominant error source, and the fusee was the main defence against it. Its long persistence in marine chronometers is the giveaway: when a clock error meant a longitude error meant a shipwreck, the extra parts were worth it.