We can now follow the action of the balance wheel. It runs perfectly correctly at, say, a temperature of 60°. Hold it over a candle. The spokes lengthen, and carry the rim-pieces _outwards_ at their fixed ends; but, as the pieces themselves bend inwards at their free ends, the balance is restored. If the balance were placed in a refrigerating machine, the spokes would shorten, but the rim-pieces would bend outwards.
As a matter of fact, the "moment of inertia" cannot be kept quite constant by this method, because the variation of expansion is more rapid in cold than in heat; so that, though a balance might be quite reliable between 60° and 100°, it would fail between 30° and 60°. So the makers fit their balances with what is called a _secondary_ compensation, the effect of which is to act more quickly in high than in low temperatures. This could not well be explained without diagrams, so a mere mention must suffice.
Another detail of chronometer making which requires very careful treatment is the method of transmitting power from the main spring to the works. As the spring uncoils, its power must decrease, and this loss must be counterbalanced somehow. This is managed by using the "drum and fusee" action, which may be seen in some clocks and in many old watches. The drum is cylindrical, and contains the spring. The fusee is a tapering shaft, in which a spiral groove has been cut from end to end. A very fine chain connects the two parts. The key is applied to the fusee, and the chain is wound off the drum on to the larger end of the fusee first. By the time that the spring has been fully wound, the chain has reached the fusee's smaller extremity. If the fusee has been turned to the correct taper, the driving power of the spring will remain constant as it unwinds, for it gets least leverage over the fusee when it is strongest, and most when it is weakest, the intermediate stages being properly proportioned. To test this, a weighted lever is attached to the key spindle, with the weight so adjusted that the fully wound spring has just sufficient power to lift it over the topmost point of a revolution. It is then allowed a second turn, but if the weight now proves excessive something must be wrong, and the fusee needs its diameter reducing at that point. So the test goes on from turn to turn, and alterations are made until every revolution is managed with exactly the same ease.