By far the greater number of watches now made both on the Continent and in America have the mechanism known as the going barrel in substitution for the chain and fusee. In the going barrel the mainspring is of great length, and only a few coils of it are brought into action. To the going barrel itself the main wheel is attached, and thus the force of the mainspring is transmitted direct to the escapement. The general adoption of the going barrel mechanism is due to the introduction of keyless winding, which can only be adapted to the fusee watches with difficulty, and to the greater cheapness of the arrangement. Moreover, it is found that for three or four coils of a long spring, at a certain degree of winding up, the tension varies very little. The going barrel is not used in the best English work, in which absolute uniformity of motion is aimed at; but for ordinary purposes it is not of so much consequence that the same rate from hour to hour should be maintained provided the daily rate is uniform.

In watches without a fusee the apparatus for preventing overwinding is different from that in the old form of watch; it goes by the name of the Geneva stop, and the principle of it is simple. If two λvheels work together, of which one has the spaces between some two or more adjacent teeth filled up, it is evident that that wheel cannot be turned quite round. And it will be the same thing if one of the wheels is only a cylinder with a single tooth in it, and the other has a certain number of notches, not going all round, through which that tooth can pass. If, therefore, a one-toothed wheel of this kind is fixed to the barrel arbor, which is turned by the key, and works into a wheel with only 4 or 5 notches in it and a blank space through which the tooth cannot pass, it will evidently allow the barrel to be wound up the 4 or 5 turns and no more; and as it unwinds it turns the stopping wheel back again with it.

The other parts of a watch do not differ from those of a clock, except in size, and the position in which they are arranged, to bring them within the circle of the dial, until we come to the escapement; and there a different state of things arises, mainly from the fact that the balance of a watch revolves through sometimes as much as 270°, while a clock pendulum only vibrates through 4° or 5°. The balance being common to all the watch escapements, it will be proper first to describe that, and the conditions to which it is subject. The two equal arms, with equal weights at each end, in fig. 3 of article Clocks (vol. vi. p. 17) are really a balance just as much as the wheel which is commonly used as the more convenient form. But in that figure there is not to be seen that essential element of a modern balance—the thin spiral spring, opening and closing itself at every vibration. The outer end of this spring is attached to the frame by a cock R (fig. 2), and the inner to the balance at S; and the time of vibration depends only on the strength of the spring and the moment of inertia of the balance, and not at all upon the extent or angle of the vibration. And, as the force of a spring varies (approximately) inversely as its length, this suggests a ready method of regulating the watch; for it is easy to make a pointer or index, or “regulator” PT, turning on a ring fixed to the watch plate, concentric with the balance, and having two pins in it at P, called curb pins, just close enough together to embrace the spring, so that, as the index is moved one way or the other, the length of the spring which is free to vibrate may become shorter or longer. When the regulator has been moved as far as it can go towards fast, suppose, and the watch still loses, the spring has to be shortened at the cock R into which its outer end is pinned; and, in order that the balance may be capable of alteration, so as still to stand square with the escapement when the spring is in its neutral state, the other end is not actually pinned to the balance, but the cock S is on a small ring which is set on the axis or verge of the balance pretty tight by friction, but capable of being turned by hand. An index-plate like that in fig. 2 enables one to see smaller movements of the index than radial marks.

It is almost impossible to move such a regulator little enough, and with sufficient accuracy, for a very small variation of rate. One way of obtaining greater accuracy, and probably the best, is to make the regulator movable by a tangent screw acting on its end, and capable of being turned by the watch key. Another mode of giving a small motion to the regulator is by putting a portion of a wheel with teeth on it, and turning it by a small pinion with the index attached to it, so that the motion of the index exceeds that of the regulator itself as much as the radius of that wheel exceeds the radius of the pinion. Chronometers are regulated in a different way altogether. It is not expedient to alter the effective length of the spring after its length is once fixed. For it has long been ascertained that a spring does not give isochronous vibrations at all lengths, but only at certain intervals; and therefore it is necessary in an accurate timekeeper to use only one of those lengths of the spring which are found to be isochronous for different arcs of vibration; and, that being fixed, the timing of the balance can only be done by altering its moment of inertia, and this is done in chronometers by screws with heavy heads in the rim of the balance, set farther in or out as it is wanted to go faster or slower. In marine chronometers, where there is plenty of room for it, the balance-spring is generally made in a cylindrical form, with the coils all of the same diameter, instead of the flat spiral used in watches,—though it does not seem to be quite clear that the cylindrical form is materially better than the other.

The timing of a watch for position, as it is called, is a matter which requires some attention. If the balance is not exactly poised on its axis, it will have a tendency to take one position when the watch is carried vertically, as it always is in the pocket; and the time of vibration will be affected by its disposition thus to act as a pendulum. The watch ought therefore to be tried with XII, IX, VI, and III successively upwards, and if it does not keep the same rate the balance is not properly poised. Marine chronometers, indeed, being set in gimbals (a ring with the two pivots into the box at right angles with the pivots which carry the chronometer), will remain horizontal, though not without some degree of motion under the motion of the ship; and this gives the balance the further advantage of having its weight resting only on the end of the axis or verge, a position in which there is much less friction than that of a watch carried in the pocket; but there it is not of so much consequence, because the balance is so much lighter than a chronometer balance.

Compensated Balances.— A pendulum requires scarcely any compensation except for its own elongation by heat; but a balance requires compensation, not only for its own expansion, which increases its moment of inertia just like the pendulum, but far more on account of the decrease in the strength of the spring under increased heat. Dent, in a pamphlet on compensation balances, gave the following results of some experiments with a glass balance, which he used for the purpose on account of its less expansibility than a metal one :—at 32° F., 3606 vibrations in an hour; at 66°, 3598⋅5; and at 100°, 3599. If therefore it had been adjusted to go right (or 3600 times in an hour) at 32°, it would have lost 7½ and 8½ seconds an hour, or more than three minutes a-day, for each successive increase of 34°, which is about fifteen times as much as a common wire pendulum would lose under the same increase of heat; and if a metal balance had been used instead of a glass one the difference would have been still greater.

The necessity for this large amount of compensation having arisen from the variation of the elasticity of the spring, the first attempts at correcting it were by acting on the spring itself in the manner of a common regulator. Harrison’s compensation consisted of a compound bar of brass and steel soldered together, having one end fixed to the watch-frame and the other carrying two curb pins which embraced the spring, as was described at fig. 2. As the brass expands more than the steel, any increase of heat made the bar bend; and so, if it was set the right way, it carried the pins along the spring, so as to shorten it. This contrivance is called a compensation curb ; and it has often been reinvented, or applied in a modified form. But there are two objections to it: the motion of the curb pins does not correspond accurately enough to the variations in the force of the spring, and it disturbs the isochronism, which only subsists at certain definite lengths of the spring.

The compensation which was next invented left the spring untouched, and provided for the variations of temperature by the construction of the balance itself. Fig. 3 shows the plan of the ordinary compensation balance. Each portion of the rim of the balance is composed of an inner bar of steel with an outer one of brass soldered upon it, and carrying the weights b,b, which are screwed to it. As the temperature increases, the brass expanding must bend the steel inwards, and so carries the weights farther in, and diminishes the moment of inertia of the balance. The metals are generally soldered together by pouring melted brass round a solid steel disk, and the whole is afterwards turned and filed away till it leaves only the crossbar in the middle lying flat and the two portions of the rim standing edgeways. The first person who practised this method of uniting them appears to have been Thomas Earnshaw, who brought the chronometer to the state in which it has remained for the last century, with scarcely any alteration except more complete compensation.

The adjustment of a balance for compensation can only be done by trial, and requires a good deal of time. It must be done independently of that for time,—the former by shifting the weights, because the nearer they are to the crossbar the less distance they will move over as the rim bends with them. The timing is done by screws with heavy heads (t,t, fig. 3), which are just opposite to the ends of the crossbar, and consequently not affected by the bending of the rim. The compensation may be done approximately by the known results of previous experience with similar balances; and many watches are sold with compensation balances which have never been tried or adjusted, and sometimes with a mere sham compensation balance, not even cut through.

Secondary Compensation.— When chronometers had been brought to great perfection it was perceived that there was a residuary error, which was due to changes of temperature, but which no adjustment of the compensation would correct. For, if the compensation was adjusted for two extreme temperatures, such as 32° and 100°, then the chronometer gained at mean temperatures; and, if adjusted for any two mean temperatures, it would lose for all beyond them. This error was observed, and attempts were made to correct it before anybody had pointed out how it arose: this appears to have been first done in a paper in the Nautical Magazine by E. J. Dent in 1833; and he gave the following illustration of it. The variation of the force of the spring proceeds uniformly in proportion to the temperature, and therefore may be represented by a straight line inclined at some angle to another straight line divided into degrees of temperature. But the inertia of a balance of the common construction cannot be made to vary uniformly according to the temperature, but will vary more rapidly in cold than heat; and consequently its rate of variation can only be represented by a curve, and the curve can only be made to coincide with the straight line representing the rate of variation of the spring in two points,—either two extremes, or two means, or one extreme and one mean point.

The same thing may be shown mathematically, as follows. Let r be the distance of the compensation weights b , b, in fig. 3 (which we may assume for convenience to be the whole mass M of the balance) from the centre at some mean temperature, and let dr be their increase of distance due to a decrease of some given number of degrees of heat, under the action of the compensation bars. Then the new moment of inertia will be M (r 2 +2rdr+dr 2 ), and the ratio of the new to the old will be 1 + (2 dr/r ) + ( dr/r) 2 ; and the term (dr/r ) 2 is now too large to be disregarded, as it might be in pendulums, where the compensation dl/l is only required to be about 2 / 20 th of the in a balance. It is found that an equal increase of temperature will produce an equal or rather a less motion ( - dr) of the weights towards the centre than from it at any given point; but, calling it only equal, the ratio of the decreased moment of inertia to the original one will be 1 - (2 dr/r) + (dr/r ) 2 , so that the increase and the decrease from the mean amount differ by twice (dr/r) 2 ; in other words, the moment of inertia of the balance varies less in passing from mean to hot temperature than from mean to cold; and consequently if it is adjusted for mean and cold it will not have decreased enough at an equal increase from mean to hot, or the chronometer will lose, and if adjusted for the two extremes it will gain at mean temperatures.

The correction of this error is called the secondary compensationi, We shall give a short description of the principal classes of inventions for this purpose.

The first that was disclosed was Eiffe’s (sometimes called Molyneux’s), communicated to the astronomer-royal in 1835. In one of several methods proposed by him a compensation curb was used; and, though, for the reasons given before, this will not answer for the primary compensation, it may for the secondary, where the motion required is very much smaller. In another the primary compensation bar, or a screw in it, was made to reach a spring set within it with a small weight attached at some mean temperature, and, as it bent farther in, it carried this secondary compensation weight along with it. The obvious objection to this is that it is discontinuous; but the whole motion is so small, not more than the thickness of a piece of paper, that this and other compensations on the same principle appear to have been on some occasions quite successful. Molyneux took a patent for a secondary compensation exactly the same as this of Eiffe’s, then before the astronomer-royal.

Another large class of balances, all more or less alike, may be represented by E. J. Dent’s, which came next in order of time. He described several forms of his invention; the following description applies to the one he thought the best. In fig. 4 the flat crossbar rr is itself a compensation bar which bends upwards under increased heat; so that, if the weights v, v were merely set upon upright stems rising from the ends of the crossbar, they would approach the axis when that bar bends upwards. But, instead of the stems rising from the crossbar, they rise from the two secondary compensation pieces stu, in the form of staples, which are set on the crossbar; and, as these secondary pieces themselves also bend upwards, they make the weights approach the axis more rapidly as the heat increases; and by a proper adjustment of the height of the weights on the stems the moment of inertia of the balance can be made to vary in the proper ratio to the variation of the intensity of the spring. The cylindrical spring stands above the crossbar and between the staples.

Fig. 5 represents Loseby’s mercurial compensation balance. Besides the weights D, D, set near the end of the primary compensation bars B, B, there are small bent tubes FE, FE with mercury in them, like a thermometer, the bulbs being at F, F. As the heat increases, not only do the primary weights D, D and the bulbs F, F approach the centre of the balance, but some of the mercury is driven along the tube, thus carrying some more of the weight towards the centre, at a ratio increasing more rapidly than the temperature. The tubes are sealed at the thin end, with a little air included. The action is here equally continuous with Dent’s, and the adjustments for primary and secondary compensation are apparently more independent of each other; and this modification of Le Roy’s use of mercury for compensated balances (which does not appear to have answered) is certainly very elegant and ingenious. Nevertheless an analysis of the Greenwich lists for seven years of Mr Loseby’s trials proved that the advantage of this method over the others was more theoretical than practical; Dent’s compensation was the most successful of all in three years out of the seven, and Loseby’s in only one. Loseby’s method has never been adopted by any other chronometermaker, whereas the principles both of Eiffe’s and of Dent’s methods have been adopted by several other makers.

A few chronometers have been made with glass balance-springs, which have the advantage of requiring very little primary and no secondary compensation, on account of the very small variation in their elasticity, compared with springs of steel or any other metal.

Dent also invented a very different method of effecting the primary and secondary compensation at once, and without any additional appendage to the balance or addition to the cost. He called it the prismatic balance, from...