ON TIME MEASURERS. 337 



that its arc of oscillation should be invariable ; and finally that it be 

 unaffected by changes in the density of the atmosphere. 



These conditions may at first sight appear very simple, but cannot 

 be so readily fulfilled as might be imagined. 



In the first place, of whatever material the pendulum may be com- 

 posed, some small alteration in its length (which we have mostly to 

 consider) will occur at every change of temperature ; and as we cannot 

 ensure a perfectly uniform temperature, a correction or compensation 

 must be applied to counteract the effects of a variable one. 



The arc of oscillation must not vary because the pendulum in its 

 swing does not describe a cycloidal curve, as it should do in order 

 that the long and short oscillations might be performed in the 

 same space of time ; but almost a segment of a circle which diverges 

 rapidly from the cycloid, consequently a change in the arc described 

 altering the amount of divergence from the true theoretical path, alters 

 the time of oscillation. This variation is known as the circular error, 

 and cannot be corrected. All we can do is to endeavour to make the 

 error a constant quantity by maintaining a constant arc of oscillation. 



The variations in the density of the atmosphere affect the duration 

 of the oscillation of the pendulum although to a much less .extent than 

 a change of temperature or a change of arc. Still the variation is 

 sufficiently large to make a correction desirable. It amounts to about 

 03 sec. in 24 hours for a change of pressure indicated by one inch fall 

 or rise of the mercury in the barometric tube. 



As we propose to invite your attention more particularly to the com- 

 pensation of the pendulum for changes of temperature, and to Sir Geo. 

 Airy's correction for changes of barometric pressure (as exemplified 

 in the clock downstairs which is almost a counterpart of the 

 Greenwich Sidereal Stand one), we will dismiss the problem of the main- 

 tenance of a constant arc of oscillation after stating that for its accom- 

 plishment we use an escapement and train of wheels actuated by a 

 weight, the whole being so constructed as to impart to the pendulum 

 at every oscillation or alternate oscillation a certain push or impulse 

 which shall be of uniform force, and just equal to the resistance 

 encountered by the pendulum in its swing from side to side. We will 

 not stop to describe the forms of escapement best fulfilling this condi- 

 tion but will simply mention that the three varieties most in favour are 



y 



