240 



and, after it was made, nothing was left behind to shew when it was 

 made. This has, however, lately been altered, and the circuit is 

 now completed by a mean-time clock, which is compared every day 

 with the transit clock, and adjusted to the true time ; their com- 

 parisons being duly entered in a ledger on every occasion, shew in- 

 contestibly the limit of error of the clock, and thereby of the fall of 

 the hall each day. 



Referring to these entries, I 6nd that, during the last fortnight, 

 the correction of the clock at a quarter before one, were, on 



April 3, — 0-0 April 11, + 0-1 



... 4, - 01 ... 12, - 0-0 



5, - 0-0 



6, + 0-1 



7, + 0-2 



8, - O'l 

 10, - 0-0 



13, + 0-2 



14, - 0-1 



15, - 0-1 

 17, - 0-1 



And as the greatest daily rate of the clock during this period was never 

 more than 0-3 seconds, the above must have been sensibly the errors of 

 the clock at one hour, and, therefore, of the drop of the ball, subject 

 only to a constant correction for the time necessary for the electricity 

 to pull the various triggers. I have not been able yet to observe this 

 quantity in any but an indirect manner, but suspect that it is under 

 0*1 second. 



1th, What is the accuracy of the approximate signals afforded 

 by the half rise and the full rise of the ball at 5 minutes and at 

 2 minutes respectively before 1 ? 



As the clock is also made to give a species of electric signal to 

 the raiser of the ball, he may and should have the windlass in mo- 

 tion within 0*5 of a second of the even minute. But, inasmuch as 

 the movement of the ball on the mast is very slow, by reason of the 

 number of intervening wheels and pinions necessary to get up the 

 requisite power, the ball will not be seen to move visibly to persons 

 outside, until the crank has made several revolutions. 



From a series of four months* excellent observations of the time 

 ball by Sir T. Brisbane, it appears that, to him in St Andrew 

 Square, the rises were seen on an average 2*5 seconds too late, with 

 a probable error of about 3 seconds. While, from another series of 

 two months' observations by Mr Swan, at a greater distance from 

 the hill, as in Duke Street, the retardation, as might be expected, 



