]\gt ; attraction is not only taken away from A but added to , so that the tilting of the beam is that due to nearly double the attraction to be measured .
Of course there are what we may term ctions , in the first position , of on , and in the second position , of on , but these may be allowed for in the calculations .
We cannot give any mathematical expression for the attraction of on the beam and suspending wires , owing to their irregularity of shape .
But this attraction is eliminated if a second experiment is made in which A and are raised equal known distances to * , and .
For the difference between the two increments of weight on the right , is due solely to the alteration of the positions of A and relative to , the attraction on 3the beam remaining the same in each .
From the observed effect of a known alteration of distance the attraction at any distance can be found .
This is , shortly , the method adopted .
The arrangement was ultimately complicated by the addition of a second mass .
Originally the mass was alone on a turn-tabJe xperiments which Irnade 1pose thangeComp nensity , when Isupposed thework winished , ation ohowed surious aieswhich Iscribe tting oacement oaffect tance .
However , after affect oound t which revolved about a vertical axis immediately under the central knife-edge of the Elevation of balance room and observing room .
The front of the case is removed , and the front pillar is not shown .
The pointer and mirrors are at the back .
there was also a secular change , the amount of tilt gradually creasing .
This secular : : change was probably due to increasing rigidity of the floor , so that it tilted over bodily , moving the snpports of the balance with it , an increase partly due , perhaps , ressure ouilding , which hrected.welve years , butchiefly , hink , ngine recently erected noing heavywork , ibrations wlainly foubt treatly dedthe floor iettling d\ldquo ; econd balancing mherefore added , half a : bottom of each cup , so that the bottom of the mass can be seen for the purpose of measuring the vertical diameter .
Two beams , metres long , run across the turn-table 26 centims. apart , with the cups between them , and across the ends are two boards , each with circular hole 12 centims. in diameter , and in either of these the smaller , or balancing mass , may rest .
These beams are braced by brass rods to brass uprights at their middle points to diminish bending .
The turn-table is moved by an endless gut lope passing round it , and fixed at one point of the rim .
The two sides of the rope pass over pulleys on to a drum in the room above .
There are stops on the circular rail , against which come brass pieces on the turn-table when the masses are in position at either end of the motion .
The drum can be turned easily by the observer at the telescope .
Since the knife-edges and planes of the balance of eel , all the moving parts of the apparatus are free from iron .
As an illustration of the necessity of this , I may mention that for some time I used what I supposed to be a brass wire rope to move the turn-table , but on looking out for the explanation of some irregularities , I found that the brass was wrapped round a core of steel wire , which acquired poles at the highest and lowest points in the position in which it always rested between different sets of weighings .
These poles had quite an appreciable action on the balance beam .
The Balance.\mdash ; This is of the large bullion balance type , with gun-metal beam and steel knife-edges and plates .
It was made specially for the experiment by Mr. OERTLING , with extra rigidity of beam .
Its performance has shown the great excellence of the design .
The central knife-edge is supported on a teel plate by a frame-work rising 107 centims. above the base plate , and the usual moveable frame can be raised or lowered from outside the case , fixing the beam or setting it free to oscillate .
The beam has often been left free to oscillate for months at a time , with the full load of 20 kilogrms .
on each side , but I have no reason to suppose that the knife-edges have suffered at all .
The length of the beam was measured by taking the length of each half separately by a beam compass , and the mean of several measurements gave centims. as the total length .
The standard scale used throughout was that of a cathetometer made by the Cambridge Scientific Instrument Company .
This scale has been verified at the Standards Office , and taking its coefficient of expansion as , it may be regarded for our purpose as perfectly correct at , any errors being at that temperature much less than the errors of experiment .
Comparing the beam compass Vanes dashpd Double Suspension Mirror ( half size ) .
Of course , if the mirror were weightless , it would not affect the sensibility of the balance , and the threads might be brought very close together .
But the weight of the mirror--it is silver on glass , 56 millims. milhms .
millims.\mdash ; has a considerable effect on the sensibility , diminishing it with decrease of distance between the points of suspension .
In practice it has been found convenient to work with the threads parallel , and from 3 to 4 millims. apart , the time of swing one way being adjusted to about 20 seconds .
A less time hardly suffices for a correct determination and record of the scale reading .
Taking 4 millims. as the distance , and supposing the bracket to be 600 millims. below the knife-edge of the balance , the mirror evidently turns through an angle 150 times as great as that through which the beam turns .
The drawback to this method of magnification is that the mirror has its own time of swin and is easily disturbed .
The swings of the rnirror and the disturbances are , however , effectually damped by having four light copper vanes attached to the end of A and were selected for use as being the nearest to each other in value .
and were kept untouched in boxes till 1890 .
In the various experiments made between 1886 and the final weighings , A and had necessarily been handled to some extent , especially through frequent breaking of the silk fibre suspension used before the subsidiary beam described above , and it appeared possible that their weights might be altered .
It was also necessary to determine whether an appreciable amount of dust was deposited on them in the course of several weeks as it was inconvenient to dust them frequently .
The riders and might be assumed to have the same weight as in 1886 , and could be taken as standards .
To make the weighings a 16-inch chemical balance was arranged with a double suspension mirror on exactly the principle already described for the large balance .
The apparatus was put together quickly with materials hand , and might easily be greatly improved .
It is only described here to show how accurate the method is , even with such rough apparatus , and that it is applicable to a small as well as a large balance .
A cork sliding on the pointer with a horizontal needle stuck in it , served to support one thread of the mirror ; a stand with a projecting arm\mdash ; one made to hold platinum wires in a Bunsen flame\mdash ; served to support the other thread .
A wire with a small copper vane depended from the mirror and was immersed in an oil dashpot .
The telescope and a millimetre scale were on alevel with the mirror about 2 metres distant on one side of the balance .
Two brass strips , parallel to each other and the beam , were fixed on the top of one arm of the beam , and in each of these were two V notches in which centigramme riders rest .
levers , worked by cams on a rod rotated by the observer , picked one rider up and let down the other , so that the effect was equivalent to the transfer of 1 centigrm .
from one notch to t , he other .
Their distance apart was such that this was equivalent to the addition of 3284 milligrm .
to one pan of the balance .
This was the arrangement described in my former paper .
Attached to one pan was a pair of brass strips parallel to each other , and such that the riders or , would just rest across them .
Two lifting rods worked up and down between these strips , so that of the two riders to be compared , one could be picked up at the The following are the values in terms of the cathetometer scale already referred to , the temperature being C. :\mdash ; 33.95 Diameter of therefore 24 .
: The value of was found by measuring the shortest distance of from the wall when respectively in the first and the second positions on the turntable .
It was found that .12 whence circleinwhichthecentreofWemayobtainfromtheseMmovesWithperfectadjustmentthisshouldbemeasures.anindependentvalueoftheradiusofthe SL at The plate is used as follows:\mdash ; Adjust it normal to the optic axis of the telescope , and move the telescope till the required point is brought as near to the cross-wire as is possible by the hand .
Clamp the telescope , and then tuln the plate till the poin is exactly on the cross-wire .
Read the position of the pointer attached to the plate on its scale .
Repeat these operations with the other telescope on the other point , then turn the instrument about its vertical axis till the telescopes sight the vertical scale placed at the same-distance away as the two pointa - Looking through one of the telescopes the cross-wire is in general not exactly on a division .
Turn the plate so that first the nearest division above , and next the nearest division below , is on the cross-wire .
Reading the position of the pointer in each case , interpolation gives us the reading on the vertical scale corresponding to the position of the pointer when the cross-wire was between the two scale divisions .
Doing this for each telescope the difference between the two points is found in terms of the vertical scale .
The plates I have used are about 9 millims. thick , and the pointers about 9 centims. long .
They move over scales such that 25 to 27 divisions correspond to a shifting of 1 millim. The lower ale is graduated from to 50 , the upper from 50 to 100 to prevent confusion .
The 50 divisions occupy a tance of 66 millims. It be observed that in this form of instrument the level error is practically entirely ob.viated .
It can only come in if the scale is not at the same distance as the height to be measured , and may then be made negligible in practice by levelling the telescopes .
Indeed , the uncertainty of measurement appears only to depend on the effect , doubtless through an air wave .
In a high wind the balance was always unsteady , partly , I suspect , through rushes of air into and out of the case with sudden pressure changes , and partly through changes of ground level , with variations of wind pressure against the building .
At all times there was a march in one direction or the other of the centre of swing .
This was especially marked soon after the frame was lowered and the beam left free .
As already remarked , readings were not taken till changes due to change in strain of the beam had subsided .
But the march was appreciable at other times , as will be seen from the diagrams .
Perhaps the change was sometimes due to tilting of the ground , with barometric variation , since the balance was a very delicate level , and sometimes due to the change in buoyancy of the air affecting two sides unequally , though I have not been able to make out any direct connection between barometric height and position of centre of swing .
I believe that the explanation is be sought for the part in unsymmetrical effect on the beam of slight changes of temperature , for I have frequently noticed that a rising temperature produced an upward march , and a falling one a downward march .
This explanation is supported by the following table of observations ' of the centre of swing , extending from May 9 to May 22 , 1890 , the balance being free , and the balance room undisturbed meanwhile .
The relation between temperature and centre of swing is represented in Diagram IX .
( Plate 19 .
) Of course , after a change in the of the attracting masses or of the riders , the balance does not at once settle in a new position of equilibrium , but oscillates about it .
Inasmuch as the balance never rests in this position , it is better to term it the centre of swing rather than the equilibrium position or resting point .
The dashpot used to damp the vibrations of the mirror reflecting the scale serves also to damp those of the balance beam , and they die down rapidly .
Instead of waiting , however , to observe directly the point on which they are closing in , it is much more exact , and also saves much time , to find the centre of swing as with an undamped balance from the extremities of the swings .
I have always observed and recorded four extremities of three successive swings , occupying in all a little more than a minute .
Notwithstanding the very considerable damping , the successive lengths of swing are still in geometrical progression , but the rate of reduction is too great to allow ordinary approximation , in which the geometrical is assumed to be an arithmetical progression .
The exact method of determining the centre of swing is as follows:\mdash ; the two positions , the vertical distances between attracting and attracted masses were measured by the cathetometer in the manner explained in Table IL , and the position of the attracted masses was altered .
A full account of all the weighings is given in Table III .
, and the results are represented in Diagrams I.-VI .
The three upper rows of points in each diagram represent the centres of swing , those in the initial position being marked .
After movement of the rider they are marked , and after movement of .
the masses they are marked O. The base lines for the different rows are altered to save space , as described on the diagrams , for on the scale adopted the rider series would always be about 10 inches above or below the initial series .
In Diagram I. the rider and mass series are also brought down and superposed on the initial series , so that each of the three has the same average height .
It will be seen that all three are affected by the same disturbances .
The advantage of the short time of swing and the mode of combining the results in threes wiJl be realised more easily from this superposition .
The base line may be regarded as a time scale , as the instants corresponding to centres of swing were almost exactly equidistant .
In each case , under the representation of the centres of swing , c$re plotted the resulting values of , and at the side will be found a representation of the distribution of results about the mean .
Assuming that each day 's mean value is correct , and that the differences for different days are to be set down to variation of distance , \amp ; c. , we can find the distribution of all the values about the mean by simply superposing the marginal curves at the side of the figures .
The result fairly shows the accuracy as far as the weighing alone is concerned .
It is represented in Diagram VII .
, where A is the comparison of the values of in Diagrams I. and II .
, shows a very curious similarity in the fluctuations , and at first I was inclined to think there was some common external disturbance producing these fluctuations .
But an analysis of the two sets of values appeared to show that the resemblance is merely accidental .
When the values of and are set out separately , it is seen that the fluctuations depend chiefly on , of which the fluctuations are slightly like each other for the two series , while those of are quite different , but such that they make the fluctuations in resemble each other much more closely than those in ] .
Further , it is not easy to see how fluctuations due to some external source would affect the values of equally in the upper and lower positions and not have any effect on R. Some periodic change of level might be suspeoted , but this ought certainly to be traced in R. I have examined all the other diagrams and plotted out the component values of and , but have found no trace of resemblance , so that I think the curious likeness in I. and II .
must be set down to accident .
There is a curious step by step descent of the centre of swing in the initial position on September 23 , Diagram VI .
, which I cannot explain .
It may be due to some error in the method of finding the centre of swing which comes in with a rapid march of that centre .
The effect on the result is probably only small , for the value of obtained with a march in the reverse direction on September 25 is very nearly the same , the two values being September 23 2112753 .
The following is a list of the weighings recorded , with the distances measured and the mean values of the attraction:\mdash ; SET I. that movement of rider or mass decreased the reading .
This was done in order to lessen errors due to want of symmetry .
If reversal had no effect , Set II .
should , with the increased distance recorded above , give a value of in the lower position of about , instead of 998 .
The larger value actually found is no doubt chiefly due to a want of symmetry in the large attracting mass M. The effect of this want of symmetry will be discussed after the investigation of the mathematical formula , and an account will be given of an independent method of detecting it .
I think there is still outstanding a small difference , due , perhaps , to want of symmetry in the turn-table or in the attracted masses .
The result of the reversal shows how necessary it was to make it .
I ] have liked to have in Set II .
as many deter- minations as in Set I. , so that the mean should be based on values of equal weight .
During June and July , 1890 , a complete set of 100 in each position , upper and lower , was made ; but , owing to the pressure of other work , I was unable to calculate the results till the completion of the set .
I then found that the value of was still more than in Set , and , on plotting out the results , it appeared that occasionally the rider value fell very considerably , and in an irregular way .
On examination , there was little doubt that the rider came in contact occasionally with the suspending frame , when it was raised and should have been clear from it .
Very likely temperature changes had brought about a displacement of the lever apparatus .
Comparison with Set I. seemed to show that during that set no such contact had taken place , for there was no comparable irregularity .
As it appeared dangerous to attempt to disentangle the good from the bad , the set of June and July was rejected , and Set II .
was taken as recorded .
When I had made the weighings giving 50 and 52 values in the two positions respectively , the balance became so irregular , through the cooler weather , that it was useless to continue work .
Rather than carry over the experiment into another season , when it might be necessary to repeat the whole of the work , I have preferred to take Set II .
as it stands , and give it the same weight as Set I. The final results are calculated from the means of Sets I. and II .
, as explained hereafter .
may here state the results obtained:\mdash ; : This is to be equated to the difference in the values of the change in attraction in the two positions , as determined by rider .
Let mass of each rider .
length of the small rider beam .
mass deflection rider deflection in lower position .
, , , , , , , , , , upper , , acceleration of gravity , or dynes weight per unit mass at Birmingham .
Then Whence we may find the gravitation constant tmass arm ng am .
, where all the quantities on the right hand are given in the Tables at the end .

