44 



NATURE 



[November 14, 1895 



No. 2. Calibration of bridge-wire and determination of error 

 due to the position of the scale zero mark. 



No. 3. Determination of the coil errors. 



No. 4. Value of the mean box-unit in terms of the Board of 

 Trade ohm. (This was tiot a necessity for the temperature 

 measurements, but it appeared advisable to ascertain it.) 



No. 5. Determination of R, R^ and 5 for each thermometer. 



It is impossible to give any full description of the operations 

 — I can but indicate the methods adopted. 



No. I has already been described supra. 



No. 2. A length of 4 cm. of an unused portion of the 

 bridge-wire was soldered across the thick posts supporting the 

 coil marked FI ; thus if balance was obtained with the plug FI 

 in, its withdrawal would compel a movement of about 2 cm. in 

 the contact-maker to restore equilibrium. As the constitution 

 of both wires was the same, and as they were included in the 

 glass case, the length moved by the contact-maker was unaltered 

 by changes in temperature. 



Cj C2 (Fig. i) were connected by a thick copper strip, while/, 

 and /j were connected with the terminals of a resistance-box 

 (r,) having a slightly larger unit than the Kew box, and the 

 extremities of ;\ were connected with two other boxes {r^, r^), of 

 which r.j could be altered by certain steps from 5 to 20,000 

 ohms, and r^,hy any quantity from o"i to 10,000 ohms ; thus r, 

 and ^3 may hp regarded as shunts to r^. All three boxes were 

 maintained at a constant temperature. 



This arrangement was adopted to enable a balance to be 

 obtained when the contact-maker was at, or very near any given 

 position ; for example^suppose contact desired at reading 

 + 15, rj and r^ were made as large as possible, and r, adjusted 

 until the bridge- wire reading somewhat exceeded 15 ; rj was 

 then reduced until the reading but slightly exceeded 15. Re- 

 ductions of some hundreds of ohms in r^ would now cause but 

 small alteration in the combined resistance of the three boxes, 

 provided r^ greatly ex-ceeded r^ and r^ when in parallel arc. It 

 was thus always found possible and easy to balance so that 

 contact was nearer than 0'5 m.m. to any desired position. 



The plug FI was now withdrawn, and the corresponding 

 movement of the contact-maker noted. Repeated observations 

 throughout the whole length of the wire were thus taken, not 

 only by the writer, but also by Mr. G. M. Clark, who performed 

 an independent series. Denote the resistance of the wire across 

 FI by U , then the reciprocals of the observed lengths give in terms 

 of U the effective resistance per cm. of the wire at the middle 

 point of each range. By means of the arrangement described 

 above, the observations were so conducted that these middle 

 points fell almost exactly on the integral numbers of the scale, 

 and, by plotting, the exact values at the integers were obtained. 



By repeating the process with the plug H the value of U in 

 terms of H could be accurately ascertained, and a check was also 

 obtained on the previous observations. The value per cm. for 

 the whole wire could thus be found in terms of H. Later in the 

 operations, when the value of H was known in terms of the 

 mean box unit, the b.-w. values were expressed in terms of the 

 mean unit and integrated on each side of the zero mark. 



No. 3. The eflfect of any inequality in the bridge arms of 

 Si Sj (Fig. i) could not be ehminated by means of the Correc- 

 tion Tables, nor could such an error be easily detected by means 

 of the box itself, as is the case in the remaining coils. Great 

 attention was, therefore, devoted to securing equality. It is 

 certain that they do not differ by i in 100,000. 



The errors of coils A to H and FI (the temporary connection 

 across FT having been removed) were now determined by a 

 method originally adopted by Prof. Callendar. No great efforts 

 were made to secure the identity of these coils with their nominal 

 values, for it was certain that some small corrections would in 

 any case be necessary, and as the magnitude of the correction in 

 no way increases the labour of calculation, the time and energy 

 expended in any exact adjustment would have been wasted. 



The procedure was as follows :— 



All plugs were inserted, and the balance obtained with the 

 contact-maker at any convenient position by adjusting r^, r^, r^, 

 as previously described. H (5) was then removed, and the 

 change in reading required to readjust balance observed. Let 

 Zj be the consequent change in reading ; fj, r„ r^ were then 

 altered until the contact-maker was brought back to about the 

 same position as it occupied when H was in. H was then in- 

 serted, and G (10) removed. Let change = Zo ; contact was 

 again brought to first position, D (20) removed, ard G and H 

 replaced. Let change = Z3, &c. 



NO. 1359, VOL. 53] 



When the process is completed we thus get a series of 

 equations 



A - (B to H) = Z, 

 B - (C to H) = Z7 

 C - (DtoH)==ZB, &c 

 By subtraction we then get 



A - 2B = Zg - Z7 



B - 2C = Z7 - Zg, &c. 



Now the values of Z, Zj, &c., in terms of H&xq already knowfi 

 from the previous operations, hence A, B, &c., in terms of H 

 can be found. 



As the right-hand extremities of the intervals Z^ Zj, &c., were 

 approximately in the same position, nearly the same portion of 

 the bridge-wire was used throughout ; thus any errors in the 

 previous calibration would but slightly affect the results. 



Knowing the values of all the coils in terms of H, it is then 

 easy to express them all in terms of the mean coil, and hence in 

 terms of the mean box unit, a corresponding correction being 

 made in the integrations of the bridge-wire. 



The zero error of the scale was next determined by reducing 

 the resistances between C,, C3, and Pj, Pj (Fig. i) to zero, and 

 replacing all plugs. The observations were checked by re- 

 versing all the connections. The displacement was found to be 

 4- o'oo5 cm.. 



No. 4. Finally the sum of all the coils was determined by 

 means of a dial box, of whose comparison with the B.A. 

 Standard full particulars are given in Phil. Trans. A, 1893, 

 pp. 407-410. The result was that the mean Kew box unit at 

 20° C. = 0*0099993 Board of Trade ohms. 



At the conclusion of the standardisation a large number of 

 readings of the same resistance were taken by different observers 

 with different combinations of coils and b.-w. readings. The 

 accuracy of all the corrections was thus exposed to a crucial test, 

 the results of which were satisfactory. 



. No. 5. The standardisation of the thermometers was per- 

 formed after the installation of the apparatus at Kew, the 

 previous observations having been made in my own laboratory. 



I find that many misapprehensions are prevalent as to the 

 nature of the operations, and it may therefore be of assistance 

 to observers who standardise their own thermometers, if I give 

 a complete example of one set of the observations as taken at 

 Kew, together with their reductions. I select thermometer K„ 

 as several observers took part in its standardisation, and it 

 therefore well illustrates the order of accuracy obtainable. 



The numbers in italics show the corrections resulting from 

 the standardisations of which an account has been given. The 

 times are always entered, since the observations of the barometer 

 cannot be taken simultaneously with the temperature measure- 

 ments, and it is necessary, therefore, when working with steam 

 and sulphur, to form a time-chart by which to ascertain the 

 correct pressure at the moment of observation. 



Thermometer Kg. Determination of R,,. 



Mean 



257-827 



