Z 
FEBRUARY 15, fs84.] 
For this purpose the heavy main conductor 
was cut, and the two ends were inserted into 
large mercury-cups, cut out in a block of wood 
an inch and a half thick. ‘These cups were 
also connected by about forty feet of number 
0 copper wire, the ends of both main and 
shunt wire being well immersed in the mercury, 
and pressed close together. These mercury- 
cups were connected with two others by means 
of short copper wires, and into the second 
pair the ends of the galvanometer wires were 
plunged. As thus arranged, about one-fifth of 
the current was taken through the galvanome- 
ter. - Even with this division of the current, it 
was found, that, when using the strong current 
from the Weston dynamo, the wires of the gal- 
vanometer were somewhat heated ; and in order 
to avoid this result, a short piece of number 0 
wire, not more than two or three inches in 
length, was bent so that it could be inserted 
in the mercury-cups, and thus cut the galva- 
nometer out, except during the few moments 
necessary for taking a reading. During all of 
the ‘resting’ periods this short wire carried 
by far the greater portion of the current, and 
thus tended to prevent the heating of the shunt 
wire proper as well as of the galvanometer. 
The determination of the ratio of the two 
parts into which the current was divided, or 
the value of the ‘shunt multiplier,’ was, of 
course, a matter of great importance. In the 
preliminary measurement of this ratio the cur- 
rent from the Thomson-Houston machine was 
of great service on account of its steadiness. 
To begin with, a number of tests were made to 
discover if the connection resistances were of 
such importance that any accidental variation 
in them would perceptibly alter the shunt ratio. 
The shunt was repeatedly lifted out of the cups 
and replaced, and the galvanometer connec- 
tions were broken and remade. Every thing 
that could be disturbed was disturbed; but, 
upon reconstruction, the result was found in all 
cases to be practically unaltered. On Sept. 28 
a series of twenty measurements was made with 
the shunt alternately in and out, using a cur- 
rent of 10 ampéres. The results agreed closely 
with each other, and gave 4.6 as the value of 
the shunt multiplier. On the following day the 
tests of the incandescent machines began; and 
the shunt was not moved from its position, 
nor disturbed, until after the conclusion of the 
entire work. On Oct. 3, after all of the reeu- 
lar tests had been completed, another test of 
the shunt was made, with a current of 10 
ampéres, as before. Ten observations were 
made, all of which agreed in giving a ratio of 
a little more than 5.0. This result was quite 
SCIENCE. 
179 
unexpected, and the discrepancy between it 
and that obtained from the first test was en- 
tirely too great to be accounted for by errors 
of observation. As circumstances prevented 
further tests in Cincinnati, it was determined 
to remove the shunt and all connections to the 
physical laboratory of the Ohio state univer- 
sity, where a thorough examination of the 
cause of the difference could be made. ‘This 
was done ; but, before any experimental exami- 
nation had been undertaken, the origin of the 
difficulty suggested itself. The two short wires 
connecting the mercury-cups had been in one 
case thrown with the galvanometer doubtless, 
and in the other with the shunt. It was 
perfectly certain, however, that throughout the 
tests they had formed a part of the galvanome- 
ter. Upon examination, this explanation was 
at once found to be correct. ‘The shunt and 
galvanometer were connected up precisely as 
they had been in Cincinnati: and a series of 
twenty-five observations gave, when the small 
wires were a part of the shunt, a multiplier of 
4.60; and, when they formed a part of the gal- 
vanometer circuit, it was 5.01. The measure- 
ments were made by comparing the resistance 
of the two parts of the circuit by means of the 
fall in potential, as shown by a Thomson’s 
reflecting galvanometer of high resistance. 
While in use in Cincinnati, the shunt was con- 
stantly carrying a portion of the current; and 
its temperature was therefore always slightly 
higher than that of the galvanometer. The dif- 
ference was small, and it could not be measured 
accurately ; but, on account of its existence, 
it was thought proper to adjust the shunt mul- 
tiplier. An excess of heat in the shunt would 
throw a greater amount of the current through 
the galvanometer than would go there if the 
two were at the same temperature : accordingly, 
the value accepted was 4.9 instead of 5.0, as 
indicated by the comparison, in which the cur- 
rents used were much weaker than those trans- 
mitted during the tests. It will be observed, 
the existence of an excess of temperature in 
the shunt favors somewhat the system in 
which the stronger current was transmitted. 
In the measurements of electromotive force, 
Thomson’s potential galvanometer, by White, 
was used. In the beginning a large number of 
comparison observations were made, in which 
the same electromotive force was measured by 
this instrument and by the well-known method 
of discharging a condenser through a high- 
resistance galvanometer. 
A condenser of one-half micro-farad capacity, 
and a reflecting galvanometer of nearly seven 
thousand ohms resistance, both by Elliot Broth- 
