FORCES ACTING IN MAGNETIC CIRCUITS. 201 
(c.) Scraped to a surface plate. 
(d.) Ground by emery wheels. 
(c.) Turned flat to a surface plate—this takes a little skill. 
(7) Optically ground by emery and diamond dust and finished 
with putty powder. 
This last requires a note. Of course the bars must be provided with 
shoes of many times their diameter to make the process a success, 
and these shoes must be of similar material to the bars. In order 
to save circumlocution I may state that both I and my assistant, 
Mr. Cook, are fairly expert at this kind of work and we met with 
no real difficulty. The use of diamond dust instead of emery . 
saves a little time but makes it more difficult to get a good result. 
I obtained two sets of bars with properly ground faces, one of 
these sets was of hard iron and was not so good as Brashear’s 
celebrated test plates, on account of a slight convexity on the part 
of one surface and a corresponding concavity on the part of the 
other. The other pair of bars were of soft Swedish iron well 
annealed ; they were less than 1 cm. in diameter and about sixty 
em. long. The surfaces were as good as the test plates—z.e. per- 
fect according to the present state of the art. I have little doubt 
they are as good soft iron surfaces as have ever been prepared. 
This means that there was no inequality comparable with a wave 
length of sodium light on either surface. 
The bars were kept straight and aligned by well fitting glass, 
or brass, or fibre tubes at the plane of contract. The fit was 
always just so good that no correction for friction was necessary. 
The magnetising current was measured by a Siemen’s Dynamo- 
meter which was compared with suitable members of a chain of 
Kelvin balances, it was found that in this dynamometer the read- 
ings were correct within the limits of accuracy of reading. The 
current was supplied by storage cells. 
The induction coils were wound on brass bobbins with proper 
precautions. The bobbins were of different diameters, and were 
compared ballistically and found to give identical results—hence 
it was concluded that they were all free from leakage errors. 
