SIR w. THOMSON'S COMPASS. 677 



parameters g, h, k, do not affect the deviation of a ship on an even 

 keel; we may then annul the components P and Q and the other 

 parameters. Longitudinal and transverse magnets are arranged on 

 the deck, which separately compensate P and Q, and then bars of 

 soft iron, or boxes of chains, are added, which compensate the mag- 

 netisation induced by the earth. 



Ordinary compasses have long and highly magnetised needles, so 

 that it is necessary to keep the correcting instruments at such a dis- 

 tance that they produce on the needle an almost uniform field, and 

 thus avoid the reaction of the induced magnetism. We are thus led 

 to use very powerful magnets and great masses of iron. 



1238. In order to approximate more closely to the theory, Sir 

 W. Thomson uses, on the contrary, a series of small needles con- 

 nected by silk threads to a very light card of paper or of mica, so 

 that the whole does not weigh more than 30 grammes. Although 

 these needles are moreover very slightly magnetised, the moment 

 of inertia of the system is so small that the compass rapidly ac- 

 quires its position of rest, without having recourse to artificial means 

 as in compasses with liquid. The smallness of the needles and 

 their small magnetic moment render it possible to bring the 

 correcting parts much nearer ; the magnets are placed in the 

 compass box itself, and the soft iron is formed of two sym- 

 metrical spheres on the binnacle, at the two ends of a horizontal 

 diameter passing through the needle. 



The component Q is compensated by a transverse magnet, and 

 the component P by a pair of longitudinal magnets symmetrically 

 placed under the compass in respect of the vertical. The semi- 

 circular error is thus eliminated. 



The quadrantal error is got rid of by two spheres placed at a 

 convenient distance, and at right angles to the plane of symmetry 

 of the ship in the usual case, or in an oblique direction if the co- 

 efficient E is not null that is to say, if the quadrantal error itself 

 is oblique. This latter circumstance occurs when the compass is 

 not in the plane of symmetry, or that the distribution of iron on 

 board is not symmetrical. In this case there is a constant error 

 A independent of the direction of the ship, and which may be deter- 

 mined once for all, if it is not preferred to correct it by a simple 

 rotation of the binnacle. 



A vertical magnet, lastly, corrects the heeling error. 



The correction is easy when the ship can be swung in sight 

 of land, or with a clear sky, for the head can then be put in different 

 directions, which separately neutralise the several terms. 



