SHUDDEMAGEN. — DEMAGNETIZING FACTORS FOR IRON RODS. 207 



of the day. With the reversal method no attempt to read the zero 

 was made, but instead a number of throws were taken alternately in 

 the plus and minus directions, and then averaged. These throws often 

 agreed regularly to about 1 part in 1000, when taken with a little 

 care. The reversal method, however, has a possible error due to the 

 time-constant of the primary circuit being comparatively large when 

 there is much iron in the solenoid S, and also to the slow establish- 

 ment of the magnetism in a thick iron rod. This was counterbalanced 

 by making the complete period of the astatic system about 25 seconds, 

 and finally 31 seconds. 



The step-by-step method was used only in one series of experiments 

 with the first solenoid >S'. This method is much harder to carry 

 through successfully, especially since the battery £J must maintain its 

 voltage without appreciable drop while furnishing an increasing cur- 

 rent for about half an hour, and the zero reading must be taken care- 

 fully every little while. Usually several curves were obtained for each 

 length of the iron rod used, so that a good average curve could be 

 constructed. As is well known, the two methods do not give the same 

 magnetization curve, the one by the step method usually, but not 

 always, lying below the reversal method curve. 



The iron rods tested in the first solenoid were all of soft Bessemer 

 steel, six feet long and of diameters ranging from 0.2381 cm. (= ^^ 

 inch) to 1.270 cms. (= ^ inch). The secondary coils consisted of 

 from 30 to 400 turns of fine insulated wire wound directly over the 

 middle of the rod. It was found necessary to reverse the magnetism 

 about six times before reading the actual throws, otherwise the read- 

 ings come out too low. After sufficient data had been collected to 

 construct a curve, equal lengths of the rod were cut off from each end, 

 so as to reduce nt from one value to the next. The ends of the rod 

 were then filed smooth and plane. Then a curve was obtained for the 

 shortened length of the rod. 



After proper reduction of the observations, the magnetization curves 

 B vs. H' were carefully constructed for all the m's used, on a large 

 sheet of millimeter paper of the dimensions 43 X 53 cms. 



The next problem was to devise some means of getting at the normal 

 curve (m = cc ). In the earlier part of the investigation frequent use 

 was made of the principle which leads to Du Bois's experimental formula 

 iVm^ = 45^ when m ^ 100. It was found that so long as B did 

 not exceed the value 8000, the formula was fairly well satisfied for 

 m ^ 150, provided only one system of simultaneous equations was 

 used. That is, supposing we had plotted out the actual magnetization 

 curves for m = 300, 250, 200, and 150. If we take all these into 



