37$ Dr. Norman Campbell : Experiments on 



may be its cause ; but it should be noted that the rotating 

 armature machine had unlaminated pole-pieces ; in most 

 modern machines the pole-pieces are laminated. 



The Peak Potential. 



16. So far the experiments have been preliminary. The 

 main object of the investigation is to determine the peak 

 potentials given by the various combinations of armatures. 

 These were determined as in the previous part, a primary 

 current of 0*5 amp. being broken ; the capacity added to 

 the secondary circuit for purposes of measurement was 

 37*4 mmf., in place of 11'8'mmi. 



In fig. 9 the resulting peak potentials are plotted against 

 C x for the following combinations :— 10,000 / 0, 8000/0, 

 6000/0, 8000/2000, 8000/4000, 8000/6000, 6000/4000, 

 6000/8000. On each curve is given the corresponding 

 value of c, calculated as explained in § 14. 



It should be noted at the outset that the divergence 

 of the experimental points from the smooth curves drawn 

 through them exceeds greatly the possible experimental 

 error of a measurement; it is due to a real variability of 

 the peak potential being measured, and not to a deficiency 

 of the method of measurement. The best proof of this 

 statement is to be found in figs. 12, 13 (which will be 

 discussed presently), in which the method of measurement 

 was exactly the same but the divergence of the points 

 from a smooth curve very much less. Again, while the 

 general shape of the curves, and even such minor features 

 as could be detected with certainty, could be repeated at 

 successive trials, the absolute values of the peak potential 

 were found to vary considerably. Thus experiments were 

 made six times on 10,000/0 ; the extreme values differed 

 5 per cent, from the mean, although on each occasion the 

 variation of the peak potential with Ci was almost exactly 

 the same. The series of measurements selected for fig. 9 

 are believed to be closely comparable, but the ratio between 

 the absolute values for any two curves cannot be relied upon 

 to less than 3 per cent. 



The reason for this variability of the peak potential has 

 not been discovered, but it is clearly connected with the 

 losses in the iron core. It has always been noticed in 

 all experiments on magnetos, and was not mentioned 

 in the previous part because the method of measurement 



