2-B] EFFICIENCY. 5 1 



multiplying be by S, gives watts eddy current loss; multiplying db 

 by S gives watts loss in hysteresis, friction and windage. 



If the eddy current loss were zero, ac would coincide with the 

 horizontal line ab; the first equation in 16 would become W = AS, 

 showing that IV 9 would be proportional to speed and the W 9 curve 

 in Fig. 2 would become a straight line. 



18. A Convenient Approximation. Since the eddy current loss 

 is commonly only a small part of the total rotation losses, for small 

 changes in speed it is nearly correct and often very convenient to 

 say the rotation losses W are directly proportional to speed. 



19. Further Separation of Losses. Hysteresis loss can be ap- 

 proximately separated from friction and windage by additional runs 

 at other field excitations. Friction and windage can not be separated 

 from each other by any simple means and hence are considered 

 together. There are various graphical and analytical methods for 

 separating losses, all based on the following facts: friction and 

 windage losses vary as first power of speed and are independent of 

 flux density; eddy current loss varies as square of speed and square 

 of flux density; hysteresis loss varies as first power of speed and 

 1.6 power of flux density. At any one speed, armature voltage is 

 taken as a measure of flux density. In any of these methods it is 

 necessary to make some assumption or approximation ; for this rea- 

 son the graphical methods are superior. (In the graphical method 

 given below the approximation consists in obtaining Oo by extra- 

 polation to zero excitation.) 



The analytical methods will not be taken up here; they consist in 

 obtaining several equations (based upon the above relations) and 

 eliminating between them after substituting numerical values obtained 

 from observation of W at various speeds and flux densities. 



20. Graphical Method. Various graphical methods for separat- 

 ing losses differ chiefly in detail; the following procedure (either 

 a or &) is suggested: 



(a) Make a series of no-load runs, as already described, at vari- 

 ous field excitations, extending these to as low a field excitation as 

 possible. Plot results as in Fig. 2, obtaining a series of curves 

 (straight lines) ac with intercepts Oa,, Oa v Oa v etc., corresponding 

 to various field currents. It is desired to find a value for an inter- 

 cept Oa for the supposed case of zero field current, for which of 



