A BRAUN TUBE HYSTERESIGRAPH 307 



d. For small values of AHc this correction is proportional to the 

 frequency of the magnetizing field, through the factor dH/dt. When 

 the correction becomes appreciable compared to Hm the effect con- 

 sidered under d becomes active and AHc increases at a slower rate 

 than the first power of the frequency. 



e. The value dB/dH is not constant as was assumed but depends 

 upon H and attains a maximum near H = He. The effect of eddy 

 currents is to make the magnetization non-uniform over the cross- 

 section of the sample, and the average value of dB/dH is then con- 

 siderably less than the maximum value. This effect again makes 

 the value of AHc increase at a slower rate than the first power of the 

 frequency and of Hm. 



f. When conditions are such that AHc is small, a sufficient approxi- 

 mation to its value is obtained by deriving dB/dH from the dynamic 

 loop so that the correction can be made without reference to the 

 static curve. 



Besides the eddy current effect, widening of the hysteresis loop has 

 also been ascribed to what has been called magnetic viscosity, a 

 property of the material which causes the induction to lag behind 

 the actual magnetizing force. The existence of this phenomenon 

 has been affirmed by some experimenters and denied by others. Not 

 enough is known about it to make its discussion here pertinent.'' 



6. NON-LlNEARITY OF THE AMPLIFIER. 



The push-pull construction of the amplifier compensates to some 

 extent for curvature of the tube characteristics. In order to have 

 full compensation, the tubes should be matched so that the tubes of 

 each pair work on a part of their characteristic having the same 

 steepness and curvature. 



Furthermore, the last pair of tubes, the tubes working into the 

 oscillograph, should have a practically straight characteristic over 

 about 30 volts on each side of the average plate voltage, or in the 

 case of 102-D tubes about 1.5 volts on each side of the average grid 

 voltage. 



7. Observational Errors. 



Because of the width of the fluorescent trace on the screen, there is 



a probable error of about .2 mm. in measuring the width or height 



of the hysteresis loop on the tube. This corresponds to probable 



errors of the order of 1 per cent in B and 2 per cent to 3 per cent in //,.. 



Probable errors of 1 per cent are also introduced in the calibration 



'For a recent consideration of time-lag in magnetization see R. M. Bozortii, 

 Phys. Rev., 32, p. 124, 1928. 



