564 BELL SYSTEM TECHNICAL JOURNAL 



considered much more reliable than those determined by differences 

 (items 1, 5, 11). The degree of agreement between the two should 

 be considered more as a check on the accuracy of the individual 

 \alues from which the latter are derived than as a check on the 

 former. 



(j Hence, we may consider 3.3 mmf. (item 2, Table II) as the stray 

 capacitance for the 6 inch square and about 2.3 mmf. (item 12) for 

 the 43^ inch circle. This checks fairly well with the theoretical 



value of — (CGS units). If the square is considered equivalent to a 



TT 



circle of equal area the value of — is equivalent to about 3.03 mmf. 



TT 



For the 4I9 inch circle the value of — is 2.0 mmf. The measured 



' " TT 



values should be somewhat higher than the theoretical since the 

 shielded bridge and other apparatus comprise a considerable mass of 

 grounded metal at no great distance from the sample. 



The fringe effect for the equal circular electrodes may be compared 

 with values computed from Kirchhoff's formula 



^ r /, 167r(6 + 0^ , t , ^ + / o\ 



where C is the fringe effect and r and / are the radius and thickness 

 respectively of the electrodes and h is their separation, all in CGS 

 units. For very thin electrodes this reduces to 



^=f.O*ir-^)- 



Values for this expression reduced to a percentage correction are 

 listed as item 13 in Table II and are plotted in Fig. 3. It will be 

 noted that the observed effect is at least a third less than that com- 

 puted. It should be borne in mind, however, that Kirchhoff's formula 

 applies primarily to electrodes in air. To completely simulate this 

 condition with a solid dielectric would require that the electrodes be 

 completely surrounded by a considerable thickness of the dielectric. 

 If the difference noted above is due to lines of force which pass partly 

 through air and partly through the sample, this difference should be 

 greatest for a sample of high dielectric constant and diminish as the 

 dielectric constant of the sample approaches that of the air. It is 

 seen from Fig. 3 that this is the case, the curve for hard rubber having 

 a dielectric constant of 3 being nearer to the computed curve than 

 that for glass having a dielectric constant of 7.7. The anomalous 



