568 BELL SYSTEM TECHNICAL JOURNAL 



about the niininiuni depth of mercury that can be depended upon to 

 cover the surface of the sample and till the angle between the sample 

 and the ring. This, of course, has slightly less effect than if the entire 

 surface of the electrode were raised as is the case when mercury is 

 poured into the ring, but the difference is probably negligible. The 

 procedure was the same as previously described for the metal ring. 

 The lower foil electrode covered the entire surface of the sample. 

 The upper foil is the same diameter as the inside of the insulating 

 ring. (In the case of the metal ring the foil was the same size as 

 the outside diameter of the ring at the lower edge). The capacitance 

 is measured first between the foils alone and then with the ring super- 

 imposed on the upper foil. The change should represent very closely 

 the increased edge effect due to a depth of Y^' of mercury in the ring. 



The results of this test are tabulated in Table III. It will be 

 noticed that the change in capacitance due to the beveled rubber 

 ring is about the same as for the metal ring, while the change due to 

 the square edged ring is materially greater than for the metal ring. 

 The fiat phenol fibre ring produces a change in capacitance two or three 

 times as great as the metal ring and the apparent power factor for 

 the rubber sample is more than doubled. This, of course, is due to 

 the dielectric loss in the ring itself and therefore the lower the power 

 factor of the sample under test the greater is the proportional error. 

 While the rubber rings had no appreciable effect on the power factor 

 of the samples tested, it is believed that in the case of very low loss 

 materials such as fused quartz the effect might still be appreciable. 

 Since the insulating rings under the best conditions are no better 

 than a metal ring and under poor conditions are very much inferior, it 

 is believed that in general metal rings will be found more satisfactory 

 for confining mercury electrodes. 



Thus far we have considered the experimental data primarily with 

 regard to the capacitance and dielectric constant. Table IV shows 

 the power factors computed from the conductance readings corre- 

 sponding to readings 1, 4, 6 and 10 in Table I. These values were 

 computed from the capacitance and conductance values for the 

 various types of electrodes and without correction for edge effects or 

 ground capacitances. The values for hard rubber illustrate fairly 

 well the variation for different electrodes which would be expected 

 on the basis of the preliminary discussion. The value obtained with 

 the 43/^" circle and guard ring should be the true value. The other 

 values should be slightly lower on account of the additional capaci- 

 tance without corresponding power less due to the flux which passes 

 partly and entirely through air. In general these variations are small 



