An old lamp was taken and wires soldered to its terminals, and also to the terminals of the socket into which it was screwed, and, by measuring as nearly simultaneously as possible the pressures on the lamp and on the socket terminals under various conditions as to heat and position of lamp, we proved that the resistance of the socket and contacts was never perceptible. There seemed, therefore, no reason to doubt the results already obtained; but as the accuracy with which the candlepower could be measured had increased with practice and with the introduction of the swinging frame, we decided after 110 hours of testing to start again with new lamps. The results obtained in these preliminary tests are shown in fig. 10a, the curves in fig. 10 being those obtained from the lamps subsequently tested. These curves show the variation of light and current with time for four groups of lamps between the terminals of which perfectly constant P.Ds. of 100, 101, 102, and 104 volts were maintained respectively, the full lines representing candlepower and the dotted lines current. The breaks in the curves connected by vertical lines indicate the times at which a lamp broke and was replaced by a new one, so that each group always consisted of three lamps. All these curves show that the light given out by these Edison-Swan lamps was greater after they had been glowing for some time than it was when the lamps were new; also that even just before the filament of one lamp in a group broke, the total light given out by the group of three was greater than when the lamps were new. This is a totally different result from that obtained in earlier tests, as illustrated in figs. 1, 2, 3, and 4, which showed that a considerable deterioration in candle-power always took place after lamps had been running for some time. Further, while the globes of earlier lamps were always much blackened, even after a run of a few hundred hours, and so became comparatively useless long before the filament broke, the Edison-Swan lamps which we have been testing showed hardly any blackening, even when the filaments lasted for over 1300 hours. The highness of the average candle-power of Edison-Swan lamps marked 100-8 after they have been run for some time at 100 volts, is corroborated by the figures in the following table, which gives the results of the measurements of the candle-power and the watts per candle of eleven 8-candle 100-volt lamps which had been run from the mains of the Notting-Hill Electric Lighting Co. As these lamps had been used in different rooms of a dwelling-house, it was Phil. Mag. S. 5. Vol. 39. No. 240. May 1895. 2 E Fig. 10 a.-Preliminary Tests on Edison-Swan 8 C.P. 100-volt Full lines represent Candle-Power and Time. impossible to say for how many hours they had been run, but the time probably varied from about 200 to 600 hours, and at any rate none of them were new. The measurements were taken at 100 volts, and Harcourt's pentane burner was the standard of light used. In these lamps, as in the others we tested, the blackening of the bulbs was very slight indeed. Since the above measurements were made, the two lamps whose candle-powers were 11.9 and 11.25 have broken after about another hundred hours of life without, as far as could be judged, any diminution in their light or any increase in the blackening of their bulbs. Moreover, although the rise of candle-power seen in figs. 10 and 10 a was always accompanied with a rise in current, yet, as the rise in candle-power was proportionally much greater than the rise in current for the lamps we tested, as will be seen from figs. 10 and 10 a, where one division represents 50 per cent. change in the light, but only 4 per cent. change in the current, the consumption of power per candle was actually less after the lamp had run for 50 hours than it was at the beginning, and the power per candle did not rise seriously during the whole life of a lamp. Examining the results for the group run at 100 volts, we see that in the first 124 hours the candle-power rose from 26.3 to 35, or by 33.1 per cent., the current in the same time rising from 1.261 to 1335 amperes, or by 5.87 per cent., so that during these first hours the watts per candle dropped from 4.79 to 3.82. During the 1820 hours of running, three lamps broke in this group, their lives being 572, 957, and 1167 hours; the three still unbroken at the end of the tests had been run for 1248, 863, and 653 hours respectively. Over the first 500 hours the average consumption of energy with the lamps run at 100 volts was at the rate of 4·0 watts per candle, over the second 500 hours 4:36 watts per candle, over the third 500 hours 4:55 watts per candle, and over the whole run the average efficiency was 0.23 candle per watt, corresponding with 4.35 watts per candle. The candle-power of the three lamps run at 101 volts rose in the first 150 hours from 26.5 to 30-6, that is by 15.5 per cent., the current rose from 1.27 to 1.33, or by 4.72 per cent., the watts per candle in the same time falling from 4.80 to 4:32. The average watts per candle over the whole run of 1340 hours of this group were 4.68, corresponding with an efficiency of 0.215 candle per watt. In the group at 102 volts which was run for 1820 hours six lamps broke, their lives being 242, 372, 516, 570, 516, and 786 hours respectively in the order of breaking; the three lamps unbroken at the end of the tests had run for 1148, 898, and 772 hours respectively. The initial rise in candle-power in this group at 102 volts was from 30.6 to 35, or a rise in 120 hours of 144 per cent., the current in the same time rose from 1.26 to 1.31 amperes, or by 3.97 per cent., so that the watts per candle diminished from 4.22 to 3.82. The average consumption of energy during the first 500 hours was at the rate of 3.9 watts per candle, during the second 500 hours 4:06 watts per candle, during the third 500 hours 4.2 watts per candle, and over the whole run the average was 4.12 watts per candle, corresponding with an efficiency of 0.243 candle per watt. Turning to the group run at 104 volts, the curves show that the breakages were very frequent. Altogether five lamps broke, their lives being 94, 175, 210, 242, and 254 hours; the three lamps left unbroken at the end of the test had been run for 165, 222, and 364 hours. The initial rise in candle-power was from 33.8 to 40.5 candles, that is 19.8 per cent.; and the current rose from 1.264 to 1.307 amperes, or by 3.4 per cent. Over the whole run the average watts per candle, for the lamps run at 104 volts, were 3.6, corresponding with an efficiency of 0.278 candle per watt. It is known from preceding tests (see the 'Electrician,' July 15th, 1892, for example) that the light given out by a new glow-lamp varies approximately as the seventh power of the pressure when the pressure is something like the normal pressure for the lamp. Now, as the groups of lamps that we tested at 100, 102, and 104 volts respectively were selected so that each group gave practically the same light initially when tested at the same pressure of 100 volts, it was to be expected that the three groups run at the three different pressures would follow the law for the light given out by the same lamp when used at different pressures. And this is practically the case, for if L100, L102, and L104 be the number of candles emitted by the three groups of three lamps each, we see from the curves on fig. 10 that at the start L100=263, L162=306, L104=32·5; L104 -1.24. L100 ... L1021·16, and L100 Hence the light is roughly as the seventh power of the pressure at the start. But after the lamps have been glowing for 100 hours this relationship no longer holds, for after 100 hours from the start we see from the curves on fig. 10 that L100=33.9, L102=341, L104=39.8, |