so that the light varies only as something like the fourth power of the pressure.

On the other hand, if we consider the results of the liminary tests recorded in fig. 10a, we have at the start L10024-8, L102=28.7, L104=30,

pre

so that

L104
L100

= 1.21;

100

while after 100 hours' run

L100 331, L102=38, L101 = 43.3;

=

so that

In this case, then, the law of the seventh power holds roughly not only for the light given out by the three groups at the start but also at the end of the first 100 hours' run.

The following table gives the analysis of the chief results. shown by the curves on figs. 10 and 10 a.

Although, as already stated, the number of lamps that could be supplied with power every night at the Central Technical College was limited to nine, and although, therefore, the total number dealt with in the investigation was small, the curves on figs. 10 and 10a, and the results given in the preceding table, are quite sufficient to enable us to arrive at the following results :

(1) When a group of Edison-Swan lamps marked 100-8 are run at 100 volts, and each lamp, as its filament breaks, is replaced by a new Edison-Swan 100-8 lamp, it may be expected that the light given out by the group will never be as small as it was at the beginning, when all the lamps in the group were new. This important result is brought about by the deterioration of the lamps with long-lived filaments being compensated by the great rise in the light given out by each new lamp when put in place of one whose filament has broken.

(2) An Edison-Swan lamp marked 100-8 when run at 100 volts will give an average illumination during its whole life of about 10 candles, and will absorb an average power of about 4:3 watts per candle, so that such a lamp must be regarded as a 43-watt lamp, and not a 30-watt lamp as is not unfrequently stated, this difference in power being about 43 per cent. (3) An Edison-Swan lamp marked 100-8 may, when run at 100 volts, emit during a large portion of its life a light of as much as 117 candles, and absorb a power of about 44-6 watts, which is about 44 per cent. greater than the nominal 30 watts.

These last two facts are of great importance to dynamoconstructors, when specifying, as is frequently done, the number of lamps of a given type to which a given dynamo can supply current to without becoming too hot. For if, in making such a calculation, it be assumed that an EdisonSwan 8-candle-power lamp absorbs the nominal 30 watts, that is 3.75 watts a candle instead of the 44.6 watts which our tests show that such a lamp actually absorbs during a long portion of its life, the current will be 44 per cent. greater, and the rate of heating of the dynamo 108 per cent. greater, than was anticipated.

(4) Groups of new Edison-Swan lamps marked 100–8, if selected so as to give the same light at 100 volts, will, when new, emit a light which is roughly proportional to the seventh power of the pressure applied to them. But after a run of 100 hours, this rule connecting light and pressure may, or may not, hold.

The rise in candle-power with time, which occurred during the early part of the life of all the Edison-Swan lamps which we tested, may be noticed also in the collection of curves published by Mr. Feldman in the Electrician,' Nov. 29th, 1892 (fig. 11). But there are certain very important differences between the results recorded by these curves and those

which we have obtained. These differences are shortly as follows:

(1) The greatest rise of light emitted by the lamps as re-
corded in Mr. Feldman's curves was 14 per cent. In
our experiments this maximum rise was 45 per cent.
(2) In spite of the first rise in the candle-power as shown
in Mr. Feldman's curves (fig. 11), the candle-power at
the end of the life of the lamps was in all cases much
less than it was when the lamps were new. Whereas

with the lamps which we tested at 100, 101, and 102 volts the light given out by a group was never as low as it was at the beginning when the lamps were new. (3) In spite of the first rise in candle-power recorded in the curves on fig. 11, it was only for the lamps whose behaviour was recorded in curves P and Z that the power expended per candle diminished at first with time. And even in the case of these two sets the power expended per candle increased again, and was greater when the filaments broke than it was when the lamps were new. Whereas in our tests the power expended per candle not only diminished considerably during the early life of the lamp, but it never rose again as high as it was when the lamps were new. Although, then, the rise in candle-power during the early part of the life of a glow-lamp is apparently not an absolutely new fact, the magnitude of the rise and the effects resulting from it were, in the older lamps, so trifling that no special attention seems to have been devoted to this important subject in the former reports of tests of lamps. Indeed, even in a prominent book connected with glow-lamps which has quite recently been brought out in this country, no mention whatever is made of the fact in question. And yet, as we have already shown, and as will become more apparent from what follows, this remarkable rise in the candle-power during the earlier part of the life of an Edison-Swan lamp has a very important effect on the economy of lighting with glow-lamps.

The next point toconsider is the way in which the cost of lighting with the modern 100-8 Edison-Swan lamps depends on the pressure at which they are run, and on the cost of a Board of Trade unit. The curves on figs. 12 and 13 give the cost of obtaining light with the various groups of lamps, including the cost of replacing lamps with broken filaments, calculated day by day from our tests as the experiment

went on.

The ordinates of the curves show the cost of 100 candle-hours at any time during the test, time being measured horizontally, and the values of the ordinates have been calculated from the following formula :

Total cost per candle per 100 hours at the end of any number of hours t equals

100 x

+100 ×

Board of Trade units used during time t × Price of one Board of Trade unit
Total Candle-hours during time t

Number of lamps used during time t x Price of one lamp
Total Candle-hours during time t

4.0

Time in Hours.