A convenient way is to use an oil condenser of very small capacity, consisting of two small adjustable metal plates, in connection with this and similar experiments.  In such case I take wires rather short and set at the beginning the condenser plates at maximum distance.  If the streams for the wires increase by approach of the plates, the length of the wires is about right; if they diminish the wires are too long for that frequency and potential.  When a condenser is used in connection with experiments with such a coil, it should be an oil condenser by all means, as in using an air condenser considerable energy might be wasted.  The wires leading to the plates in the oil should be very thin, heavily coated with some insulating compound, and provided with a conducting covering—­this preferably extending under the surface of the oil.  The conducting cover should not be too near the terminals, or ends, of the wire, as a spark would be apt to jump from the wire to it.  The conducting coating is used to diminish the air losses, in virtue of its action as an electrostatic screen.  As to the size of the vessel containing the oil, and the size of the plates, the experimenter gains at once an idea from a rough trial.  The size of the plates in oil is, however, calculable, as the dielectric losses are very small.

In the preceding experiment it is of considerable interest to know what relation the quantity of the light emitted bears to the frequency and potential of the electric impulses.  My opinion is that the heat as well as light effects produced should be proportionate, under otherwise equal conditions of test, to the product of frequency and square of potential, but the experimental verification of the law, whatever it may be, would be exceedingly difficult.  One thing is certain, at any rate, and that is, that in augmenting the potential and frequency we rapidly intensify the streams; and, though it may be very sanguine, it is surely not altogether hopeless to expect that we may succeed in producing a practical illuminant on these lines.  We would then be simply using burners or flames, in which there would be no chemical process, no consumption of material, but merely a transfer of energy, and which would, in all probability emit more light and less heat than ordinary flames.

[Illustration:  Fig. 8.—­Effect produced by concentrating streams.]

The luminous intensity of the streams is, of course, considerably increased when they are focused upon a small surface.  This may be shown by the following experiment: 

I attach to one of the terminals of the coil a wire w (Fig. 8), bent in a circle of about 30 centimetres in diameter, and to the other terminal I fasten a small brass sphere s, the surface of the wire being preferably equal to the surface of the sphere, and the centre of the latter being in a line at right angles to the plane of the wire circle and passing through its centre.  When the discharge is established under proper conditions, a luminous hollow cone is formed, and in the dark one-half of the brass sphere is strongly illuminated, as shown in the cut.