When using the ordinary ten-inch induction coil, and when the capacity charged by it does not exceed a small fraction of a microfarad, it is quite sufficient to employ brass or steel balls separated by a certain distance in air, at the ordinary pressure, as the arrangement of the discharger. When, however, we come to deal with the discharges of very large condensers, at high electromotive forces, then it is necessary to have special arrangements to prevent the destruction of the surfaces between which the spark passes, or their continual alteration, and many devices have been invented for this purpose. The author has devised an arrangement which fulfils the above conditions very perfectly for use in large power stations, but the details of this cannot be made public at the present time. * * * * * We have to consider in connection with this part of the subject the dielectric strength of air under different pressures and for different thicknesses. It was shown by Lord Kelvin, in 1860, that the dielectric strength of very thin layers of air is greater than that of thick layers.[18] The electric force, reckoned in volts per centimetre, required to pierce a thickness of air from two to ten millimetres in thickness, at atmospheric pressure, may be taken at 30,000 volts per centimetre. The same force in electrostatic units is represented by the number 100, since a gradient of 300 volts per centimetre corresponds to a force of one electrostatic unit. It appears also that for air and other gases there is a certain minimum voltage (approximately 400 volts) below which no discharge takes place, however near the conducting surfaces may be approximated. In this particular practical application, however, we are only concerned with spark lengths which are measured in millimetres or centimetres, lying, say, between one or two millimetres and five or six centimetres. Over this range of spark length we shall not generally be wrong in reckoning the voltage required to produce a spark between metal balls in air at the ordinary pressure to be given by the rule: _Disruptive voltage_ = 3,000 × _spark-gap length in millimetres_. If, however, the air pressure is increased above the normal by including the spark balls in a vessel in which air can be compressed, then the spark length, corresponding to a given potential difference, very rapidly decreases. Mr. F. J. Jervis-Smith[19] found that by increasing the air pressure from one atmosphere to two atmospheres round a pair of spark balls he reduced the spark length given by a certain voltage from 2·5 to 0·75 centimetre.