The flash-arc in high-power valves

The ?flash-arc? phenomenon, sometimes called the ?Rocky Point effect?, is the spontaneous and complete breakdown of the high insulation normally afforded by a good vacuum between metallic electrodes. The breakdown does not, in general, follow immediately upon the application of the voltage, but is preceded by an interval or ?time-lag? of widely variable duration. Part I of the paper gives a general experimental survey of the properties, including visible effects, of the flash-arcs obtained by the discharge of condensers (capacitances 00001 to 0.0016 ?F) through inductances (3.1 to 40 ?H) in series with the discharge tube and with a device for measuring the peak current. The latter is a quick-acting peak-voltmeter responding in a very small fraction of a micro-second to the voltage developed across a non-inductive resistor. The completeness of the breakdown with the larger condensers is shown by the small damping of the oscillatory current. With the smallest condenser, however, the oscillation is very heavily damped. The maximum amplitude of the oscillations is generally much less than the value to be expected from the charging voltage. It follows that energy is expended in building up the arc path, and it is shown that 1 to 3 watt-seconds will give an arc of high final conductivity, whereas 0.2 watt-second will not. From the recorded values of condenser charge and current maximum a minimum estimate of the duration of the initial building-up stage is found to be 2 ? 10?7 to 4 ? 10?7 sec., which is in agreement with Snoddy´s value of 5 ? 10?7 sec. from observation in a rotating mirror. This is compared with the time of flight of positive ions, which approaches 10?7 sec. The discharge is easily transferred from one electrode to another; e.g. in a triode valve the grid and filament both take their share of the current. Residual gas in the tube only affects the flashing voltage indirectly. A considerable rise of gas pressure has no effect, but a subsequent clean-up of t- he gas on to the electrodes lowers the flashing voltage. Autelectronic currents from edges formed by fracture of the cathode also have little effect on the discharge voltage. When a flash-arc is artificially encouraged by pilot discharges across loose contacts it is found that the initial expenditure of energy is no longer demanded, and that the flashing voltage may be very low. The chief and most general visible after-effect in valves which have flashed is that widespread tree-like markings appear on the grid and filament supports, demonstrating the great mobility of the arc. The fine wires in the active part of the grid are, however, never damaged. Part II of the paper compares the incidence of flashing in operation at various installations. At Carnarvon, Daventry 5XX, and Motala, and at short-wave stations, it has seldom or never occurred; at Daventry 5GB (two circuits) and at Rugby GBR (two circuits) it occurred frequently in some circuits a few years ago, but is now rare. It is concluded that: (a) Direct paralleling of anodes is dangerous if it permits concentration in the flashing valve of currents exceeding 5 to 10 amperes. (b) The smaller the high-frequency output condenser, the safer is the circuit. (c) The larger the inductance between each valve and the high-tension supply, the safer is the circuit. (d) Individual anode resistances are a good safeguard, but they cannot be used for valves above the 10-kW size. (e) Resistances in the high-tension supply are useful. When several valves are used, subdivision of the circuit by providing separate input chokes and output condensers gives marked improvement. Recent larger valves of the 100-kW size have been found to compare well with the earlier smaller valves, although the circuit conditions are necessarily less advantageous. Some general comments on rectifier valves are made. Part III of the paper describes and discusses the time-lag. Short time-lags of the order of a few minutes are sensitive to voltage, a 3