An analysis of the operation of resonant cavity magnetrons is made without regard to the mechanism of conversion of the d.c. input power into the r‐f output power. The magnetron is represented by its equivalent circuit and is assumed to be essentially in equilibrium at every point of its energy build‐up curve. The last assumption is shown to be reasonable on the basis of measured starting times. From the data of load impedance charts and the rise of the r‐f pulse, the relation between power generated and vane voltage is deduced. By considering the division of the energy between that stored in the resonant system, and that dissipated in the load, the law of build‐up is derived. The dependence of starting time on load is calculated and agrees with experiment when the extra energy stored in the connecting line is taken into account. The starting time is affected slightly by the initial noise level and becomes infinite below a minimumQ. For highQvalues the starting time can be varied only by changing the energy stored in the line, which may enable one to design a line for ``mode'' suppression.