The Cyclotron Resonant Maser (CRM) is a device in which electrons gyrating in an external magnetic field produce coherent EM radiation. A DC electron beam current must be converted to an AC beam current to create RF energy. There are two possible approaches: phase bunching (O-type) and spatial segregation (Mtype). In phase bunching, electrons are either accelerated or decelerated depending on when the electrons enter the interaction region, causing phase bunching. The electron bunches are then slowed down by the RF field for energy extraction. Not all electrons lose energy; some even gain energy. In spatial segregation, electrons entering the interaction region at different times are deflected in different directions. With an appropriate spatially varying RF field, all electrons can lose energy leading to very high conversion efficiency. A CRM with a smooth walled cylindrical waveguide interaction cavity and an annular electron beam passing through it can generate very large amount of RF energy. Depending on the electron beam position a gyrotron (O-type device) and a gyro-peniotron (M-type device) are possible. In this work, first, a nonlinear theory to study CRMs with a smooth walled cylindrical waveguide interaction cavity is presented. The nonlinear set of differential equations are linearized to study the starting conditions of the device. A gyropeniotron operating in the TE0,2 - mode is studied using the theory presented. It is found that a gyro-peniotron operating in a low order mode can be self excited without mode competition from gyrotron modes, leading to the possibility of a very efficient high power RF source. A higher order mode gyro-peniotron experiences severe mode competition from gyrotron modes. The cavity Q required for gyropeniotron operation is very high, which can lead to excessive heat in the cavity walls due to ohmic losses. Hence, a gyro-peniotron operation seems practical only in low order modes and in short pulses. Second, an existing linear theory of gyrotrons is extended to include effects of magnetic field tapering, cavity wall profile, finite beam thickness, velocity spread and axially dependent beam coupling to the fields of competing modes. Starting currents are calculated for the operating and the most dangerous competing mode in a 140 GHz gyrotron, which was developed at Communications and Power Industries (CPI). Start-up scenario of this device is also studied using the non-stationary code MAGY, which is a tool for modeling slow and fast microwave sources.