CMOS RF circuit design has been an ever-lasting research field. It gained so much attention since RF circuits have high mobility and wide band efficiency, while CMOS technology has the advantage of low cost and better capability of integration. At the same time, IC circuits never stopped scaling down for the recent many decades. Reliability issues with RF circuits have become more and more severe with device scaling down: reliability effects such as gate oxide break down, hot carrier injection, negative bias temperature instability, have been amplified as the device size shrinks. Process variability issues also become more predominant as the feature size decreases. With these insights provided, reliability and variability evaluations on typical RF circuits and possible compensation techniques are highly desirable. In this work, a class E power amplifier is designed and laid out using TSMC 0.18 µm RF technology and the chip was fabricated. Oxide stress and hot electron tests were carried out at elevated supply voltage, fresh measurement results were compared with different stress conditions after 10 hours. Test results matched very well with mixed mode circuit simulations, proved that hot carrier effects degrades PA performances like output power, power efficiency, etc. Self- heating effects were examined on a class AB power amplifier since PA has high power operations. Device temperature simulation was done both in DC and mixed mode level. Different gate biasing techniques were analyzed and their abilities to compensate output power were compared. A simple gate biasing circuit turned out to be efficient to compensate selfheating effects under different localized heating situations. iv Process variation was studied on a classic Colpitts oscillator using Monte-Carlo simulation. Phase noise was examined since it is a key parameter in oscillator. Phase noise was modeled using analytical equations and supported by good match between MATLAB results and ADS simulation. An adaptive body biasing circuit was proposed to eliminate process variation. Results from probability density function simulation demonstrated its capability to relieve process variation on phase noise. Standard deviation of phase noise with adaptive body bias is much less than the one without compensation. Finally, a robust, adaptive design technique using PLL as on-chip sensor to reduce Process, Voltage, Temperature (P.V.T.) variations and other aging effects on RF PA was evaluated. The frequency and phase of ring oscillator need to be adjusted to follow the frequency and phase of input in PLL no matter how the working condition varies. As a result, the control signal of ring oscillator has to fluctuate according to the working condition, reflecting the P.V.T changes. RF circuits suffer from similar P.V.T. variations. The control signal of PLL is introduced to RF circuits and converted to the adaptive tuning voltage for substrate bias. Simulation results illustrate that the PA output power under different variations is more flat than the one with no compensation. Analytical equations show good support to what has been observed.