ROBUST DESIGN OF LOW NOISE MICROWAVE AMPLIFIERS

Abstract

In this thesis, the design of low noise microwave amplifiers was studied with special focus on the area of robust design of low noise amplifiers and the synthesis of impedance transforming networks of such amplifiers. Statistical techniques like design centring and yield optimisation are often used to increase the yield of microwave circuits without the need for post production tuning. However they are iterative in nature and do not address specific circuits. In the thesis, the most sensitive S-parameters and noise parameters that affect the transducer gain and noise figure of low noise amplifiers (LNAs) were identified. The effects of passive component variations such as changes in microstrip line widths and relative permittivity of the substrate on characteristic impedance of the microstrip line and effective relative permittivity were also studied. It was shown subsequently that the effect of transistor parameter variations on transducer gain and noise figure of LNAs is dominant. A circuit topology (variance reduction network) was proposed to minimise the effect or circuit performance variation due to transistor parameter variations. In the area of impedance transforming network synthesis, an alternative direct synthesis method was proposed. It is comparatively simpler than the analytical or real frequency approaches and also yielded topologies that are comparable with single frequency synthesis methods. Moreover, the proposed method provides gain and noise figure considerations over the prescribed bandwidth which the single frequency synthesis method does not account for. These two aspects were applied to the design of a 9.5-10.5 GHz LNA. Good experimental results were obtained. Yield simulations performed showed higher yield for LNAs designed with the variance reduction networks compared to those without. It also showed that the effect of transistor parameter variations on transducer gain and noise figure of LNAs is more important than the effect of passive component variations.