新穎的金氧半電晶體雜訊模型與應用於超寬頻系統低雜訊放大器之設計

Abstract

我們已經發展出新穎的微帶線結構用來直接量得NFmin而不需要複雜的校正手續(de-embedding)，以取代傳統的CPW結構。在10GHz、0.18μm MOSFET 8 gate fingers條件下，非常低的NFmin，0.9dB，可以直接量測得而不需要任何的校正。在精準的雜訊量測結果為基準下，我們發展出新穎的金氧半電晶體雜訊模型可以來預測元件的雜訊表現與特性。此外,我們修改了窄頻低雜訊放大器所使用的源極電感回授匹配方式，並且將它應用於設計超寬頻低雜訊放大器。該放大器採用台積電0.18微米製程，在3~10GHz的範圍裡達到輸入與輸出的阻抗匹配並提供10dB的功率增益。在1.8V的供應電壓下消耗27mW的功率，而第一級放大及僅銷耗15mW。在這篇論文中，我們將會說明如何修改源極電感回授匹配及其原理，以及如何將之應用於超寬頻低雜訊放大器中。

A novel micro-strip line layout is developed to direct measure the min. noise figure (NFmin) accurately instead of the complicated de-embedding procedure in conventional CPW line. Very low NFmin of 0.9 dB at 10 GHz is directly measured in 8 gate fingers 0.18μm MOSFETs without any de-embedding. Based on the accurate NFmin measurement, we have developed the novel NFmin model to predict device noise characteristics. Besides, we also designed an UWB LNA with Modified Source L-degenerate by using TSMC 0.18μm technology. The LNA provides a forward gain (S21) of 10dB over the 3 ~ 10 GHz range with a low noise figure of 3.5dB (at 6 GHz) while consuming 27mW from 1.8V power supply. To achieve its wide-band characteristics, a novel input matching mechanism is proposed, which modifies L-degenerate approach for the wide-band matching. We will present a detail analysis of this LNA architecture.