Modeling and Analysis of High Frequency Noise in Bicmos Transistors
The importance of high frequency noise performance is increasing in advanced bipolar and complementary metal-oxide semiconductor (BiCMOS) technologies because of the high demands of radio frequency (RF) and mixed-signal integrated circuits used in the 5G communication, automatic-driving sensors and internet of things (IOT) applications. While the characterization and modeling of high frequency noise of BiCMOS transistors have been a topic for many years, some important issues have not been clarified. For example, the noise correlation is not well predicted for bipolar devices, the excess noise factor is not well understood for MOSFET devices and the temperature dependence of high frequency noise in BiCMOS devices is not well studied. Focused on these issues, this research establishes the approach to extract the noise transit time from the high current compact model (HICUM), demonstrates an efficient methodology for high frequency noise prediction for silicon-germanium heterojunction bipolar transistors (SiGe HBTs) and validates the prediction methodology over size, bias and temperature. One of the issues of high frequency noise modeling of bipolar transistors is the noise correlation effect. This research explores the physical origin of high frequency noise correlation, studies the noise model of SiGe HBTs and creates an approach to extract the noise transit time from the HICUM compact model. The extracted noise transit time is validated by the tuner-based noise measurement results of sample SiGe HBTs by comparing the four noise parameters between the calculated and measured data over transistor size, bias and temperature. The results show that the noise transit time can be independent of frequency but dependent on bias and temperature. Furthermore, a complete high frequency noise prediction system is established. Based on the extraction methodology of the noise transit time from the HICUM model, this dissertation demonstrates a low-cost and time-friendly methodology to predict the full high frequency noise properties of the bipolar devices directly from the S-parameter measurement, DC measurement and the parameters from the HICUM model without the tuner-based noise measurement. Compared with the conventional tuner-based noise measurement, this methodology can save the measurement time as well as achieve a good accuracy. For MOSFET devices, the importance of excess noise factor is increasing with the transistor size scaling down to sub-100nm for high frequency noise modeling, but it has not been well studied so far. This research analyzes the excess noise factor based on the device physics and characterization results, investigates the noise sources contribution and models the high frequency noise based on Y-parameter methodology.