TABLE OF CONTENTS The noise modeling of III-V HBTs is still a topic of research. While the basic mechanisms are quite well known, it is in most cases no problem to resimulate a certain measurement, especially in the small-signal domain. However, predicting the noise behavior is not as easy, and the large-signal models available cannot be considered to be satisfying in any way.
The main questions concern the correlation of the shot-noise sources, and how to describe the low-frequency noise under large-signal excitation. While the first yields a reduction of modeling accuracy at higher frequencies, the second prevents for example reliable simulation of oscillator phase noise.
The review of noise theory is necessarily short. Detailed introduction into the topic is found in the literature [1, 2].
Noise is observed when electrical properties of a sample show fluctuations around a mean value. For example, the random motion of electrons inside a resistor yields voltage fluctuations at the resistor's ports. In semiconductor materials, the limited number of carriers leads to additional noise sources because their number fluctuates due to generation and recombination, or since a macroscopic continuous current is caused by a limited number of individual current pulses (e.g., the current across a diode junction). Noise sources commonly are classified by their frequency dependence. Mainly two different categories exist: white noise that shows a spectrum constant with frequency, and low frequency noise that is also often referred to as 1/ f noise. Of course, neither white noise nor 1/
