Characterization and modeling of low frequency noise in 0.13 µm BiCMOS SiGe:C heterojunction bipolar trasnsistors

Abstract

Low frequency noise (LFN) in 0.13 µm BiCMOS SiGe:C was characterized both as a function of base current bias IB and emitter area Ae. The LFN exhibits typical behavior of 1/f noise for frequencies up to 1 KHz after which the shot noise 2qI is visible, in transistors with large emitter area (Ae > 1µm 2 ). The 1/f noise is modeled following the SPICE compact model, and the LFN parameters KF and AF were calculated. The extracted figure of merit KB = KF * Ae, has an excellent value of 1.10 −10 µm 2 . The transistors with small emitter area (Ae < 1µm 2 ) can be affected by the presence of generation-recombination (G-R) components. In some cases, where generation-recombination presents large amplitude and high cut-off frequency, it is related to random telegraph signal (RTS).

Low-frequency noise in downscaled silicon transistors: Trends, theory and practice

“…Data for SiGeC HBTs (■) were treated separatedly as they show a clear improvement over the rest, with (A E ×K F ) avg = 1.25×10 −10 μm², 16.5dB below the range for npn HBT. Data for SiGeC are from [40,64,65,228,229,230]. In left-hand figure, the data are for npn SiGe HBTs from [224] processed in similar conditions ( ), and () from [35] and ( ) from [225] processed at variety of conditions, which resulted in scattering of data.…”

“…Rather, other factors, such as aforementioned bias, layout (especially proximity of trench isolation and polysilicon emitter overlap), SOI, annealing, etc., impact the 1/f noise in HBTs. While overall Ge does not impact the noise in SiGe HBTs, the addition of carbon (C) to the SiGe semiconductor does improves the K F =f×S I B /I B ² in SiGeC HBTs, with (A E ×K F ) avg = 1.25×10 −10 μm², 16.5dB below the range for npn SiGe HBTs (data shown with (■) in Figure 40 and taken from [40,64,65,228,229,230]). Also, as observed in [64], SiGeC HBTs have a much lower noise response compared with the HBTs from other technologies.…”

“…Nevertheless, the universality of thermal noise provided suitable reference, and the noise factor (or noise figure), NF -see eqs. (226) and (228), became a norm in measurement and reporting the results for 1/f noise, which continued to mid-1970's, and caused some inconvenience and delay to establish that the factors for 1/f noise are different from the factors that determine the white noise. Apart from the shape of the spectra, the main difference between white and 1/f noise is that the white noise can be described by electrical DC quantities, current for shot noise or resistance for thermal noise, while the 1/f noise is in addition inversely proportional to the size of statistical populations of carriers and/or traps.…”

“…Data for SiGeC HBTs (■) were treated separatedly as they show a clear improvement over the rest, with (A E ×K F ) avg = 1.25×10 −10 μm², 16.5dB below the range for npn HBT. Data for SiGeC are from[40,64,65,228,229,230].…”