A new method based on the lumped-element network representation of the pad-set parasitics is developed to extract the
intrinsic drain current noise source and gate resistance from raw measurement data instead of direct de-embedding. The
length dependence of BSIM noise model is also corrected using a sub-circuit in the model file. With the new method, we
can finally integrate an improved and hardware verified noise model into design kits.
We present a comprehensive study of low-frequency noise mechanisms in 210 GHz SiGe HBTs using a variety of measurement techniques, and explain a unique scaling effect. The implication of these noise mechanisms on SiGe HBT compact modeling methodologies are also discussed.
We present a comprehensive investigation of the fundamental differences in low frequency noise behavior between <i>npn</i> and <i>pnp </i>SiGe HBTs. Geometry effects on the low frequency noise are assessed, as well as the impact of interfacial oxide(IFO) thickness on <i>pnp</i> noise characteristics. Temperature measurements and ionizing radiation are used to probe the fundamental physics of 1/f noise in<i> npn</i> and <i>pnp</i> SiGe HBTs. The<i> npn </i>transistors show a stronger size dependence than the <i>pnp</i> transistors. The 1/f noise for <i>pnp</i> SiGe HBTs exhibits an exponential dependence on IFO thickness, indicating that IFO produces the main contribution. In most cases, the magnitude of the 1/f noise has quadratic dependence on the base current(I<sub>B</sub>), the only exception being for the post-radiation <i>npn</i> transistor biased at low base currents, which exhibits a near-linear dependence on I<sub>B</sub>. In the proton radiation experiments, the <i>pnp</i> devices show better radiation tolerance than the <i>npn </i>devices. The observed temperature dependence for both types is quiet weak, consistent a tunneling mechanism.