Research On The Modeling And Parameter Extraction Of The On-wafer Transmission Line And Inductor For Millimeter Wave Integrated Circuit

Transmission lines and spiral inductors are two kind of widely used passive components in the millimeter-wave integrated circuit.A high-precision equivalent model is crucial for circuit simulation.This work focuses on modeling and parameter extraction of the on-chip transmission lines and spiral inductors for millimeter-wave applications.A set of coplanar waveguides(CPWs),grounded coplanar waveguides(GCPWs),and spiral inductors are fabricated on HLMC 40 nm RF CMOS process of Shanghai Huali Microelectronics Corporation.The open-short and L-2L method are used to de-embed the on-wafer RF measurement.The devices for electromagnetic(EM)simulation,which have the same structures and material parameters with the fabricated devices,are designed and simulated.The results of the EM simulation are validated by adjusting the dielectric constant,conductivity,etc.,in comparison with the above de-embedded measurement data.The parameters extraction data are produced by the validated EM simulation data.In this work,an empirical model is used to model the transmission lines and spiral inductors,and a parameter extraction and parameter optimization method was proposed to improve the model's fitting accuracy.The multiple regression method is used to get the scalable function of each element in the empirical model,and the form of the scalable functions are configurable.For the multi-parameter optimization of complex models,a genetic algorithm for multi-parameter optimization in modeling is designed and implemented.With the genetic algorithm,the multi-parameter optimization becomes efficient and accurate.With above methods,the problems during the millimeter-wave passive component modeling,such as lacking of versatility,narrow frequency band,and manual adjustment of parameters can be solved.The above methods are verified for CPW with the model scalable with length,signal line width,and signal line and sidewall space.The root mean square fitting errors of the CPW scalable model's RLGC parameters are all less than 6% in the 0~100GHz frequency range.These methods are applied to GCPW and spiral inductor,and a similar high-precision scalable model is also obtained.Compared with the results in the papers,the scalable models of CPW,GCPW and spiral inductor have the advantage of wide frequency range and high precision.The parameter extraction method,parameters optimization and model scaling are all programmed and the processes are automated.The scalable models of the CPW,GCPW and inductor,are all described in SPICE netlists.The netlists are verified by the Cadence.