Design of RF and wireless packages using fast hybrid electromagnetic/statistical methods

Abstract

A method for the design of packaging elements using a combination of electromagnetic simulation and statistical modeling techniques is presented. In this paper, fite-ground coplanar waveguide (FG-CPW) lime coupling is addressed. The electromagnetic performance of two FG-CPW limes in parallel is determined with FDTD. The results of these simulations are analyzed using the design of experiments (DOE) and response surface modeling (RSM) techniques. The result is a statistical model that can he used to determine the lime geomeby based on performance requirements. Introduction Modem RF microsystems require increasingly higher levels of integration to create compact modules that incorporate the maximum amount of functionality. One method of accomplishing this is to place progressively mure circuit components in the package. These components are placed in very close proximity and their interaction is inevitable. Crosstalk and coupling between closely spaced lines can have a catastrophic effect on the performance of a circuit. It is imperative that this coupling he quantified so that its effect can be accounted for in design. One common type of line used in these circuits is the fite-ground coplanar waveguide (FG-CPW)(l). This paper explores the coupling between two finite-ground coplanar waveguides placed in parallel and quantifies the effect based on the circuit parameters using a methodology that integrates simulation and statistical tools. Two identical, parallel, finite-ground coplanar waveguides can he characterized by five parameters, the width of the signal line, the distance between the signal lime and the ground plane, the width of the ground plane, the spacing between the limes, and the height of the substrate. By comparing the performance of configurations generated by varying these parameters, the effect of each parameter can he determined. The number of cases required by varying these parameters can he large and their fabrication can be time consuming and expensive. In order to reduce this, these cases can be supplemented by simulation. By using an accurate simulator, such as the full-wave f~te-difference time-domain technique (FDTD)(Z), which has been shown to effectively model the topology performance in a wide frequency range(l). the parameter variation can he carried out numerically in a significantly quicker and inexpensive way. Statistical techniques can then he applied to the simulation results to develop accurate and highly efficient models that can predict performance based on any combination of parameters in a design space. The hybrid design procedure for this investigation begins with identifying the parameters to be varied and determining the design space, that is, the ranges for the variation. The design space of the parameters has to be chosen such it includes the fabrication value range while providing sufficient variation in performance. In the case of the parallel FG-CPW lines, the parameter of interest is the signal coupled to one line when the other is excited (parasitic crosstalk). The results of these FDlD simulations are then used in design of experiments (DOE)(3) and response surface modeling (RSM)(4) statistical methodologies in order to develop a model that explains the effect of each parameter on the performance of the circuit. This proof of technique is useful in a number of ways. First, it allows the interaction of circuit interconuects used in SOP and SOC geometries to be characterized in terms of their geometrical and material parameters in a swift manner. Furthermore, the technique can he applied to more complex wireless transceiver packages that can be extremely difficult to optimize, even in electromagnetic simulation. This allows the designer to optimize results from highly accurate electromagnetic simulation techniques with powerful statistical methodologies and apply them to highly-integrated RF structures using diakoptics approaches.