A Frequency-Domain, Volterra Series-Based Behavioral Simulation Tool for RF Systems

Iason Vassiliou
(Professor Alberto L. Sangiovanni-Vincentelli)
(SRC) 98-DC-324-010

To accelerate the design of analog and mixed-signal systems, we have proposed a top-down, constraint-driven design methodology for analog circuits[1]. The expansion of our design methodology in the RF domain, requires system-level behavioral simulation tools that capture all the important block-level characteristics of RF building blocks, without making any assumption about their implementation. Typically, performance specifications for an RF system and its building blocks, are given in the frequency domain.

Using the Volterra theory of nonlinear systems[2] and the theory of non-linear, time-variant systems, we have developed a purely frequency-domain behavioral simulator that models both weak nonlinearities and periodically time-varying nonlinearities. The behavioral simulator uses high-level frequency domain models extracted from frequency-domain specifications or from transistor-level simulations, to calculate performance metrics such as second and third order intermodulation, noise figure, 1 dB compression point, and total signal-to-noise-and-distortion-ratio, for an arbitrary topology of interconnected blocks. It includes three parts: (1) a library of RF modules and translation of specifications or simulation results to Volterra n-port kernels, (2) a user interface so that the user can give an arbitrary interconnection of RF blocks with variable parameters, and (3) a solution engine to compute figures of merit needed.

The RF system-level simulation is being done in three steps: first, frequency domain RF block-level performance metrics, such as second and third order intermodulation performance, conversion gain and spurious effects, filtering characteristics, phase noise, and noise figure are being mapped onto generalized Volterra kernels and noise sources. The generalized Volterra kernels are the ones of the "sideband" transfer functions, i.e., the transfer functions that take into account the frequency translation effects of periodic strongly nonlinear excitations. In this way, strongly nonlinear modules such as mixers can be modeled. Second, the generalized Volterra kernels are being used to formulate a harmonic-balance, nonlinear system of sparse equations that is being solved using Newton-Raphson iteration and sparse-matrix techniques. So far we have implemented the simulation of the deterministic effects. A typical flow of the overall behavioral simulation methodology is shown in Figure 1. Figure 2 showsa typical output spectrum for a node.

 
Figure 1: Behavioral Simulator Flow Diagram

Figure 1: Behavioral Simulator Output

[1]
Henry Chang, Edoardo Charbon, Umakanta Choudhury, Alper Demir, Eric Felt, Edward Liu, Enrico Malavasi, Alberto Sangiovanni-Vincentelli, Iasson Vassiliou, ``A Top-Down, Constraint-Driven Design Methodology for Analog Integrated Circuits'', Kluwer Academic Publishers , Boston MA, 1997
[2]
M. Schetzen, "The Volterra and Wiener Theories of Nonlinear Systems",John Wiley & Sons, 1980.
Send mail to the author : (jvassil@eecs.berkeley.edu)