Wave digital modeling provides a framework for translating analog circuit structures into digital implementations using incident and reflected waves. For engineers working with nonlinear circuit models, its treatment of three-terminal devices offers a useful case study in balancing model flexibility, numerical solvability and computational cost. The approach described here was developed for virtual analog audio applications, but its underlying modeling choices are relevant to anyone evaluating how nonlinear components can be represented within a modular simulation architecture.
How wave digital modeling represents a circuit
In a wave digital filter (WDF), each circuit port is described by voltage and current alongside two wave variables: an incident wave and a reflected wave. A port resistance links these quantities. Circuit elements become blocks with scattering relations, while connections between elements are represented by adaptors.
This structure separates component models from circuit topology. Dynamic elements such as capacitors and inductors can be discretized using the trapezoidal rule, preserving important energy-related properties of their analog counterparts. Nonlinear elements require an additional step: their port variables must satisfy a nonlinear device relationship as well as the networkâs Kirchhoff constraints.
Three-terminal devices create a particular modeling challenge. A transistorâs behavior is generally described by two independent variables, but its three terminals can be connected in multiple ways within a circuit. A model limited to one port configuration may not fit every topology efficiently.
A flexible port model for three-terminal devices
The proposed framework starts with a six-port representation spanning the three device terminals and a common ground node. Other configurations, from one to five ports, can be derived as special cases by treating selected ports as open or short circuits. This allows the device model to be adapted to a circuit rather than requiring every circuit to use the same port arrangement.
The deviceâs nonlinear behavior remains concentrated in a two-variable mapping. The surrounding equations account for the selected port resistances, incoming waves and circuit constraints. Once the two independent device variables are found, the reflected waves can be calculated.
This flexibility matters when choosing between a compact device model and a more complex connection network. A two-port nonlinear element may require an R-type adaptor to represent a particular topology. Such adaptors can have large scattering matrices, increasing computational cost; multiple connections between the adaptor and nonlinear element can also create delay-free loops, which require an implicit solve. A model with more ports may avoid or reduce those complications.
MOSFET and JFET models: explicit solutions
The MOSFET and JFET examples use piecewise device models that distinguish cutoff, ohmic and saturation operating regions. In cutoff, the channel current is zero. In the other regions, current depends on terminal voltages and device parameters.
For each transistor, the model selects two independent variables suited to its operating equations. Substituting the device behavior into the wave-domain circuit equations produces:
- A linear system for cutoff.
- A third-order polynomial in the ohmic region.
- A third-order polynomial in saturation.
Because these cases can be solved in closed form, the scattering relations are explicit: the model does not need an iterative solver to determine the device state. When the operating region is not known in advance, the implementation checks candidate solutions against the conditions defining each region. Past operating states can help prioritize those checks, while closed-form calculations make the worst-case solution procedure predictable.
The JFET model also includes a large gate resistance to represent its small gate current. This resistance can be separated from the nonlinear channel behavior. In the demonstrated preamplifier configuration, that separation allows a port to be made reflection-free and helps avoid a delay-free loop between two nonlinear JFET elements.
BJT modeling and a more robust nonlinear solve
The BJT example uses the Ebers-Moll model, which represents the base-emitter and base-collector junctions with exponential current relationships. In the wave domain, these relationships produce coupled transcendental equations rather than the polynomial equations obtained for the MOSFET and JFET models. They cannot be solved analytically in the same way, so the implementation uses an iterative method.
A conventional Newton-Raphson solver updates both unknowns using the systemâs Jacobian. Its convergence can depend on the initial estimate, and large updates can move the solution away from a useful operating range. The proposed modified Newton-Raphson method applies compensation functions to limit overshoot in each junction-voltage variable. The compensation is designed around the modelâs two separate diode-like exponential contributions.
In the reported tests, the modified method converged for all tested cases, while the conventional method did not always converge. The modified approach also required fewer iterations on average in the comparison. These results illustrate an important implementation principle: solver robustness is part of a nonlinear device modelâs practical value, not merely a numerical detail.
What the circuit examples demonstrate
The models were applied to three audio circuits: a MOSFET overdrive effect, a JFET guitar preamplifier and a BJT common-emitter amplifier. Their wave digital implementations were compared with circuit simulations. The examples show how port selection affects the surrounding network: some configurations use conventional series and parallel adaptors, while others use R-type adaptors. In the BJT amplifier, a three-port device model avoids the eight-port R-type adaptor used in an alternative implementation.
These demonstrations are specifically about audio-circuit emulation, not industrial control validation. Their broader engineering lesson is that model architecture, circuit topology and numerical method should be considered together. A compact port count is not automatically the most efficient choice if it introduces larger junctions or implicit equations.
Practical implications
For teams evaluating nonlinear circuit simulation, the approach offers a structured way to compare implementation options. The MOSFET and JFET cases show how choosing device variables and exploiting operating regions can produce explicit solutions. The BJT case shows when an iterative solver is unavoidable and how controlling its updates can improve robustness.
Ultimately, wave digital modeling makes port configuration a design choice. Selecting an appropriate number of ports can preserve modularity while limiting adaptor complexity and solver requirements. The most suitable model depends on the device equations and the topology it must supportâa useful principle whenever nonlinear components are embedded in computational circuit models.
Access Official Technical PDF