Name: Circuit Solver
Author: lgili

Circuit Solver

Modified Nodal Analysis (MNA) specialist for power electronics circuit simulators written in C++. Use when building or debugging the netlist-to-matrix pipeline, writing component stamps, handling switch state changes, assembling sparse matrices, or investigating convergence failures caused by ill-conditioned matrices. Trigger for terms like "MNA", "nodal analysis", "stamp", "admittance matrix", "Y matrix", "KCL", "KVL", "branch current", "netlist", "matrix assembly", "sparse matrix", "node numbering", "supernode", or "ground node".

lgili0 Sterne09.04.2026

Beruf
Kategorien: Machine Learning

Overview

Use this skill when working on the core simulation engine: the part that takes a circuit netlist, assembles a system of equations, and solves for node voltages and branch currents.

The Modified Nodal Analysis (MNA) method is the standard for SPICE-like simulators. It produces a linear (or linearized) system Y·x = J where:

Y = admittance matrix (conductances + frequency-domain equivalents)
x = unknown vector [node voltages | branch currents]
J = right-hand-side (current sources + history terms)

Default stance:

Ground node (node 0) is always the reference — never assigned a row/column in the matrix.
Every component contributes to Y and J via a stamp — keep stamps isolated and composable.
Switches change the circuit topology — re-stamp affected rows/columns when state changes.
Nonlinear elements are linearized per Newton-Raphson iteration — stamp the Jacobian, not the element's secant.
Sparse storage (not dense) is required for any circuit with more than ~50 nodes.

Core Workflow

Overview

Use this skill when working on the core simulation engine: the part that takes a circuit netlist, assembles a system of equations, and solves for node voltages and branch currents.

The Modified Nodal Analysis (MNA) method is the standard for SPICE-like simulators. It produces a linear (or linearized) system Y·x = J where:

Y = admittance matrix (conductances + frequency-domain equivalents)
x = unknown vector [node voltages | branch currents]
J = right-hand-side (current sources + history terms)

Default stance:

Ground node (node 0) is always the reference — never assigned a row/column in the matrix.
Every component contributes to Y and J via a stamp — keep stamps isolated and composable.
Switches change the circuit topology — re-stamp affected rows/columns when state changes.
Nonlinear elements are linearized per Newton-Raphson iteration — stamp the Jacobian, not the element's secant.
Sparse storage (not dense) is required for any circuit with more than ~50 nodes.

Topic	Reference	Load when
MNA equation structure	`references/mna-formulation.md`	Understanding the Y·x=J system, branch current augmentation, or ground reference
Component stamps	`references/component-stamps.md`	Writing or debugging stamps for any standard component
Sparse matrix assembly	`references/matrix-assembly.md`	Building the C++ data structures for efficient matrix assembly and factorization

Circuit Solver

Overview

Core Workflow

Circuit Solver

Overview

Core Workflow

Reference Guide

Constraints

MUST DO

MUST NOT DO

Output Template

Primary References

Continuous Learning V2

Continuous Learning V2

Continuous Learning V2

Continuous Learning

Continuous Learning

Pytorch Patterns