HierarchicalScatteringCircuitSolver

class pmrf.models.composite.interconnected.circuit.solvers.scattering.HierarchicalScatteringCircuitSolver(eps: float = 1e-12, linear_solver: AbstractLinearSolver = AutoLinearSolver(well_posed=None))

Bases: AbstractScatteringCircuitSolver

Hierarchical S-parameter reduction solver.

Operates sequentially by applying a generalized block Schur complement (sub-network growth) to each internal net block-by-block. Unrolls the circuit topology into the algorithm, which may result in increased compile times.

Best suited for circuits that contain a mix of chain-like topology and arbitrary interconnections.

run(s_block_diagonal: Array, z0_ports: Array | None, z0_ext: Array | None, topology: PortRepresentation) → ScatteringResult

Executes the scattering circuit solving algorithm.

Note on Dangling Ports

Implementations MUST safely handle mathematically isolated external ports (where a port in topology.ext_idx is the ONLY port on its defined net). To prevent solvers from evaluating these as perfect open circuits (S=1.0), implementations should dynamically inject matched virtual VNA probes (e.g., padding the S-matrix with 0.0s and rewiring the indices) prior to solving.

param s_block_diagonal:

A 2D array of shape (N_total, N_total) containing the uncoupled S-parameters of all components placed sequentially along the main diagonal. All off-diagonal elements are exactly zero.

type s_block_diagonal:

jnp.ndarray

param z0_ports:

A 1D array of shape (N_total,) containing the characteristic impedances for all ports.

type z0_ports:

jnp.ndarray

param topology:

The static map dictating the port connections, nets, and partition logic.

type topology:

PortRepresentation

returns:

The fully solved S-parameter result.

rtype:

ScatteringResult

eps: float = 1e-12

linear_solver: AbstractLinearSolver = AutoLinearSolver(well_posed=None)