qspice-power-electronics

QSpice: Power Electronics

A workflow for switching converters (buck, boost, SMPS): get a working transient, then characterize efficiency, transient response, and control-loop stability.

When to use

• Designing or analyzing a DC-DC converter or other switching power stage.
• You need efficiency, output ripple, load-step response, or loop-gain / phase-margin numbers — not just "does it run."

When NOT to use

• The run won't converge or stops early → qspice-convergence-debugging first.
• Generic single-node waveform reads/metrics → qspice-waveform-analysis.

Workflow

1. Get a circuit. materialize_reference_circuit for a bundled starting point (e.g. buck_converter_cpp), or author your own. inspect_schematic to confirm topology; check_schematic for a conservative ERC pass (floating nets, single-pin nets, missing ground) before spending sim time.
2. Transient first. prepare_transient to stage a .tran directive long enough to reach steady state plus the events you care about, then run_simulation.
3. Characterize the output. read_waveform on the switching node and output; measure_step_response for rise time / overshoot / settling on a load or reference step; measure_efficiency from input/output power traces.
4. Assess the loop. For a switched (non-averaged) SMPS prefer .bode: prepare_bode_analysis on the settled circuit, run_simulation, then measure_stability_margins. For small-signal/averaged models use prepare_loop_gain_analysis (Tian or Middlebrook injection) + prepare_ac.
5. Verify suspicious Bode points. .bode uses aggressive numerical methods and can carry artifacts (spectral leakage, aperture diffraction). Spot-check the crossover with prepare_meas(kind="fra", ...) — it stages .meas <name> fra <freq> <input> <output>, the most reliable frequency-domain measurement QSpice offers — then re-run and read_measures.

Closed-loop .bode details

• Injection point: the perturbing source goes between the SMPS output and the top of the feedback divider (low impedance driving high impedance).
• Settling first: run prepare_transient + run_simulation to learn how long the supply takes to settle; pass that as settling_time to prepare_bode_analysis. No settling acceleration is attempted by QSpice.
• Reference node: if the feedback reference is not at AC ground, pass reference_node (stages .options boderef=<node> alongside .bode).
• Amplitude shaping: perturbation amplitude rises away from the geometric mean of the sweep by default. Tune with bode_amplitude_frequency, bode_low_power, and bode_high_power (.options bodeampfreq/bodelopow/bodehipow), or set bode_amplitude_frequency="0" for constant amplitude.
• Low FSTART is expensive: each decade down multiplies simulation time.