| name | qspice-waveform-analysis |
| description | Inspect, derive, and measure QSpice waveform results after a simulation. Use when reading signals from a `.qraw`, comparing nodes, computing derived expressions like V(out)-V(in), measuring RMS/peak/gain/THD, or exporting traces for further analysis. |
| license | MIT |
| metadata | {"author":"qspice-mcp","version":"1.0"} |
QSpice: Waveform Analysis
A workflow for turning a completed .qraw result into answers: which signals
exist, what a derived quantity looks like, and what the scalar metrics are.
When to use
- A simulation finished and you need to read or quantify the result.
- You need a derived signal (difference, product, ratio) the raw file does not
store directly, e.g.
V(out)-V(in) or instantaneous power V(n1)*I(L1).
- You need scalar metrics (RMS, mean, peak-to-peak, gain) or a Fourier/noise
summary.
When NOT to use
- The run failed / produced no waveform → use
qspice-convergence-debugging.
- You still need to author or simulate the circuit →
qspice-getting-started.
Workflow
- Enumerate.
list_signals (and list_steps for stepped/.step runs) to
learn the exact, case-correct signal names. QSpice lowercases net names, so a
node drawn as VOUT is read as V(vout).
- Read a single trace.
read_waveform with a signal and optional
t_start/t_end. Responses are budget-capped (2000 points); narrow the
window instead of asking for everything.
- Derive across signals.
evaluate_waveform_expression with an arithmetic
expression over signal tokens — V(out)-V(in), V(out)*I(L1),
2*V(fb). Only + - * / **, parentheses, and numeric constants are allowed;
all referenced signals must share one axis (same raw, step, window).
- Quantify.
measure_waveform for RMS/mean/max/min/peak-to-peak on one
signal; read_fourier for .four THD summaries; read_noise for .noise
integrated/spot output.
- Visualize / export.
plot_waveforms for a PNG preview; export_waveform_csv
when downstream tooling needs the full series outside the MCP budget.
Conventions
- Always
list_signals first; never assume casing.
- Prefer
evaluate_waveform_expression over manually reading two traces and
subtracting by hand — it aligns axes and applies the budget.
- For complex AC data, pick an explicit
component (magnitude/phase) rather
than relying on auto.
- Confirm a result is physically sane with
measure_waveform, not just that the
read succeeded.
