Meti, Designing Before Building
The Simulator Is Not the Study
A circuit simulator can tell me what one design does at one operating point. The actual engineering question is usually less polite. Which switching frequency still works at the hot corner? How much magnetics loss buys a smaller enclosure? What survives the standard, the tolerance stack, and the part that purchasing can actually obtain?
Meti is the layer I wanted around those questions. It is not a circuit simulator. It drives external tools such as PLECS and LTspice, arranges the experiments, and keeps the results in a form that can be compared rather than admired one waveform at a time.
Studies, Not Loose Scripts
The core is a Rust workspace with study definitions, simulator adapters, parameter series, optimization methods, signal analysis, magnetics models, standards, and report generation. A study can expand a design space with grid, random, or genetic searches, evaluate candidates across operating conditions, and collect the indicators that matter to the system.
Python bindings make the engine usable from the analysis environment where a great deal of engineering already happens. The command-line and desktop surfaces cover repeatable runs and inspection. Reports carry plots and design cards without tying the underlying study to one rendering backend.
Keeping the Argument
The useful output of design exploration is not merely the winning row. It is the argument: which constraints removed the other candidates, where the losses moved, and which assumptions deserve another look. Meti keeps that path visible so a result can be revisited after the prototype disagrees with it, as prototypes enjoy doing.
The project is in active development and has grown well beyond its first collection of power-electronics scripts. It still follows the same practical aim. I want to spend less time moving numbers between tools and more time deciding what those numbers mean before copper is ordered.