Filterpole: Filter Design

The nominal curve passes; the production batch does not. An anti-alias RC that sits exactly on its corner on paper drifts across the requirement line once 5% capacitors arrive, and a biquad that looked clean in floating point grows passband ripple - or an unstable pole - after its coefficients are rounded to fixed point. Filterpole works in that gap between the textbook response and the response your components will actually build. Work moves through four instrument surfaces: 1. SPEC - choose analog RC (low/high-pass) or digital biquad (low, high, band-pass, notch); enter values, tolerances, distributions, and dB requirement bands on a log-frequency grid. 2. REALIZE - component values snap to the E12/E24/E96 preferred series; digital designs show quantized coefficients and a pole-zero map that redraws as you step precision from 4 to 30 fractional bits. 3. ENVELOPE - a seeded Monte Carlo study sweeps your stated component spreads and plots the P05-P95 production envelope against every band, with estimated yield and a Wilson 95% interval. 4. RUNS and PACKET - inspect the worst observed sample and each requirement crossing, then export the study as JSON, CSV, or a PDF evidence packet. Every study is reproducible from its stated seed and carries a SHA-256 checksum, and editing any input visibly invalidates the previous result, so a stale plot can never pose as current evidence. Figures are sampled-model estimates for study, not production sign-off. Projects live solely in the app's local container on your device; Filterpole makes no network connections and requests no permissions.