Zero Overshoot Demo — Bounded vs Unbounded Recursive Filtering Tool

How to read the demo

quick start

Start with Step + EMA (α=0.50) to see overshoot.
Switch to Zero Overshoot (GSRF) and compare how the output stays within the target band.
Use Max overshoot and Band crossings as a quick boundedness read.

Synthetic, domain-shaped signals (not real-world data). Filter math is unchanged. Max overshoot is the largest amount the output leaves the target band. Band crossings counts how often it repeatedly exits/returns to the band. Time to stability is the first step where the output stays within the band for the stability window. Read these together to compare bounded vs unbounded behavior.

Learn more: deterministic safety, runaway feedback, paper abstracts.

Controls

seed: 2026

Industry preset Signal model Signal type

GSRF

EMA (α=0.10)

EMA (α=0.50)

Show all

Noise (σ)

σ value

Target band (±)

Target center

GSRF parameters

β (0–1]

k_return (>0)

mem [0–0.5]

b (|b|<2)

—

Max overshoot (selected)

—

Band crossings (selected)

—

Time to stability (steps)

Tip: start with Step + EMA (α=0.50) to see overshoot, then switch to GSRF. "Show all" overlays all filters on the same signal.

Signal + Filter Output

mode: gsrf

Industry context

Typically optimizes

—

Common runaway failure pattern

—

What Zero Overshoot constrains

—

Examples: —

Signal model: — (synthetic)