Measurement Science Deep Dive — groundSpring¶

The gap between what models predict and what instruments measure. This notebook explores groundSpring's core domain: noise decomposition (bias vs variance), Anderson localization (signal propagation through disordered media), and the tolerance architecture that makes every comparison traceable to a mathematical bound.

Data sources: experiments/results/experiment_catalog.json, benchmark_timing.json, security_gaps.json

For other springs: This is your domain-specific showcase. Replace the science narrative with your most compelling discovery story.

The Five Pillars of Measurement Science¶

1. Signal vs Noise — Sensor drift, calibration error, environmental interference
2. Inverse Problems — From observations back to causes
3. Sensing Systems — How instruments distort reality
4. Temporal Dynamics — How systems drift over time
5. Spatial Propagation — How signals travel through media

Tolerance Architecture¶

Every floating-point comparison in groundSpring uses a named constant with documented provenance. 13 library tiers from DETERMINISM (1e-15) to EQUILIBRIUM (0.1), 5 epsilon guards, 6 upstream contract pins binding to barraCuda v0.3.12 API behavior.

Anderson Localization: The Flagship Domain¶

Anderson localization — how disorder causes wave functions to exponentially localize — threads through 8 experiments across 4 scientific domains: pure math (Exp 008, 009, 012, 018), uncertainty bridging (Exp 015, 022), immunology (Exp 033), and hardware (Exp 028, NPU classification).

Validation Summary¶

Provenance: All data from groundSpring V124 (April 27, 2026). See Spring Catalog on primals.eco.