Entropy Budget per Diagnostic

The fabric tracks how many bits of real information each diagnostic contributes, so sampling, storage, and trust are matched to content.

Bits, not bytes

Two channels can produce the same number of bytes yet carry very different amounts of information: a slow, quiet thermocouple and a fast, structured Mirnov coil. The fabric tracks an entropy budget — the real information content per diagnostic — and matches sampling rate, retention fidelity, and compression to it, so effort follows information.

Estimating information content

• Entropy of the sample distribution bounds the lossless storage a channel needs.
• Effective bandwidth times log of signal-to-noise bounds its rate of information.
• Redundancy with other channels discounts a channel's unique contribution.
• Regime dependence: a channel can be information-rich during a disruption and quiet otherwise.

Matching effort to content

A high-entropy channel earns full-rate sampling and lossless retention; a low-entropy channel can be sampled slower and compressed harder without losing physics. Because the budget is regime-dependent, the fabric can retain more during interesting windows (a breeder disruption, a burner plug event) and less during quiet flat-tops, all recorded explicitly.

Feeding governance and quality

The entropy budget informs the data-quality score (a channel delivering far below its expected information is suspect) and the compression policy (bounded-loss only below the noise floor). It is the quantitative backbone of the pipeline's information theory, applied per diagnostic across both machines. It is an analysis discipline for a stack whose FOAK machine begins construction Q2 2027.