Calibration & Drift Tracking

Every channel carries a live calibration; the fabric watches for slow drift and corrects or flags it before it biases a feature.

Calibration is data

A raw ADC count means nothing without the calibration that maps it to physical units: gain, offset, feedthrough parasitics, filter delay, and sensor sensitivity. The fabric stores this per channel and applies it as the first step of normalization. Because the calibration is versioned, any archived shot can be re-reduced with the exact calibration in force when it was taken.

Sources of drift

• Thermal drift in front-end gain and offset as the cryostat and hall temperatures move.
• Neutron-induced changes in sensitivity, from 14 MeV neutrons in the breeder and the 5.44%-fraction flux in the burner.
• Mechanical relaxation of bonded sensors such as REBCO strain gauges over load cycles.
• Baseline wander in integrated diagnostics like flux loops.

Detecting drift

Drift is detected against references and against redundancy: a channel that steadily disagrees more with its neighbors, or whose baseline walks between shots, is flagged. Where an in-situ reference exists, the fabric corrects continuously; where it does not, it flags the channel for the next calibration and lowers its quality score in the meantime.

Correcting without erasing history

A correction never overwrites raw data. The archive keeps raw counts; corrections live as calibration versions applied at read time. This preserves forensic value — a disruption in the breeder or a plug collapse in the burner can be re-examined with either the original or a later, better calibration — and keeps lineage exact.

Calibration and drift models are design specifications for machines under construction from Q2 2027; the drift sources listed are the physically expected ones for the FOAK devices.