Skip to content

MAST locked-mode authority

validation/mast_locked_mode_authority.py is the locked-mode admission gate for canonical locked_mode_amp. It distinguishes a source-authoritative MAST locked-mode observable from the historical SCPN-CONTROL compatibility recipe. The gate measures source facts and the candidate filter's physical time scale, but it never filters the saddle array or emits a locked-mode amplitude.

Primary-source boundary

Kirk et al. describe MAST locked-mode formation as a growing n=1 radial magnetic-field perturbation measured by an outer-midplane saddle-coil array. The modes in the reported experiment form locked without a rotating m=2,n=1 signature. The paper does not define SCPN-CONTROL's 201-sample complex boxcar as the estimator, filter, calibration, or uncertainty contract.

The gate therefore requires the existing saddle row/geometry authority and locked-mode-specific declarations for:

  1. a radial measured component on the outer mid-plane;
  2. a stationary n=1 estimator in the machine frame;
  3. a physical-frequency filter policy, edge handling, and a positive cutoff below the source Nyquist frequency;
  4. background and poloidal-field pickup correction policies;
  5. a vacuum-vessel response policy;
  6. positive one-standard-deviation uncertainty;
  7. a content digest for the estimator authority; and
  8. the exact primary-source citation.

Free-text naming, a smooth-looking trace, or a chosen sample count cannot supply these declarations.

Legacy compatibility candidate

The replay producer currently calculates:

abs(boxcar_window(A_1_complex(t), 201 samples))

The rationale is that a rotating complex n=1 phasor may average towards zero while a stationary component survives. This is a useful synthetic recipe, but the window is expressed in samples and was not tied to a source-authoritative physical cutoff, background/pickup subtraction, vessel response, or uncertainty. The authority gate records this formula as compatibility_only_not_source_authorised.

Real two-shot result

Generation-pinned shots 30421 and 30424 both have a 50 kHz saddle timebase. A 201-sample window spans 4 ms between its first and last samples and has 4.02 ms of discrete support. Its exact first boxcar null is 50000/201, approximately 248.756 Hz. These measured numbers are identical for both shots, but they do not prove that this frequency is a valid locked-mode boundary.

All fourteen locked-mode-specific declarations listed by the gate are absent. The inherited saddle row/geometry blockers also remain: rows 9 through 12 contain no finite samples and the ordered row join, selected vertical set, released geometry, calibration, baseline, and bad-channel authority are incomplete. The result is therefore blocked; locked_mode_amp remains a compatibility candidate and all training, scientific-validation, facility-prediction, and control-admission claims remain false.

The byte-reproduced report is retained as non-public operational evidence; its file SHA-256 is 1ddc292e...9cac2e and its payload digest is 54c9fccb...e31f70.

Reproduction

Run only against a verified SourceObjectManifest-v2 root:

python -m validation.mast_locked_mode_authority \
  --manifest /external/source-v4/source_object_manifest.json \
  --artifact-root /external/source-v4 \
  --shot-id 30421 --shot-id 30424 \
  --json-out /external/l2f12d-locked-mode-authority.json

The output is deterministic and self-digested. It contains metadata, blockers, timebase statistics, and digests only; no raw saddle arrays or derived locked-mode traces are emitted.