F650 Open SCL: expose report-backed MMXU1 A/PhV/PPV phase measurements #493

Field evidence: GE F650 structured measurements missing in Live Values (2026-10-09)

Opened BCU_GE.iid with ARSAS 1.6.40+1ac41648549db92902190ae91ee6302da2d0afe0 and pinned engine 352c81e6a798635c6addcee0683235ca87ad416d.
Actual IID file has not been supplied: do not claim exact vendor phase-reference paths, and do not publish the private IID or machine/IP details.

• 1 LD, 153 LN, 2228 DO, 14668 DA; 8 Static DataSets with 124 members (124/124 inventory-present), 20 unresolved primary-value descriptors.
• 98 selected live signal rows and 16 independently selected control objects, 53 unique runtime monitor points; 53/53 displayed only means all admitted points have values, not complete measurements.
• Source DataSet semantic members include MMXU1.A / PhV / PPV (in CSWI, MMXU_SOGI1/2/3 and REPORT1) and AuxV (SOGI1/2/3). They are missing from the listed selected leaves. Scalar TotW/TotVA/TotVAr/TotPF/Hz and BusHz.mag.f are admitted.
• 6/6 actual configured RCB streams routed; 1140/1140 SCL-assisted MMS initial reads successful, 11532 projected leaves, 0 read errors, 0 cyclic MMS process polling. Root concern is schema-to-runtime projection/admission, not missing network access.
• Independent warnings: semantic gate INCOMPLETE and two BRCB BufOvfl findings. Audit separately; do not declare SOE continuity proven.

Root-cause source review

Engine Iec61850DataSetSemanticBindingResolver.PrimaryValue returns a value only when exactly one primary attribute exists; a top-level three-phase FCD can have several valid phase primaries. Iec61850DataSetSignalInventoryProjection therefore preserves the FCD as unresolved rather than inventing a scalar. This is intentionally conservative.
Consumer Iec61850DataSetSignalInventoryService.ResolveRuntimeBinding only resolves an unresolved structured member if SchemaSafeAggregateProjectionService.TryResolveStaticDataSetPrimaryLeaf succeeds. That method rejects top-level DataObject references, only supports descendant FCDs (e.g. A.phsA). SignalDefinition.IsThreePhaseMeasurementAggregate handles THD types but not ordinary A/PhV/PPV. StaticDataSetAuthoritySelection chooses CanPublishToRuntime candidates and only one row per member; the result is current/voltage phase content not published. StaticDataSetReportProjectionAccumulator only handles supported THD/demand aggregate parents.

2026-10-09 exact private IID template review (user-provided, kept out of GitHub)

• The original BCU_GE.iid was supplied directly for private analysis. It was generated by IEDScout 4.10; IEDName BCUGE; MMXU LNType BCUGEF650.MMXU1. The file itself is not committed or attached to this public issue.
• Explicit MMXU1.A DOType includes typed FLOAT32 cVal.mag.f and instCVal.mag.f for phsA, phsB, phsC, neut, net, res. Operator descriptions: Phasor Ia/Ib/Ic/Ig/In/Isg Primary; fc=MX phase q/t; units SIUnit=A with multiplier=k.
• Explicit MMXU1.PhV DOType includes typed FLOAT32 cVal.mag.f and instCVal.mag.f for phsA, phsB, phsC, neut. Operator descriptions Va/Vb/Vc/Vn Primary; fc=MX phase q/t; SIUnit=V, multiplier=k.
• Explicit MMXU1.PPV DOType includes typed FLOAT32 cVal.mag.f and instCVal.mag.f for phsAB, phsBC, phsCA. Operator descriptions Vab/Vbc/Vca Primary; fc=MX phase q/t; SIUnit=V, multiplier=k.
• Therefore 13 exact schema-defined phase magnitudes exist for the three missing FCDs; plus AuxV is a separate single complex measurement (cVal.mag.f and instCVal.mag.f), not a 3-phase object.
• Static FCDA members have doName=A/PhV/PPV and daName omitted, fc=MX, proving whole-DO FCD membership. Those FCDs appear in CSWI, MMXU_SOGI1, MMXU_SOGI2, MMXU_SOGI3, REPORT1; AuxV appears in MMXU_SOGI1/2/3. No additional RCBs or dynamic DataSets are justified.
• The precise reported numerical values, actual report leaf payload and physical scaling are not present in the IID. They must be proven with live report evidence. Multipliers in IID must not blindly be multiplied on displayed values without checking engine's value/unit policy.
• Required fix must retain all six A channels (including neut/net/res) and all four PhV channels rather than truncating to ABC, and must preserve q/t/units and one-parent-to-many children. Use typed schema to safely prefer exact canonical cVal.mag.f and refuse ambiguous/nonexistent leaves.

Acceptance / safe solution

1. Obtain actual BCU_GE.iid (private, not committed) and inspect exact FCD CDC/type-template/FC=MX, phase paths and live decoded report child values. Identify each of 20 unresolved descriptors by reason; avoid guessing paths or units.
2. Add source-neutral typed one-FCD-to-many-phase-leaf projection using engine-authoritative ResolvedAttributes and exact schema. WYE phase A/B/C and DEL AB/BC/CA should become distinct readable rows only when proven. Never choose the first numeric child or fabricate missing phase values.
3. Preserve literal DataSet member and RCB provenance while selecting all proven exact phase leaves. Route the engine-expanded InformationReport children to those rows, with separate quality/time/engineering units; no extra RCB subscriptions per phase.
4. Preserve Static DataSet report-only: no cyclic MMS process polling, no dynamic DataSet writes, no change to canonical command path/CSWI controls or unrelated report continuity.
5. Deterministic regression tests with WYE, DEL, missing/ambiguous attributes, shuffled child ordering, optional vendor instantaneous paths, FC/type mismatch, quality/timestamp, aliases, and actual GI report updates; compare SCL vs discovery parity.
6. Show explicit diagnostics: scalar-ready, structured phase-ready, unresolved with exact cause; keep semantic gate INCOMPLETE and BufOvfl investigations distinct.
   Physical acceptance is a real GE F650 IID + captured reporting proof for current and phase/line voltage, without regression to existing control/reporting.

Related: draft PR #492 (F650 canonical control + native-dialog UX), issue #446. This is an ANALYSIS ticket only: no code modification/CI requested yet.