[masarray]

Field incident

Operator reports intermittent Process Bus Insight waveform jump/flicker while capturing ESP32-P4 4I+4V SV. Independent analysis of a sanitized 10.077 s, 80,648-frame PCAP (not committed): canonical 40,324 SV frames and diagnostic mirror 40,323, zero smpCnt jumps and duplicates in both, no PTP frames while Studio reports PTP STOPPED. The canonical stream wraps every 4000 samples; diagnostic mirror is a separate untagged broadcast stream with 16-bit smpCnt. Both streams carry byte-identical eight instantaneous channel values for all 40,323 sample pairs. The PCAP contains no sample-value discontinuity; Npcap host arrival jitter is not proof of publisher send jitter.

Deterministic view bug

In SvStreamState.AddScopeFrame, the analyzer correctly builds a monotonic unwrapped ScopeSampleFrame.Index to survive 4000 / 65536 rollover. However BuildCoherentScopeWindow preferred SampleCounter % (80 spc * 2 cycles = 160) rather than that unwrapped index. Since 65536 % 160 = 96, when the mirror wraps 65535 → 0, its newly received slot 0 belongs to a different phase than the retained prior samples. This visibly splices the waveform despite consecutive wire samples. The bug is masked for the canonical 4000 wrap because 4000 is a multiple of 160.

Fix

• Use the already owned unwrapped sample index for the coherent scope slots. No new caches, no copy or allocation added to 4 kHz receive hot path, and no effect on raw SV decoder, standards fields, SCL mapping, per-stream ownership, power scaling, PTP or publisher firmware.
• Regression uses generated 4I4V 50 Hz data; captures 320 samples just before 65535→0 and one after. For a stable repeated sine, 160-slot waveform must be identical immediately before/after and sequence/missing counters stay 0. Demonstrated discriminating legacy implementation yields up to 7608-count artificial seam, corrected monotonic slotting yields 0 for exact same data.

Other findings (outside scope)

ARStack Studio PTP reported STOPPED / UNLOCKED and smpSynch=0, corroborated by zero 0x88F7 in capture. SMV sample GPTimer is independent of PTP clock discipline; separate milestone arstack61850 #154 covers phase-locked PTP→smpCnt, and #162 next scope. Do not claim physical clock lock from packet continuity. Await green CI before merge.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-16T06:17:58.913841Z	c7b993e	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c7b993e7e3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[masarray]

Merged handoff — independent physical PCAP diagnosis, rollovers and final Windows candidate

• Merged PR fix(sv): remove periodic waveform phase splice at 16-bit mirror smpCnt rollover #28 into main at 8f88482a356fdaabf74ed519c7bd02c95108fb43, tree aa524bf876bec87289d596ea591793f99e8354a2 exactly equals reviewed CI head c08c6da75170e926beb05f1195b73a629d26a611. 7/7 checks PASS incl CodeQL, Windows candidate release, runtime multi-stream, dependency review. 61/61 Windows TRX tests PASS; both review findings resolved.
• Final Windows candidate GitHub Actions run #38860719197, artifact id 12028898615 (ProcessBusInsight-v1.4.0-beta.2-win-x64-candidate-64). ZIP portable SHA256 9e6958eda4e0ea09688a6786c14f1d35201827b2a6854904401cf547ebdc8bff matches SHA256SUMS.txt; manifest commit c08c6da... and ProcessBusInsight.exe present.
• User-held 10.077 s classic PCAP: 80,648 total frames, 40,324 canonical tagged/multicast 0x4000 + 40,323 diagnostic untagged/broadcast 0x4F01, zero smpCnt jumps and duplicates on either; 40,323 paired 8-channel payloads exactly identical; all smpSynch=0 and quality word zero; 0 L2 PTP 0x88F7/UDP 319/320. Host Npcap arrival can show max ~15 ms batching even with contiguous smpCnt, thus no physical TX jitter proof. Never commit operator PCAP/identifiers.
• Root cause fixed: BuildCoherentScopeWindow wrongly selected slots via raw 16-bit smpCnt % 160 instead of already-unwrapped sample index; 65536 % 160 = 96, creating a phase splice during the diagnostic mirror's wrap every ~16.384s. Synthetic discriminating example old algorithm spliced by ~7608 raw counts, new algorithm remains constant across wrap.
• Review found second issue: ResolveSequenceWrapBase guessed maxObservedCounter+1 as rate modulus and could conceal packet loss right before counter wrap. Fixed to accept only observed supported boundaries without s.m.pRate (3999,4799,65535), else explicitly reject/hold last complete window. Adversarial deterministic regression 3998→0 with missing3999, and normal 65535→0. No second ring/parser/cache; no per-packet overhead.
• Separate PTP milestone under ARStack #154/#162: operator diagnostics PTP STOPPED/UNLOCKED, hence no 0x88F7 expected, and released firmware free-runs GPTimer for SV, not continuously EMAC-PTP disciplined. PTP two-step SOURCE code not independently verified on wire, no claims of IEC 61850-9-3 or smpCnt↔PTP phase lock. PR #166 M3 Draft remains isolated.
• Field acceptance remaining: Windows candidate 30+s capture of diagnostic mirror crossing its 16-bit wrap to confirm no flicker; separately START PTP SOURCE in Studio, check Announce/Sync/FollowUp & PTP 0x88F7 on receiving NIC, no external IED/switch required for basic visibility; full timing accuracy requires independent GM/PPS and hardware timestamp evidence.