From 41c151021bf09e52f31df45651ba9faa6e6d4523 Mon Sep 17 00:00:00 2001 From: Richard Date: Fri, 14 Aug 2026 18:54:55 -0700 Subject: [PATCH] docs: renumber the tracked-black-level ADR to 0028 Two ADRs landed as 0027 within a day of each other from parallel branches: the SQM docs refresh (#606) and the optical-train docs (#608). The recurring cause is each worktree taking "next integer" off a main that does not yet have the other's ADR. Resolved with the agreed rules: the most-referenced file keeps the contested number, and latecomers move to the lowest globally-free slot. The FOV-gate ADR keeps 0027 -- it is linked from CONTEXT-MAP and referenced by ten bare `docs/adr/0027` mentions in the shipping code and tests of #609, and churning code comments is exactly what the tiebreak is meant to avoid. The tracked black level ADR moves to 0028, the lowest slot free across every branch and remote (0027 is also reserved by #571, which will need 0029 when it merges). Rename plus its three inbound references; the ADR's own text is untouched and carries no "renumbered from" breadcrumb, per the same rules. The bare "See ADR 0027" in the SQM glossary becomes an explicit link, since a bare number is what made this ambiguous to read in the first place. Co-Authored-By: Claude Opus 5 (1M context) --- ....md => 0028-tracked-black-level-supersedes-stored-bias.md} | 0 docs/ax/sqm.md | 4 ++-- docs/ax/sqm/CONTEXT.md | 3 ++- 3 files changed, 4 insertions(+), 3 deletions(-) rename docs/adr/{0027-tracked-black-level-supersedes-stored-bias.md => 0028-tracked-black-level-supersedes-stored-bias.md} (100%) diff --git a/docs/adr/0027-tracked-black-level-supersedes-stored-bias.md b/docs/adr/0028-tracked-black-level-supersedes-stored-bias.md similarity index 100% rename from docs/adr/0027-tracked-black-level-supersedes-stored-bias.md rename to docs/adr/0028-tracked-black-level-supersedes-stored-bias.md diff --git a/docs/ax/sqm.md b/docs/ax/sqm.md index d63e55e19..b27532a4a 100644 --- a/docs/ax/sqm.md +++ b/docs/ax/sqm.md @@ -225,7 +225,7 @@ dark current from sky, because both are linear in exposure. Until a confident fit exists, `pedestal()` returns `None` and the caller falls back to the profile constant. Trust is a lease rather than a latch: an accepted fit expires after `max_age_seconds` and must be re-earned. See -[ADR 0027](../adr/0027-tracked-black-level-supersedes-stored-bias.md) for the +[ADR 0028](../adr/0028-tracked-black-level-supersedes-stored-bias.md) for the decision and its gates. Read noise is zero-mean RMS uncertainty and is never subtracted as signal. @@ -359,5 +359,5 @@ See [`sqm/CONTEXT.md`](./sqm/CONTEXT.md) for canonical terminology, [`ADR-0022`](../adr/0022-sqm-radiometer-first.md) for radiometer-first ownership, [`ADR-0002`](../adr/0002-sqm-published-value-uncorrected.md) for the no-altitude-correction decision, and -[`ADR-0027`](../adr/0027-tracked-black-level-supersedes-stored-bias.md) for why a +[`ADR-0028`](../adr/0028-tracked-black-level-supersedes-stored-bias.md) for why a tracked black level outranks any stored bias. diff --git a/docs/ax/sqm/CONTEXT.md b/docs/ax/sqm/CONTEXT.md index 8db63235d..5cc73b775 100644 --- a/docs/ax/sqm/CONTEXT.md +++ b/docs/ax/sqm/CONTEXT.md @@ -129,7 +129,8 @@ moves with sensor state, so this measurement supersedes both the profile constant and any wizard-measured bias offset once its fit is **leased**. It needs no lens cap and no dark frame. Say "tracked black level", not "auto bias" or "dynamic calibration" — it measures one specific quantity and does -not calibrate anything else. See ADR 0027. +not calibrate anything else. See +[ADR 0028](../../adr/0028-tracked-black-level-supersedes-stored-bias.md). **Lease**: The gate that decides whether a tracked black level is trusted enough to