CCR deco/TTS computed as open-circuit first-tank gas (understates decompression)

[readme42]

Summary

For CCR (rebreather) dives, decompression and TTS are computed as if the diver breathed the first tank's gas open-circuit, ignoring the constant-ppO2 loop. This understates nitrogen loading, so the app reports a shorter TTS / shallower obligation than reality.

Example: dive 003 (test/dives/003_ccr_with_setpoint_switch_and_calculated_po2.ssrf.xml), GF 45/75, at minute 40 (44.1 m, loop ppO2 1.26 bar):

• App shows TTS 17 min
• Subsurface shows TTS 24 min

Root cause

A rebreather holds ppO2 at a constant setpoint, so the inspired inert-gas partial pressure is ambient − ppO2 (split by the diluent's He:N2 ratio). The app instead uses a fixed open-circuit fraction from the first tank:

• lib/features/dive_log/presentation/providers/profile_analysis_provider.dart (~L648): o2Fraction is taken from dive.tanks.first.gasMix — for dive 003 that is EAN40 → fN2 = 0.60.
• lib/features/dive_log/data/services/profile_analysis_service.dart (~L573-585): gas segments are only built for DiveMode.oc; for CCR/SCR it calls processProfile(fN2, fHe) with that fixed fraction.
• The resolved rebreatherPpO2Curve is passed into analyze(...) but used only for CNS/OTU, never for inert-gas loading. BuhlmannAlgorithm has no CCR mode.

Modeling the loop as EAN40 (60% N2) loads far less nitrogen than the real air-diluent loop at a constant ~1.3 bar setpoint, so deco comes out too short.

Evidence (probe)

Parsing dive 003 and running the deco algorithm directly at minute 40:

Model	fN2	TTS @ min 40
App's current CCR path (EAN40, first tank)	0.60	16 min (matches displayed 17)
Air diluent as fixed OC fraction	0.79	60 min
CCR constant-ppO2 (correct physics)	ambient − ppO2	~24 min (Subsurface)

The app's 16 min reproduces its displayed 17, confirming it models the CCR dive as open-circuit EAN40.

Proposed fix

Make the deco path CCR-aware:

1. Drive inert-gas loading from inspired_inert_pp = ambient − loopPpO2, split by the diluent cylinder's He:N2 ratio (not the first tank).
2. Hold ppO2 at the setpoint through the TTS ascent simulation as well (inert fraction changes with depth as ppO2 stays constant), mirroring Subsurface.
3. ppO2 source: measured O2 cells / dc-supplied ppO2, falling back to the setpoint — consistent with the existing CNS/OTU ppO2 source rule (resolveRebreatherPpO2). The curve is already resolved and passed into analyze(...).

Scope: CCR first, its own PR. SCR is a follow-up (depth-varying loop FO2 rather than a fixed setpoint).

Notes

• Not a regression — a pre-existing modeling gap, independent of the GF-ascent (fix(deco): match Subsurface's gradient-factor ascent (deep GF-low anchor + clear-to-next-stop) #451) and TTS-mandatory-only (fix(deco): keep TTS mandatory-only, report the safety stop separately #454) work.
• Separately: the "GF 45/75" shown for this dive is the recorded Shearwater deco model from the import, not the user's configured GF — worth deciding whether the app's calc should use the configured GF instead.

[ericgriffin]

Status update: PR #484 (dive planner Phase 1 engine) lands the engine half of this fix, but does not close this issue — the logged-dive analysis path still models CCR as open-circuit first-tank gas, so dive 003 at minute 40 still shows TTS ~17 min.

Mapping this issue's three proposed-fix items against #484:

Proposed fix item	Status after #484
1. Inert loading from inspired inert = ambient - loopPpO2, split by diluent He:N2	Engine capability done: BreathingConfig.ClosedCircuit(setpoint, diluent) implements exactly this (constant-ppO2 loading, alveolar-corrected), validated by test/core/deco/buhlmann_ccr_test.dart and the ccr-sp13-dil1845-60m-25min-loading golden vector against an independent Python model. Not wired: profile_analysis_service.dart still gates gas segments on diveMode == DiveMode.oc, so CCR dives fall through to processProfile(fN2, fHe) with o2Fraction from tanks.first.
2. Hold ppO2 at the setpoint through the TTS ascent simulation	Not done — documented Phase 1 limitation: deco schedules/TTS still breathe OC via AscentGasPlan. Loop deco/bailout schedules are scheduled for Phase 4 of the planner redesign.
3. Feed the resolved measured-cell/setpoint ppO2 curve (resolveRebreatherPpO2) into inert-gas loading	Not done — the resolved curve still drives only ppO2 display/CNS/OTU. The Phase 1 plan explicitly defers measured-ppO2-driven CCR loading to Phase 4.

One subtlety for whoever picks this up: wiring items 1+3 alone (loading only) would move dive 003 toward Subsurface's 24 min but overshoot it — at shallow stops a 1.3 bar loop is far O2-richer than OC EAN40 (at 6 m, ambient 1.6 bar means inert 0.3 bar, roughly an 81% O2 equivalent), so stops compute longer until the ascent simulation also holds the setpoint (item 2). The three items are one coherent fix and should ship together, per this issue's original scoping ("CCR first, its own PR").

Prerequisite chain: #484 merges the engine (DecoModel/BreathingConfig API now frozen for later phases) -> the fix here becomes a wiring PR on the analysis provider/service plus the loop-deco ascent, currently slotted in Phase 4 of the planner redesign program (spec: docs/superpowers/specs/2026-07-05-dive-planner-redesign-design.md).