How we validate the numbers
A decompression planner is safety-relevant software, so "trust us" isn't good enough. This page states exactly where DecoLog's algorithms come from, what they're checked against, and what the automated tests guarantee on every change.
Lineage: published models, reference implementations
Nothing in the deco engine is invented here. The ZHL-16C engine uses Albert Bühlmann's published coefficients (from Tauchmedizin) — 16 compartments, nitrogen and helium in parallel — with gradient factors applied per Erik C. Baker's formulation. The VPM-B engine is a faithful scalar port of Subsurface's implementation (core/deco.cpp and core/planner.cpp), the open-source codebase widely used as a reference in the diving community. The port reproduces the critical-volume iteration (CVA), the crushing-pressure model in both its gas-permeable and impermeable regimes, and Boyle's-law compensation, using Subsurface's own pre-scaled constants rather than re-derived ones. That code is GPL-2.0; DecoLog's port carries the same notice, crediting the bubble model to Yount & Hoffman and VPM-B's Boyle compensation to Erik C. Baker.
Reference-profile calibration
The VPM-B engine is calibrated against the published deep-trimix reference profile used by Subsurface: 120 m for 20 minutes on trimix 18/50 at conservatism +1. DecoLog plans it at 139 minutes total runtime versus the reference's 141 — within about 1.5% — with a near stop-for-stop schedule match, including the first stop at 63 m. The ZHL-16C engine's no-stop limits are likewise checked against published values. These calibration dives are pinned as automated test fixtures, so a change that drifts the engine fails the build.
Safety regression tests
Beyond calibration, a dedicated regression suite asserts properties that must hold for every plan, not just the reference dives:
- ppO₂ limits on every segment. Every engine-chosen ascent, stop and gas-switch segment must respect the 1.6 bar decompression ppO₂ ceiling at the segment's deepest point — so a deco-gas switch can never be scheduled below that gas's MOD.
- Depth continuity. Consecutive runtime segments must join exactly — the plan can never "teleport" to a stop deeper than the diver's current depth.
- Altitude behaves correctly. Lowering surface pressure must produce more decompression, never less.
The wider test suite covers the gas calculators too — MOD, EAD/END, SAC/RMV, CNS/OTU accumulation, and the partial-pressure and Van der Waals real-gas blending solvers.
Why we show you two models
Running ZHL-16C and VPM-B side by side is itself a validation habit: if two independently derived models agree on a plan's rough shape, that's information; where they disagree, that's a decision you should be making consciously — not one a single default algorithm should be making for you.
The honest limits
Validation proves the software computes the model correctly. It cannot prove the model matches your body on that day — decompression sickness is probabilistic, and every algorithm is an approximation fitted to limited trial data. So the disclaimer in the app is not boilerplate, it is the actual epistemic state: DecoLog is a planning aid, not a substitute for a dive computer or proper training. Verify every plan independently. Dive a computer. Stay within your certification.
⚠ Sources: A. A. Bühlmann, Tauchmedizin; D. E. Yount & D. C. Hoffman, Varying Permeability Model; Erik C. Baker, VPM-B and gradient-factor papers; Subsurface (subsurface-divelog.org), GPL-2.0 reference implementation.