TechnicalSCUBA decompression

The evidence base is thinner here than in recreational diving, so the model carries more of the work. Each class below moves a different part of the decompression problem. A result drawn from one class does not carry over to another, so the record keeps them apart.

• Bühlmann ZH-L16 with gradient factors

  Two gradient factors set how close the ascent comes to the M-value line, one at the first stop and one at the surface. Moving them redistributes time across the ascent. The 16 compartment half-times stay fixed, so a schedule quoted without both factors is underdetermined.

• VPM-B

  A bubble model with a critical volume criterion and a published set of constants. An implementation that alters those constants is a different model and needs to be recorded as one.

• RGBM

  Implementations carrying this name differ widely and several are proprietary. Treat the name as a label on a family, and record the specific implementation that produced any number.

• Table lineage

  Several technical schedules descend from naval and commercial tables fitted to a specific exposure range. That fit range matters when a schedule is used outside it.

1. Do deep stops redistribute inert gas between fast and slow compartments in a way that changes outcome, and for which exposures does the redistribution matter?
2. How should helium kinetics be represented when the same profile is modelled once by a dissolved-gas set and once by a bubble model?
3. What does an isobaric counterdiffusion penalty look like inside a model that carries no separate term for it?
4. Does setpoint handling or the decompression model account for more of the spread between rebreather outputs?
5. Which technical exposures appear in public profile data at all, and which are missing from every available dataset?

A technical schedule is reproducible only when a second party can regenerate it from the record. The 6 fields below are what makes that possible.

1. Model family, version and every configuration constant, gradient factors included.
2. Gas fractions for each mix and the depth of every switch.
3. Implementation identity: source repository, commit and build.
4. Full input profile at its recorded sample rate, not a summary of it.
5. A run manifest that links the published number to the exact inputs behind it.
6. Checksums on the inputs, so a rerun can be compared against the original.

• Trial reports and tables

  Published decompression trials and the tables derived from them, with their exposure range recorded.

• Profile datasets

  Dive-computer exports and logged profiles, each with its provenance and its access terms.

• Model implementations

  Reconstructed implementations, each with a verification level and a calibration lineage.

• Benchmark protocol

  How a model and a dataset become eligible, and how metrics are frozen before a run.