The medium

Every other area of this library measures something that happened in a medium. This one is the medium. A density and a speed of sound stand behind every sound power level, every absorption coefficient, every transmission loss and every propagation calculation in the tree, and for most of the library’s life they arrived the same way: as a number somebody typed once.

That is what this area exists to stop. phonometry.fluids computes the state of the fluid from the conditions that were actually measured, keeps the conditions beside the result, and says which model produced it.

Three media are built here. Humid air has the most carefully stated model in the acoustic literature and gets a guide of its own below. Sea water has four rival sound-speed fits and carries the one it was asked for. Any other gas is fluids.ideal_gas, which takes the ratio of specific heats and the molar mass that a gas table prints and returns the speed and density that follow, with how far that closure goes written into the result.

A density and a speed of sound stand behind every level the library computes, and they do not all come from the same place. Four states sit in the tree, one beside each model or standard that fixes it, and they disagree:

None of them is wrong. Each is the air its own document assumes, and substituting one for another would change the number that document prints: the 1,7 per cent between the metrology annex and the building standard is a real disagreement between two committees, not a rounding.

Each one stays where it is, beside the clause that prints it. Gathering them into this package would invert the dependency it exists at the bottom of: any domain may import fluids without an architecture edge, and a catalogue here that reached into materials, building and simulation would make the medium depend on three of the domains that stand on it. The comparison is a documentation artefact instead, and the published catalogues page lists all four side by side.

What fluids.PUBLISHED_FLUIDS does hold is the states this library read from a printed page: the three fluids Bies prints at the head of his Table C.1, air, fresh water and sea water, each naming that page, with the printed folio, in its model, because for a state that was read rather than computed the table is what produced it.

It is not a table to look values up in. Air at conditions that were measured is fluids.air, sea water is fluids.sea_water, and both compute the state rather than recall it.

Sixteen of the twenty-one packages are domains of application: you go to building because you are measuring a building, to underwater because you are working in water. You never go to fluids because you are measuring a fluid. You go there because whatever else you are measuring happens in one.

So it sits with filters, signals and metrology in the transverse toolbox: any package may import it without an architecture edge, because a medium is not a subject some domains have and others do not. It was the fourth member of that set, the first added since the library split the original toolbox in three, and solids is the fifth.

The reason a library ends up with a dozen different values for the density of air is that three different things wear the same clothes.

The physics of the fluid is what lives here. It answers “what is this air, at these conditions”, and better physics is an improvement: when the model gets more accurate, every caller who asked for air should get the more accurate answer.

A standard’s own simplified formula is not that. When ISO 10534-2 prints $c_0=343,2\sqrt{T/293}$, that expression is part of the procedure, and a measurement that claims to follow the standard has to use it. Those stay in the module that implements their clause, with the citation beside them, and they never move here.

A constant frozen by a conformance row is a third thing again. The Johnson-Champoux-Allard model carries a Prandtl number of 0,71 as a published constant of the model. The air at the reference state has 0,728. Substituting the physical value into the model would not correct an error, it would change the model, and it moves the impedance it computes by 1,5 parts in a thousand. That constant stays frozen where it was published.

Keeping the three apart is what lets better physics reach a caller without a single measurement silently ceasing to reproduce the standard it cites.

• No solids. A Fluid carries no shear speed, and the elastic materials of the wave solvers keep their own type with its own precondition. The two are different quantities that happen to share the word “medium”, and a solid’s properties are tabulated where a fluid’s are computed: they live next door, in phonometry.solids.

• No fields, only states. A Fluid is one fluid at one point. The stratified profiles that ray tracers march through stay in the packages that own their marchers, in the ocean and in the atmosphere, because a profile is a description of a place rather than of a substance.

• No frequency dependence. The speed of sound here is the zero-frequency one. Molecular relaxation makes sound speed depend on frequency, and the model that describes it lives with the atmospheric absorption that needs it, in phonometry.environment.

• Humid air: the CIPM-2007 formulation of IEC 61094-2:2009 Annex F, what it fixes and what it does not, how much each condition is worth, and why the library asks for the temperature but assumes the pressure out loud.