[PATCH 2/2] gas model: replace Redlich-Kwong with least-square quintic

[Linus Torvalds]

On Wed, Mar 2, 2016 at 4:24 PM, Linus Torvalds
<torvalds at linux-foundation.org> wrote:
>
> I'll post a test-patch soon, working on it right now.

Ok, suggested patch posted. It does the linear mixing, *except* for
the case of actual air. I'd rather use the real air table than trust
my linear mixing model.

It has some funny - but believable - effects.

I see 80(77)cuft for air in an AL80 as expected.

And now with that gasmix approximation thing, I see 80(78)cuft when I
have a 32% nitrox mix.

And the reason I say that is believable, is that it actually sounds
very reasonable. Oxygen seems to compress *better* than an ideal gas
(and much better than Nitrogen), and a pure oxygen mix would be 80(84)
because the compression factor at 3000psi is actually 5% lower volume
than an ideal gas according to baue. So a higher oxygen content would
be expected to compress better.

NOTE! I did not use your new air coefficients, I used our old ones
from Wikipedia rather than replacing them with your new ones that were
based on the table at http://www.baue.org/library/zfactor_table.php

It would be interesting to see a plot of

 (a) the wikipedia coefficients that we use

 (b) the coefficients for air you calculated from the air table on baue.org

 (c) the linear 21/79% mix of the O2/N2 (that we don't actually use,
since I decided to keep air special, partly because I trust that table
more, but partly because we just have a special fractional
representation of air anyway)

and see how well they match. If they *don't* match well, that would be
cause for worry.

               Linus