Possible Bug with Stats and Info

[Linus Torvalds]

On Mon, May 8, 2017 at 3:23 PM, Ryan McLean <pvtryan100 at gmail.com> wrote:
> Attached log.
>
> Would a semi-correct model be to allow a the user to specify the 1st switch
> in bar so to take 35bar as an example on 2 cylinder charged to 200
> 1st switch is at -35bar (165bar remaining)
> 2nd cylinder then at 2x35bar (-70bar) from full (130bar remaining)
> 1st cylinder then at -70bar from last use (95bar remaining)
> ...
> continue until end of dive.

I suspect we should aim to have a model where we can just mark both
cylinders as used at the same time (this is not just a sidemount thing
where people alternate regulators - it's also true of oxygen vs
diluent for example).

But right now that's now how we track gas usage, and it would probably
be pretty painful to implement.

Right now you are realistically forced to either consider it a single
dual cylinder manifold, or you'll have to manually add a fake gas
switch event somewhere in the middle of the dive or so.

> I've probably not understood gas compressibility correctly but I took a 10L
> cylinder to hold 2000L at 200bar. So if I had used 80 bar in each cylinder
> then I had used 1600L total...

Yeah, no.

It's an approximation, but it's really not a particularly good one.

First off, 200 bar is not 200 atm, although it's close. It's about 197.4 atm.

So there's a 1.3% error there.

But more noticeably, the compressibility factor of air at 200 bar is
1.036 - it's not an ideal gas. So that's an additional error of the
magnitude 3.6%, and it's in the same direction as the bar-vs-atm one
(ie both errors are in the direction of less real air at STP).

So 10L at 200 bar is actually only 10*197.4/1.036 L of air at STP. So
about 1906 L.

And 120 bar is 118.4 atm, and at that point air still acts pretty much
like an ideal gas, so a compressibility factor of 1.0.

So 10L at 120 bar is 10*118.4 L at STP, so 1185 L.

So the amount of gas you used (in one cylinder) is 1906-1185=721L.

Note that the small percentage errors all ended up being bigger due to
(a) being in the same direction, and (b) mainly being noticeable at
the higher pressure, so even though we're talking about low
single-digit percentages for the compressibility factor (and even less
for the bar-vs-atm factor), you ended up with about a 10% error
between the rough approximation (800 L used per cylinder) and the more
accurate one (721 L used per cylinder).

               Linus