Odd calculated deco "ripples" (was Re: RFC: color change for calculated deco)

[Linus Torvalds]

On Fri, Jan 11, 2013 at 8:49 PM, Dirk Hohndel <dirk at hohndel.org> wrote:
>
> Maybe this looks better?

It looks a bit boring, but at least it doesn't clash with the
divecomputer ceiling. So I guess boring is good.

That said, I'm noticing something odd in my last trimix profile. The
calculated ceiling looks very different for the same dive with the
Uemis data, and with the data from the suunto's.

Now, ignore the *big* differences. They are due to the fact that my
Vyper Air can only handle two gases (so it doesn't know about the
period where I was breathing 40% O2, and only sees the 20/30 and 80%
O2 cylinders). And the gas switch event for the Uemis is mixed up (due
to the old confusion about cylinder indexes rather than percentage
numbers), so subsurface thinks that the Uemis had me start breathing
80% O2 at 100 ft, and switch to 40% O2 at 30ft. I get an impressive
pO2 of 4.0 on that one ;)

No, ignore the big differences. Instead, go look at my dive in
full-screen mode, and in particular, look at the stuff *before* the
gas switches. In particular, minutes 21 and 22. There's a very odd
"rippling" effect there. It actually continues after the gas switch
(and is more easily seen there), so it makes be a bit worried that we
have some odd

The ripples *seem* to be at 20-second intervals. That initially made
me suspect the sample -> profile plot event interpolation, but that
doesn't seem to match the data either. That interpolation is done at
10-second intervals, and only done if the samples are more than ten
seconds apart, which they aren't (the Uemis samples are mostly 4 or 8
seconds apart, but sometimes 3s - odd).

So I think it's something from the deco algorithm itself. Anything
special about 20 seconds?

Robert - Dirk has access to my dive data from other sources, but I've
distilled down that one dive (and fixed the gas change event and
removed the Vyper Air data) and am including it here as "deco.xml" for
you to see what I'm talking about. Switch between the two computers
with arrow left/right.

(Maximize the subsurface window and do ctrl-2 to see just the graph,
and then you can zoom in too with the mouse. The ripples are not big,
but they are clear when you look close enough).

Any ideas? I don't think the ripples *matter* (they are not a big
effect), but I'd like to understand why they exist. The last time we
had odd effects in the deco graph, it was because the deco
calculations were wrong.

           Linus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: deco.xml
Type: text/xml
Size: 46927 bytes
Desc: not available
URL: <http://lists.hohndel.org/pipermail/subsurface/attachments/20130112/74424640/attachment-0001.xml>