Discrepancies between DC and subsurface's deco

[Robert Helling]

On 06.11.2013, at 08:11, Patrick Valsecchi <patrick at thus.ch> wrote:

Hi Patrick (and others),

> I was trying to figure out why my DC (Shearwater Petrel) and subsurface where showing different stops, subsurface being on the safer side. See attached defaultSubsurface.png.
> 
> Looking at the code, I saw the #define GF_LOW_AT_MAXDEPTH 0 and tried to set it to 1. Now, the stops are more similar. See attached maxDepthSubsurface.png.
> 
> Can somebody explain to me what is the meaning of this GF_LOW_AT_MAXDEPTH?

sorry, this mail has been on by “to be answered” pile for far too long.

First of all (maybe we should have a FAQ list for this): Everybody (and thus every program) does deco a little bit different. There is not THE correct procedure to get you out of the water safely. Everything that does not bend you and does not let you become too old while waiting for the permission to surface is a good algorithm. So there is not much point in exactly comparing deco times and depth between two implementations, even if they claim to do the same thing (e.g “Buehlmann with gradient factors” in our case).

Of course you can reasonably say things like “Computer A always gets me out of the water much faster than computer B”. Fair enough. If you do a lot of field research you might even have empirical evidence for statements like “The chance of getting bent following computer A is x% while it is only y% percent following the schedule of computer B”. I promise these percentages will never be zero, so the only thing you could reasonably do is to decide which compromise between getting out of the water quickly and probability of getting bent you accept. BTW, in recreational diving 1 accident in about 10000 dives is usually accepted. So there is not the correct deco.

It is a different thing to ask for the correct implementation of a given algorithm. The problem is: For decompression even the algorithms are not very well defined. The original Buehlmann algorithm is explained in great detail in a book by that author. Problem is, that algorithm has some parameters (admittedly only  a few) and there are several opportunities to add fudge factors for increased safety. For example you can always exaggerate your depth slightly when on-gasing. That book by Buehlmann not only contains the description of the algorithm but also decompression tables. Unfortunately, to compute those tables he added some of those fudge factors that are not documented in the book but rather his trade secret. So even with the “Buehlmann algorithm” (something that probably everybody agrees on what it is) does not reproduce the “Buehlmann tables”. 

And for implementation in dive computers (with limited computational abilities) different manufacturers came up with simplifications of that algorithm together with other fudge factors, none of them of course documented.

In addition there is some doubt that plain vanilla Buehlmann is a good algorithm since it has the tendency to lead to bubble formation something that is considered to be bad. Thus people came up with again other modifications or entirely new (bubble based) algorithms that all have in common to start decompression earlier at greater depths. One is RGBM which is not documented in any detail beyond marketing blah in the public. Then there is VPM-B which seems to be (I ranted about this earlier) defined in terms of a FORTRAN source code (and some text documents that I don’t find conclusive as to what actually has to be computed). I am not aware of any implementation in the open which is not directly based (as in machine translated) on that FORTRAN source. 

And then there are gradient factors which probably fall more in the “add fudge factors” to Buehlmann category. Those are defined in terms of a couple of text documents by Eric Baker which are sufficiently clear so that those can serve as a definition of that algorithm. Even if those are fudge factors they give very similar decompression profiles than the other two above. Which is good news in the sense explained in the beginning.

Basically they say that you only allow for a percentage of the gas overloading that Buehlmann would allow for. That percentage depends on depth with a given percentage at the surface (the GFhigh) and and a given percentage at depth (GFlow) and a linear interpolation in between. The question is now, what exactly is “at depth”? My first reading was that this is the max depth of the dive. A rereading of the Baker texts however shows that this is not the case, rather one should use the depth of the deepest deco stop. These two interpretations are what is selected by the “GFlow at maxdepth” option. 

I thing Figures 3 and 4 of http://liquivision.com/docs/deepstops.pdf  are quite clear: The GFlow = .2 (20 in our units) applies at the deepest stop (54m in Fig 3 rather than the max depth of 90m). So, I would say if you want to follow Baker, there is only one correct choice and that is “glow at maxdepth” off. I had it as an option only for the historical reason that I had it wrong before. 

So, what’s the upshot? Probably you can do both. But “Baker’s way” is with that option turned off. Maybe the Shearwater people did the same mistake as I initially. Who knows. Maybe that leads to good deco (quite likely, the empirical base is very very thin). Let me mention as well once more that the OSTC implementation is also not quite to the letter of Baker’s description. But still the differences are tiny and most likely not relevant. An of course, also Baker made up that algorithm off the top of his head. He wanted a modification of Buehlmann that leads to deeper initial stops and all three implementations do.

Hope that helps.

Best
Robert

PS: Some more material for further reading: I myself have written a text on deco considerations (in German), to be found here:
http://euve10195.vserver.de/~robert/dekotheorie.pdf

Recently, some navy people actually did a real study to test the benefits of deeper decompression, see https://www.dropbox.com/s/0lamtozt3v4xact/TR%2011-06_Deep%20Stops.pdf

The upshot being that the benefits are at least doubtful. At least for their type of diving.

This has lead to quite some discussion in particular on Rebreather World http://www.rebreatherworld.com/general-rebreather-diving/46994-deep-stops-debate-split-ascent-rate.html
where also the deco researcher including Bruce Wienke participated and Ross Hemmingway (author of V-planner, a vpm-b software) makes some claims that are not shared by the majority (to say the least).

-- 
.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oOo.oO
Robert C. Helling     Elite Master Course Theoretical and Mathematical Physics
                      Scientific Coordinator
                      Ludwig Maximilians Universitaet Muenchen, Dept. Physik
                      Phone: +49 89 2180-4523  Theresienstr. 39, rm. B339
                      http://www.atdotde.de

Enhance your privacy, use cryptography! My PGP keys have fingerprints
A9D1 A01D 13A5 31FA 6515  BB44 0820 367C 36BC 0C1D    and
DCED 37B6 251C 7861 270D  5613 95C7 9D32 9A8D 9B8F




-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.hohndel.org/pipermail/subsurface/attachments/20131122/369d6a68/attachment.sig>