"Accidental" deco with 1-day group, what to learn?

[Basking Ridge Diver]

True, but so far I wonder why. This is working.
I've heard the warnings. Some of those warnings are being handed down by people that certainly have a LOT more experience and knowledge than me. OTOH, the way I'm looking at it a lot of computers are black boxes, some of them have 'adjustments', and it's not at all clear WHAT THOSE ADJUSTMENTS ARE, or what tweaks have been made to the base algorithm.
When I change my Aladin I have no actual clue what I've done but it's now more conservative simply going from L0 > L1. I can't imagine diving it at L5, it would be teeth chipping conservative. OW1 students might be going in to deco.

We've made an adjustment. It's also becoming evident that we don't know what those adjustments will do across the entire course of a dive and even more on multi-day, multi-dive vacations. But we do know exactly WHAT we did. For now we dive it and watch. We aren't really going to know until that multi-day, multi-dive vacation.

Note - there is always a backup computer on board seeing how that's prudent and the damn thing has a frustrating tendency to refuse to cooperate once in awhile, usually the transmitter, but occasionally the main computer. Don't ever give up your analogue SPG!

Some day in the not to far distant future - you will come back and reread your thread - my guess is you will cringe at what you are saying...
Do some research - it is out there and all for the taking...

Safe diving

 

[danielmewes]

A more precise description would be: these penalty functions are not justified neither by the decompression model nor by experimental findings.

If my understanding is correct, that is not so.

Those adjustment functions are fitted to risk predictions made by the (real) RGBM model. This is why Wienke calls the M-value implementation of the RGBM that we have in the Suunto computers (except for the EON) RGBM "folded" onto a Haldanean model. Basically those functions were found to be suitable to reduce the predicted risk in recreational diving according to the (real) RGBM. So they are justified by the decompression model.
The RGBM itself is justified by physics, in-vitro and in-vivo measurements for many of its components, and certain simplifications and assumptions. But it is pretty clear at this point that it's not properly modelling some aspects of what's really happening. Buhlmann is almost certainly worse in how closely it models reality, but it has been more aggressively tweaked by now to fit actual data compared to the bubble models. I'm pretty sure that you could tweak the RGBM to for example not recommend deep stops anymore by tweaking the risk function that you're minimizing, but that's a separate discussion (maybe I will run some experiments to that extend at some point).

Please note that these issues of the RGBM in dive computers have nothing to do with dual-phase models. It's still a dissolved gas model, plus some penalty functions on top.

Only partially true I think. Yes, the computer doesn't model phase separation explicitly (except Suunto EON), but the adjustment functions encode predictions that are made by the dual-phase RGBM. So it has something to do with dual-phase models, in that the functions are derived from those models.

Secondly, VPM and RGBM have a tendency to assign less, not more, relevance to the slow compartments; that's the reason why they create deep stops.

That is not true. The RGBM assigns the same relevance to all compartments. In fact RGBM staging in most implementations I believe is controlled by the compartment with the highest predicted separated phase (including the surface period after the dive), regardless of whether that compartment is slow or not. Since slow compartments have more time to excite bubbles into growth during the surface interval, they are arguably more relevant in the RGBM. If you would vary the saturation in different compartments upon exiting the water while keeping their sum constant, the RGBM would assign higher risks to those outcomes where saturation is shifted into the slow compartments. Ironically, Buhlmann staging doesn't have any notion of bubble growth once you're out of the water, and hence doesn't model anything that would make you want to assign more importance to slow compartments. Instead these effects are meddled into M values that are set lower for the slower compartments. Don't get me wrong, that's a valid approach, and those values are fitted to actual experimental data I suspect. But again this is a case where deficiencies in the model are made up for by tweaking to fit actual data. In the RGBM, increased significance for slow compartments follows naturally from the model.

The reason that the RGBM creates deep stops is because it models gas diffusion into bubbles, and it assumes a population of bubble seeds with a non-uniform distribution (namely seed numbers proportional to exp(-beta r) with r bubble radius and beta a constant). Fast compartments tend to get more saturated more quickly, and, during decompression, this leads to more of the smaller bubble seeds to be excited into growth by gas diffusion into the bubbles. Since the RGBM postulates those small seeds to be very numerous compared to large ones, it penalizes high degrees of over-saturation super-linearly. Additionally, the RGBM models the Oxygen window as well as bubble surface tension, both of which further decrease the impact of compartments that only reach low levels of over-saturation.

If you were to model a saturation dive for example where all compartments reach about the same saturation, the RGBM would put a lot more focus on desaturating the slow compartments before surfacing (which of course would desaturate the fast compartments as well as a side effect, but they wouldn't control the risk estimate).