Deco Theory 101, 201, 301, and 401

[KenGordon]

What I think would be REALLY interesting is an analysis of IS that decouples on-gassing and off-gassing rates. E.g. Instead of a 4-minute compartment, which is defined to be one that goes halfway from current pressure to (a higher) ambient in 4 minutes and then halfway from it's current down to a lower ambient in 4 minutes, maybe the compartment has an on-gassing half-life of 4 minutes, but an off-gassing half-life of 5 minutes.

Like a Suunto? Some of the navy ones too.

You REALLY should read the long and bitter deep stops threads on RB World, TDF (maybe also Yorkshire Divers) and a bit here. Ross (who is usually rossh) and supporters made a lot of arguments against IS and how it is calculated and they were demolished (maybe a bit strong) by various people. The coloured heat map style representation of saturation you see in Subsurface comes from graphics used to make some of the points. While it might have looked like an awful internet argument I found it helpful for my own understanding and I am grateful to Simon for putting the effort in.

A thing programmers need to be careful of is thinking in too much detail about the model. The model is only a model, it is not the thing.

 

[stuartv]

A thing programmers need to be careful of is thinking in too much detail about the model. The model is only a model, it is not the thing.

Understood. But, there is a simplistic model for distance traveled. It says that the distance traveled will be equal to speed times time. Then there is a more sophisticated model that factors in coefficient of drag and gives a more accurate model of the distance a speeding object will go. Neither model is really "the thing".

It seems to me that a model that assumes on-gassing and off-gassing behave in half-lives, and that the half-lives of on- and off-gassing will be the same is likely "simplistic" and, though it's not "the thing", a more sophisticated model might still be more accurate. Refinement of the model by adding model elements that have direct physical world analogs seems like a reasonable way to achieve better quality models. I don't KNOW, but I imagine that that is how things like flight simulators have evolved to their current state of the art.

 

[KenGordon]

Understood. But, there is a simplistic model for distance traveled. It says that the distance traveled will be equal to speed times time. Then there is a more sophisticated model that factors in coefficient of drag and gives a more accurate model of the distance a speeding object will go. Neither model is really "the thing".

It seems to me that a model that assumes on-gassing and off-gassing behave in half-lives, and that the half-lives of on- and off-gassing will be the same is likely "simplistic" and, though it's not "the thing", a more sophisticated model might still be more accurate. Refinement of the model by adding model elements that have direct physical world analogs seems like a reasonable way to achieve better quality models. I don't KNOW, but I imagine that that is how things like flight simulators have evolved to their current state of the art.

There are more sophisticated models available today. They have proven to be less popular, on SB at least, than the Bulhmann ones. Even ones that take account of apparently obvious improvements, such as heart rate, get dismissed. Suunto, who take a bit of account of various extra profile and time factors and might actually use asymmetrical on/off gas rules, gets completely slammed here.

One of the other big deco scientist blokes, whose name escapes me due to a lack of sleep, has a very nice diagram of the factors which contribute to DCS. Very many are not the profile. Adding those in might help an illusion of improvement but what we have, and basically have had for a century, works quite well, few people get bent.

 

[atdotde]

I would not think that introducing different on- and off-gassing rates would be an improvement. After all, the underlying model is not that there is exponential in time on- and off-gassing but that there is diffusion, that is that at any moment in time inert gas atoms/molecules go in and out of the tissue in random ways and that the rate of the ingoing ones is proportional to their number outside (i.e. the ambient pressure) while the rate of the ingoing ones is proportional to pressure inside. This happens all the time. But what you observe is only the net-rate, the ingoing ones minus the outgoing ones. And this net-rate is then proportional to the difference in pressures which in turn leads to the exponential decay of the pressure difference. But it's not like gas is ingoing if the ambient pressure is higher than the tissue pressure and outgoing in the opposite case. At any time it is both in- and out-going, but which one wins depends on the sign of the pressure difference. Thus there is no fundamental difference between the two and it makes no sense to treat them at different rates.

On a different matter, adding the contributions of the different tissues might be a step in a good direction. But a direct addition might not be warranted as it could well be that there is much more of one tissue in the body than another and thus the first would contribute much more. So at least one should take a weighted sum but then it's unclear how to determine the weights.

All such modifications (different rates, taking into account tissue weights, taking into account heart rate, diver weight, fitness level, you name it) all sound like they would have the possibility to better reflect reality. But they all come with different parameters. And those are notoriously hard to evaluate empirically and guesswork almost certainly does not help. You could frame the whole deep stop discussion as arguments about the best values of two additional parameters (the gradient factors), and even there, just two numbers, it took sophisticated experiments with lots of dives and then years of discussion about the proper interpretation just to have a somewhat better idea of the optimal values for just two additional parameters.

 

[stuartv]

There are more sophisticated models available today. They have proven to be less popular, on SB at least, than the Bulhmann ones. Even ones that take account of apparently obvious improvements, such as heart rate, get dismissed. Suunto, who take a bit of account of various extra profile and time factors and might actually use asymmetrical on/off gas rules, gets completely slammed here.

Speaking only for myself, I would shun the use of any algorithm that does not allow me to determine the complete ascent plan in advance of doing the actual dive.

Using heart rate as a parameter means that you have no way to know if what you planned in advance will match what your computer actually prescribes in the water and that is unacceptable (to me, anyway).

I would rather use an algorithm that allows for worst case on parameters like that, but gives predictable results.

I would not think that introducing different on- and off-gassing rates would be an improvement. After all, the underlying model is not that there is exponential in time on- and off-gassing but that there is diffusion, that is that at any moment in time inert gas atoms/molecules go in and out of the tissue in random ways and that the rate of the ingoing ones is proportional to their number outside (i.e. the ambient pressure) while the rate of the ingoing ones is proportional to pressure inside. This happens all the time. But what you observe is only the net-rate, the ingoing ones minus the outgoing ones. And this net-rate is then proportional to the difference in pressures which in turn leads to the exponential decay of the pressure difference. But it's not like gas is ingoing if the ambient pressure is higher than the tissue pressure and outgoing in the opposite case. At any time it is both in- and out-going, but which one wins depends on the sign of the pressure difference. Thus there is no fundamental difference between the two and it makes no sense to treat them at different rates.

On a different matter, adding the contributions of the different tissues might be a step in a good direction. But a direct addition might not be warranted as it could well be that there is much more of one tissue in the body than another and thus the first would contribute much more. So at least one should take a weighted sum but then it's unclear how to determine the weights.

All such modifications (different rates, taking into account tissue weights, taking into account heart rate, diver weight, fitness level, you name it) all sound like they would have the possibility to better reflect reality. But they all come with different parameters. And those are notoriously hard to evaluate empirically and guesswork almost certainly does not help. You could frame the whole deep stop discussion as arguments about the best values of two additional parameters (the gradient factors), and even there, just two numbers, it took sophisticated experiments with lots of dives and then years of discussion about the proper interpretation just to have a somewhat better idea of the optimal values for just two additional parameters.

All good points, making for interesting food for thought.

I am still curious, though, why the experts (e.g. Dr. Mitchell) are looking at the sum of the Integral Supersaturation from all 16 compartments, instead of looking at the Max IS from among all the separate compartments.

 

[dmaziuk]

On a different matter, adding the contributions of the different tissues might be a step in a good direction. But a direct addition might not be warranted as it could well be that there is much more of one tissue in the body than another and thus the first would contribute much more. So at least one should take a weighted sum but then it's unclear how to determine the weights.

You can't bend divers. Therefore you can't determine the weights based on empirical DCS incidence, so have to decide on a a proxy measurement you can derive from existing dive profiles, e.g. ask subsurface cloud users to donate theirs. Vast majority of them did not result in DCS so... what were we doing this for, again?

 

[atdotde]

Under controlled settings and with a chamber at hand you can (ethically) bend divers. But that is expensive, not only financially. The attempt to collect profiles from subsurface users for research never really got off the ground but it might be time to revive that. DAN seems to have a big collection of profiles with at least some information on outcome (bent or not). It would be great to dig into that. And of course, with this o'Dive Doppler device one could collect interesting data given that it provides more fine grained information than bent/not bent. On the other hand, the relation between bubble count and danger of DCS is still not 100% determined...

 

[dirkhh]

I will categorically block any user who tries to harvest data from the Subsurface Cloud. That explicitly includes credential sharing.
I will also block any IP address that attempts to collect data from multiple accounts. That would be a clear violation of the privacy policy that has been established for the Subsurface Cloud.

HARD NO.

 

[atdotde]

@dirkhh that was always understood. There was talk about a website where people could upload their xml file to "donate it". It was never intended to go through the subsurface cloud.

 

[stuartv]

Maybe Subsurface should have some data elements added so that users could record outcomes in some kind of structured way?

I mean, a checkbox for "I got bent on that dive" would be a start, but I bet you smart fellows could come up with something even better.