Computer Algorithms, Objective Evidence of Whether One is "Safer"

[guyharrisonphoto]

There is an interesting discussion going on in the computer thread about selecting an algorithm to justify the types of dives you want to do. For example, getting the most "liberal" computer possible for live-aboard multi-deep-repetitive dives over a week period.

This raised a question for me. Is there OBJECTIVE evidence, either statistical (incidents of DCS per diver using the algorithm) or medical (doppler studies) showing the relative effectiveness of the different common algorithms in computers. For example, the Suuntos are typically called "conservative" and I beleive they run some sort of RGBM? The Buhlman is considered middle of the road, and is in Oceanic and Uwatec (which I have) computers, the DSAT is considered very "liberal" and is in oceanic computers. I am sure there are others.

What I want to know is whether the objective evidence shows any real benefit to the more conservative algorithms? In other words, do the more conservative algorithms unnecessarily penalize bottom time for no objective benefit?

And yes, the scenario I am most interested in are multiple deeper dives (80-130 fsw) over several consecutive days, but any evidence would be welcome.

I have the same question, but in the context of decompression dives. I am not looking for theoretical discussion, but for objective studies or evidence.

If there are links please post them.

Thanks!

 

[Lorenzoid]

I'm posting mainly just to get subscribed to this thread. I suspect you have read the same threads as I have, where various people have stated repeatedly that despite one computer giving you shorter or longer NDLs than another for a given dive, there is not enough data to conclude that one computer will, on average, keep you "safer" than another. I believe this is especially true for "multiple deeper dives (80-130 fsw) over several consecutive days" because very little data has been collected from such dives, and the rate of DCS is already considered "low" over ALL dives for which data has been collected.

 

[redacted]

No data but one point. While we may consider computer A to be conservative and computer B to be liberal, I don't believe that means that for all dive profiles, computer B will provide more bottom time than computer A. There may well be conditions where the "more conservative" algorithm provides greater NDLs.

 

[Lorenzoid]

No data but one point. While we may consider computer A to be conservative and computer B to be liberal, I don't believe that means that for all dive profiles, computer B will provide more bottom time than computer A. There may well be conditions where the "more conservative" algorithm provides greater NDLs.

... which is why, based on not much more than gut feeling, I chose a Suunto. But the OP said he's looking for data, not a "theoretical discussion."

 

[tech_diver]

" While we may consider computer A to be conservative and computer B to be liberal, I don't believe that means that for all dive profiles, computer B will provide more bottom time than computer A. There may well be conditions where the "more conservative" algorithm provides greater NDLs."

This is an excellent point I was pondering while reading the other thread. In some cases, classifying an algorithm or computer as liberal or conservative is an oversimplification. The example I used when I used to teach (years ago but I think the example is still current):

PADI table gives a No D limit at 60' as 55 minutes which is more conservative than the US Navy table which read 60 minutes at 60'.
But the Navy tables don't clear you for 12 hours after a dive (max SI) which is more conservative that the PADI table which clears you after 6 hours (max SI).

So what does 'conservative' mean?


We can talk about a table's No-D limits as more or less conservative than another or we can talk about an algorithm being generally more conservative but keep in mind that these are just generalizations and may not describe the algorithm's behavior on a specific series of dives.

 

[TSandM]

To my knowledge, such data do not exist. There is a very small amount of work being done on technical divers, using Doppler bubble measurement. I don't know of any work being done in recreational divers. To create a good study to compare algorithms would be quite difficult, because the DCS risk is so low across the board. So you are either looking at a very large number of divers, or many, many dives, before you have a useful study. Even using Doppler, you need a large number of divers, because some divers bubble easily, and some won't bubble almost no matter what they do.

 

[Subcooled]

[abstract] PERFORMANCE OF DIVE COMPUTERS EXPOSED TO PROFILES WITH KNOWN HUMAN SUBJECT RESULTS

 

[JohnN]

[abstract] PERFORMANCE OF DIVE COMPUTERS EXPOSED TO PROFILES WITH KNOWN HUMAN SUBJECT RESULTS

The paper looks interesting. I'm not able to look at it, do you have any pointers, or could you repost the paper?

 

[guyharrisonphoto]

Yes, that certainly does look interesting. But, like John, I could not link to the actual paper from the abstract.

 

[Lorenzoid]

Looks like this is the original paper: http://dornsife.usc.edu/assets/sites/222/docs/computer-tests.pdf