Dive computer algorithms

gcbryan · Jul 2, 2009

Are most dive computers just hardware with a standard algorithm ( one of several industry standard algorithms) implemented or are the differences we associate among the various computers mainly due to "tweeks" done in house?

When one computer allows 20 minutes at 100fsw on a first dive it's considered "liberal" but when that same computer goes into deco on a second dive and shows a 15 minute obligation after only one minute of exceeding the NDL now it's "conservative" I do understand that in some cases two different algorithms are being implemented depending on deco or no deco but that's not really my question.

What I'm interested in is whether or not this behavior is due to in house tweeks to the algorithm or is there a standard algorithm that just has this result as a part of the algorithm.

Some computers punish certain behavior more than others. Is this just a result of choosing one of several standard algorithms or is it a more company specific thing?

If a company uses Buhlmann ZH-L16 will their computers show the same results as any other company who uses this same algorithm or are there likely to be differences due to changes each company makes to this algorithm?

Rhone Man · Jul 2, 2009

In theory if two dive computers use the same alogrithm and the same sampling rate and the same settings, they ought to produce the same result....?

Or have I misunderstood?

gcbryan · Jul 2, 2009

I'm asking what really happens.

They all (or many) list the algorithm that they are based on but most different computer brands do act differently so I'm asking how much of what we observe is the listed algorithm and how much is software tweeks done to the algorithm by the computer manufacturers?

All the penalties for reverse profile, consecutive deco dives, going up too fast, etc...how much of that (if any) is in the listed algorithm and how much of that is added on by the computer manufacturer?

TSandM · Jul 2, 2009

My understanding is that they are actually different algorithms - maybe using different numbers for M-values, and different numbers of compartments and half-times. Some also incorporate some considerations for bubble dynamics, although I have read that the RGBM that Suunto uses is just a gradient-factor type tweak to the underlying Buhlmann type model.

Kern · Jul 2, 2009

This is what Wienke has to say about the implementation of RGBM in recreational computers in RGBM ALGORITHM OVERVIEW: CONCEPTS, BASES, VALIDATION, TESTING AND REFERENCES. It's a "modified RGBM recreational algorithms (Haldane imbedded with bubble reduction factors limiting reverse profile, repetitive, and multiday diving), as coded into Suunto, Mares, Dacor, ABYSS, GAP, HydroSpace, Plexus decometers, lower an already low DCS incidence rate of approximately 1/10,000 or less.". He go's on to ay that the Abyss software package "offers the modified RGBM (folded over the Buhlmann ZHL) and, especially, the full up, deep stop version for any gas mixture, has a fairly large contingent of tech divers already using the RGBM and has not received any reports of DCS"

He also states that, "within decometer implementations of the RGBM, not a single DCS hit has been reported in non stop and multidiving categories, beyond 300,000 dives or more, up to now."

I find it interesting that when anyone mentions how conservative computers are, Suunto is always there but Mares & others are almost never mentioned.

Sorry, posted this in the wrong place. I suppose it sort of fits so I'll leave it for now.

ptyx · Jul 2, 2009

Kern:
I find it interesting that when anyone mentions how conservative computers are, Suunto is always there but Mares & others are almost never mentioned.

According to this thread, and assuming the table is accurate, it looks like Suunto algorithm/parameters actually is the most conservative for NDL.

(Which doesn't mean it is for actual deco dives, but I'd be surprised if it wasn't the case).

As far as getting different results are concerned, I don't know much about the specific algorithms those computers use, but an algorithm is just a set of operations for computing your decompression. All of those algorithms try to simulate a human body, so there are a lot of 'magic numbers' injected to try to fit the mathematical model to what manufacturers/scientist believe is the reality. Those can differ wildly from one manufacturer to another.

A good analogy is the surface of a circle: formula is S=pi*r^2, but Suunto lawyers might decide that large circles are unsafe, and set pi to 2.5 instead of 3.14. (Re-reading the OP, you don't need that analogy - I'll leave it for others less mathematically inclined

)

Kern · Jul 2, 2009

Bryan, a look at the Suunto RGBM implementation over the years make me believe that there has been obvious tweaks to the original algorithm. Some of the newer version offer the oxymoronic 50% RGBM option to make them less conservative & their trimix computer will allow you to make the model more or less conservative.

As RGBM is a proprietary model owned by Wienke & used under license I doubt Suunto just added the tweaks it wanted but got Wienke to do it for them. Just supposition on my part as the nuts & bolts are a secret.

Kern · Jul 2, 2009

ptyx:
According to this thread, and assuming the table is accurate, it looks like Suunto algorithm/parameters actually is the most conservative for NDL.

Sure, but the Mares computer used in that table isn't using a version of RGBM.

gcbryan · Jul 3, 2009

Kern:
This is what Wienke has to say about the implementation of RGBM in recreational computers in RGBM ALGORITHM OVERVIEW: CONCEPTS, BASES, VALIDATION, TESTING AND REFERENCES. It's a "modified RGBM recreational algorithms (Haldane imbedded with bubble reduction factors limiting reverse profile, repetitive, and multiday diving), as coded into Suunto, Mares, Dacor, ABYSS, GAP, HydroSpace, Plexus decometers, lower an already low DCS incidence rate of approximately 1/10,000 or less.". He go's on to ay that the Abyss software package "offers the modified RGBM (folded over the Buhlmann ZHL) and, especially, the full up, deep stop version for any gas mixture, has a fairly large contingent of tech divers already using the RGBM and has not received any reports of DCS"

He also states that, "within decometer implementations of the RGBM, not a single DCS hit has been reported in non stop and multidiving categories, beyond 300,000 dives or more, up to now."

I find it interesting that when anyone mentions how conservative computers are, Suunto is always there but Mares & others are almost never mentioned.

Sorry, posted this in the wrong place. I suppose it sort of fits so I'll leave it for now.

I think Suunto is usually mentioned and not others just because they are the largest company for dive computers (I think Oceanic and Suunto are the largest...could be wrong). My Tusa IQ-700 is just as conservative at least using the dive planner numbers ...60 feet for 45 minutes on air!

Which brings up your quotes from Weinke about no reported DCS hits...I guess not if you shave 15 minutes off the Navy NDL numbers for 60 fsw!

Kern · Jul 3, 2009

gcbryan:
I think Suunto is usually mentioned and not others just because they are the largest company for dive computers (I think Oceanic and Suunto are the largest...could be wrong). My Tusa IQ-700 is just as conservative at least using the dive planner numbers ...60 feet for 45 minutes on air!

Which brings up your quotes from Weinke about no reported DCS hits...I guess not if you shave 15 minutes off the Navy NDL numbers for 60 fsw!

Absolutely, it depends a lot on what an individual thinks is a reasonable probability of taking a hit. 1 in 10,000; 1 in 100,000; 1 in 1,000,000; none.
Also we have to decide if Wienkes' figures are correct, or verging on the advotorial (sp).

On the other hand even the US Navy has a modified NDL table for sports divers based on Doppler studies. For 18 metres (~60') the NDL is 50 minutes.

So, should we use the models based on empirical studies of what's happening in our bodies, or use thel model based on what was thought to be happening instead? As all the recreational computers I know of have a neo/Haldaneian/Buhlmann algorithm at their base why not just use the most liberal, or the one we think looks the best? Beats me, they all seem to work. Maybe we're back to what probability of a hit we're willing to accept.

I'v done a little reading on this stuff over the years, but I have more questions now than when I started.

Dive computer algorithms

Please register or login

gcbryan

Contributor

Rhone Man

Contributor

gcbryan

Contributor

TSandM

Missed and loved by many.

Kern

Contributor

ptyx

Contributor

Kern

Contributor

Kern

Contributor

gcbryan

Contributor

Kern

Contributor

Similar threads