Deco for Divers: RGBM vs other algorithims

[scubadada]

What is the consensus on the RGBM used by many dive computers?

No consensus, it comes down to personal choice. I think it should be an informed personal choice rather than "that's the computer I see most people on the boat with"

For me, the Cressi, Mares, and Suunto RGBM algorithms are too conservative, they are obviously fine for many others. The Atomic Aquatics RGBM is considerably more liberal and they may have a wrist computer this year. There are several other decompression algorithms to choose from. The more you know, the better choice you can make

 

[KenGordon]

Ok, now it's making more sense.

So, I guess the big mystery question is "What is the RGBM doing?"

I'm surprised no one has reverse engineered their algorithm or no one has leaked the details yet.

Edit:

RGBM is supposedly proprietary, right? But to have any course to take legal action against someone unauthorized to use or duplicate it they would have to have a patent, right? And to have a patent you must publicly reveal and disclose what the idea or products is, right?

So, is there a patent? A quick google search turned up nothing.

You could google RGBM and follow the links from Wikipedia. Some of those are written by Bruce. One of those includes the following

"For recreational exposures, the RGBM collapses to a Haldane dissolved gas algorithm. This is reflected in the risk estimates above, where estimates for both models differ little."

My opinion is that no manufacturer wants it for performance reasons, but that it was good marketing in the past. So nobody would reverse engineer it, but people might license the name. For. Example I might sell a computer as "Ken RGBM" but implement Bulhmann with Pyle stops.

 

[CuzzA]

Or you could imagine the dives you might do and plan those both with the software your computer manufacturer supplies and something like Multideco and compare.

If they come out similar then no need to change computer.

At least if you dive the Mares plan (you would hope) then Mares claim that it is fit for purpose. With a GF based computer you take that on for yourself. Do you feel lucky?

Well, the Mares computer software is poorly written. It relys on you to draw a profile rather than input data.

I believe GAP is the only planner available that will compute the RGBM algorithm in a traditional data entry way.

Are you suggesting by using the Mares plan, I have possible legal recourse if I were to get seriously hurt, yet no recourse if I used GF's? I always felt like the risk always fell on the diver no matter what the computer said, unless of course you could prove a serious flaw.

 

[KenGordon]

Well, the Mares computer software is poorly written. It relys on you to draw a profile rather than input data.

I believe GAP is the only planner available that will compute the RGBM algorithm in a traditional data entry way.

Are you suggesting by using the Mares plan, I have possible legal recourse if I were to get seriously hurt, yet no recourse if I used GF's? I always felt like the risk always fell on the diver no matter what the computer said, unless of course you could prove a serious flaw.

I like being able to walk, it makes a lot of things much easier. Getting some payout from a dive computer manufacturer would never make up for not being able to walk. I really do not want to get bent.

What I mean is that a reputable manufacturer has to be able to justify their choices and so is likely to do due diligence when deciding on how conservative a computer will be. You do not have to do that when picking your GF numbers. You get to read stuff on the Internet and pick numbers accordingly. If you chose to do that and you get bent then you REALLY have nobody to blame but yourself, those posting that 60/85 or whatever was good for them owe you no duty of care.

If I were to sell a computer which defaulted to raw Bulhmann I would be doing it knowing it would bend more people than one set like a Zoop. If I did not make that very clear to users wouldn't it be fair for them to get together and try to sue me out of existence?

 

[Dr Simon Mitchell]

I would like to hear more from the community on gradient factors. One thing I found confusing/contradictory about his presentation is that from using the Navy study the conclusion was drawn that bubble models increase the chance of slow tissue DCS, yet GF are an attempt to mimic bubble models. So if that were the case, then why use GF at all, or if you're going to use them for conservatism purposes, shouldn't the first factor/percentage be much higher and closer to supersaturation to eliminate slow tissue on gassing? Are GF's simply a result of divers not willing to let go of the bubble model idea?

Also, while we all appreciate the Navy's study, can someone explain Richard Pyle's observations. Deep Decompression Stops

Is there a place for deep stops in decompression diving? For example, would there be any benefit to add say one deep stop at half your depth for a minute or two and then continue on with a straight buhlman strategy (maybe adjusted for conservatism)?

Hello,

These are good questions.

The issue of why use GFs at all is complex. Here is my take on it.

When bubble models emerged among technical divers in the late 1990s / early 2000s there was a perfect storm of factors that led to their widespread adoption. One was the fact that the widely used alternatives around at that time still resulted in "unexpected" DCS on occasion. Another was the theoretical attraction of the bubble model concept; that is to say that it seemed to make sense that the approach would work, (though this was NEVER tested). Yet another was Rich Pyle's contemporaneous observation that the insertion of a deep stop in his own practice seemed to benefit him (I will return to that issue below). Finally, a number of influential figures around at the time enjoyed getting on the bandwagon; I presume to seem knowledgeable and at the cutting edge. These people aggressively promoted the concept in the absence of any supportive data. Under these influences the use of bubble models, or GF modifications to make Buhlmann look more like a bubble model became very popular, indeed, almost ubiquitous. It went somewhat unnoticed that one of the most popular bubble models required adjustment to its early iteration because of an apparently high incidence of DCS associated with its use.

If we fast forward to today there are now two human studies which have demonstrated that a bubble model approach (somewhat against expectations) either resulted in more venous gas emboli after surfacing or a greater incidence of DCS. There have been extensive arguments about these studies (particularly the NEDU study) on the internet, but cutting through all the confusing detail, there is a strong signal in the data that emphasising protection of the fast tissues from supersaturation early in an ascent (which is what deep stops do) does not provide the hypothesised advantages. Indeed, it appears disadvantageous because it results in greater supersaturation of slower tissues (which continue to take up gas during deep stops) later in the ascent. Notwithstanding the argument that the NEDU dive profiles are dissimilar to those dived by technical divers, it has been convincingly demonstrated that this same disadvantageous pattern of tissue supersaturation identified in the NEDU study is seen in undisputedly real-world technical diving profiles controlled by a bubble model compared to a common GF approach (illustrated using Kevin Watts’ diagrams in my presentation that someone linked to on this thread).

This brings me to your question about why we should use GFs to impose any stops deeper than Buhlmann prescribed. The answer is that use of deep stop approaches became an “industry standard” as I have described above. The available data now tell us that use of a bubble model that emphasises deep stops is almost certainly not optimal. But these studies were not designed to actually define optimal decompression. We don’t know how raw Buhlmann would compare and we cannot exclude the possibility that some degree of stopping deeper than he prescribed might be useful. Thus, in the prevailing setting where the vast majority of divers have been using deep stops it seems sensible to be cautious in the approach to moving away from emphasising them. As I have said, my own approach is to raise the GF low to around 50. That is still a deep stop compared to Buhlmann, but it is not as deep as would be prescribed by a bubble model. I agree with those who have said that these adjustments are only going to have a significant impact on dives requiring substantial decompression.

I would like to briefly mention the issue of Richard Pyle’s “deep stops”. His practice of inserting one or more stops into an ascent deeper than his prevailing ceiling has often been confused or equated with a bubble model approach to decompression. This is incorrect. They are two different things. I talked at some length with Rich about this last year, and it is a source of some annoyance to him that “Pyle stops” are often equated with bubble model decompressions. He does not use a bubble model. I do not have a strong opinion either way on Rich’s practices because it has never been tested in any formal manner.

Finally, there are several other studies of relevance to this matter that will hopefully be published in the near future.

Simon M

 

[dmaziuk]

So, I guess the big mystery question is "What is the RGBM doing?"

Actually it's what is your RGBM computer is doing. It is a common practice to approximate compute-intensive algorithms with simpler "close enough" formulae. The pathological case is to replace the calculations with a lookup in a pre-computed table: zero compute cost, at the expense of extra storage for the table. There's in-betweens with curve-fitted intervals and pre-computed boundary points.

And then you could program it all into software running on a general-purpose CPU, like cellphone chips inside OSTCs (?), but low-end devices have it burned into firmware adapted to whatever microprocessor they picked. These days it's cheap enough to even bake it into a puprose-made microchip (ASIC).

So even if you know what the model is dong, an implementation esp. in a low-end device, is something else. And black-box testing of it is not easy because you have to build a hardware platform that will feed the device's pressure sensor input and somehow hook into the outputs.

So I'd say for all practical purposes what RGBM is doing is magic. (According to @RonR, with the exception of Cobalt, that's there to reinforce the rule.)

 

[fsardone]

And then you could program it all into software running on a general-purpose CPU, like cellphone chips inside OSTCs (?), but low-end devices have it burned into firmware adapted to whatever microprocessor they picked. These days it's cheap enough to even bake it into a puprose-made microchip (ASIC).

Just in case, OSTC is running ZHL-16 (plain and GF) and does it (according to the variant) in oc or ccr mode (static and dynamic setpoint). You can buy an implementation of VPM from Ross for the high end OSTC. Fact is the VPM has an higher computational load and requires a more powerful cpu.

Also since they are open source (not the VPM implementation) you can actually look at the innards of the calculations. Somebody has implemented the computations on mac and you can simulate dive computer computations and compartment loading while dry at yor desk. Huge advantage during planning.

Most of the time I plan my tables with VPM and compare with buhlman GF 40/80. I like to align the deeper stop VPM gives me for a given profile with the deeper stop buhlman GF calculates by tweaking GF LOW, then leave GF HIGH at 75/80 and do the longer shallow stops.

More deco? So be it! Also the OSTC has the possibility to change GF while diving, so in case of emergency I can switch alternate GF 60/85 shorten up deco and get out of the water at higher risk of DCS: if bent you can go in a chamber if drowned you can only go in a coffin...

In my view I do not think I am using ZHL-16 GF to simulate dual phase alghoritms like VPM or RGBM but using those to increase the dept a dissolved gas alghoritm calculates its first stop and then do the longer deco time. This for me is important when I dive deeper. In a former thread on deco, Dr. Mitchell pointed out that, while there is a widespread belief that the presence of Helium imposes longer deco times, a NEDU comparative study between trimix and heliox dives proved that trimix is not more efficient than heliox.

The ensuing discussion highlighted the probability that any dive requiring helium especially in high percentages is a longer, deeper dive and the deco obbligation is probably higher than initially tought, therefore the so called helium penalty in reality is a time and depth penalty.

Bottom line I am happy to take the deeper stops of the dual phase model and the longer shallow stop time of the dissolved gas model.

Cheers

 

[filmguy123]

Based on all this...

For my first time dive computer as a new diver, I'm looking between the:

Suunto Zoop Novo ($330), proprietary RGBM computer

VS

Aqualung i300 ($300), with their algorithim: "Dr. Lewis wrote the proprietary Pelagic Z+ (PZ+) algorithm employed in Aqua Lung’s computers. The PZ+ algorithm is based on the Bühlmann ZHL-16C algorithm and is a favored recreational algorithm for the safety conscious diver."

I am also open to other options within a $350 price ceiling limit.

Not sure if it's worth shopping a dive computer based on the Buhlmann algorithim OVER the RGBM for non tech diving at a novice level, or better to have such a ubiquitious system such as the Suunto ZOOP line.

Thanks for input!

 

[KenGordon]

I use two Suunto computers and a Shearwater computer. I would consider OSTC a and Seabear computers too. While I own a lot of Aqualung kit I would not personally buy a computer from them or any of the other "we sell everything" brands. Developing fins, wings and even regulators is not at all like developing software.

Neither Suunto or Shearwater are perfect. Suunto has constant problems breaking their PC software and particularly the drivers for downloading dives. Some releases of the DM software go significantly backwards. The Shearwater PC software is really only for downloading dive details and firmware updates, it cannot plan a dive, you need to buy third part software and hope that they mean the same thing by ZHL16C.

The Shearwater computers only have two buttons. This makes the UI a bit crappy sometimes with the options being next, select or 'wait for a while to revert'. The classic Suuntos have three or four buttons and usually provide a back or quit button which can make discovering where stuff is in the menu structure easier. They are still not excellent, I find navigating getting details in the dive history and then moving to do the same in another dive tedious. I may be doing it wrong of course.

 

[CuzzA]

My advise would be if you were pretty certain you are not going to advance your diving further than basic recreational profiles I'd just buy a $150 Mares Puck and call it good, use the money saved for more dives.

If you're confident you want to advance your diving, but are capped at $350, take a look at the NitekQ. A little more than $350, but not much considering the computers capabilities.

Both computers are pretty well reviewed.