Why no accurate computers?

[BRT]

thats not a rec dive which the OP ,, I think,,, is referencing to because of the ;mention of NDL's , You can not give a tech answer to a rec question. What he is seeing is common. The problem lies in the overal conservativess of the computer on top of the conservative setting.

I also believe that if you got the lawyers out of things they all would read much closer to each other.

I have read articles that were based o n the excessive padding of NDL info and they suggested that low conservatism is more than adiquate, inferring med and high unnecessary/ way overkill.

I've never been more than a minute into deco.

 

[drk5036]

I've never been more than a minute into deco.

I know what you meant of course, but I do agree it’s misleading. I think it’s clearer to say that the standard dive is a 60 minute dive with a maximum depth of 100 ft.

 

[KenGordon]

Based on my understanding that it was based on the limitations of computer processing, I have been predicting that computers would make that switch as computer processors improved.

ZHL16A, B and C all have identical processing requirements. The difference is just values of constants used in the middle compartments. Some wiki article claims this difference but it is untrue. In fact all the simplistic dissolved gas models are practically identical resources wise, just changing with the number of compartments. GF adds some extra multiplies to interpolate the limits.

if you want the same probability of injury over a large number of dives then the computer model has to make up for not rounding depth and time. Imagine 1000 dives to 31m for 31 minutes. On tables the calculation would be the based on 33/33 and quite padded vs the actual profile so result in fewer injuries than the computer. Thus the computer model needs to back off a little bit.

edit - reading on I see the B vs C business has been well covered.

 

[ThatDogDontHunt]

ZHL16A, B and C all have identical processing requirements. The difference is just values of constants used in the middle compartments. Some wiki article claims this difference but it is untrue. In fact all the simplistic dissolved gas models are practically identical resources wise, just changing with the number of compartments. GF adds some extra multiplies to interpolate the limits.

if you want the same probability of injury over a large number of dives then the computer model has to make up for not rounding depth and time. Imagine 1000 dives to 31m for 31 minutes. On tables the calculation would be the based on 33/33 and quite padded vs the actual profile so result in fewer injuries than the computer. Thus the computer model needs to back off a little bit.

edit - reading on I see the B vs C business has been well covered.

There is the family of ZHL8 algorithms for lower power processors. Uwatec Aladin Pro uses this. It has 8 compartments instead of 16 so likely takes half the power?

 

[dmaziuk]

There is the family of ZHL8 algorithms for lower power processors. Uwatec Aladin Pro uses this. It has 8 compartments instead of 16 so likely takes half the power?

Yes, it will take roughly half as many CPU cycles and memory bytes. But none of the Uwatec's stuff's: "-8", "ADT", whathavethey, been published anywhere I could find it, who knows what's really in there.

ZH-L12 that was published by Herr Doktor himself, has 16 compartments but only 12 M-value constants (i.e. the amount of compute power needed is the same as for ZH-L16). That was derived from empirical data, i.e. the actual dives.

My understanding is ZH-L16 is a result of a "curve fit" exercise that produced a magic number: combined with the tissue compartment's half time, it gives you the M-value numbers for that TC (the "A" set). With that you can design an implementation with any number of TCs with any half-times. Including Uwatec's 8.

 

[lowwall]

My understanding is ZH-L16 is a result of a "curve fit" exercise that produced a magic number...

This is an important point. Everything about deco science as applied to the human body is subject to false precision errors. Computers can easily calculate a given algorithm to tiny fractions of a second. But what are those algorithms built on? Tissue compartments are approximations, perfusion models are approximations, bubble models are approximations, M-values are approximations, even ideal gas laws are approximations of the real world.

To their credit, computer manufacturers do understand this. Which is why NDL and stops are rounded to minutes rather than the microseconds the algorithm produces. And why they all add fudge factors of various sizes. I hope divers realize this and don't treat the NDL as a binary "safe/unsafe" display.

Personally I prefer a computer that displays aggressive NDL numbers. I always add a margin of safety by extending time spent in shallower waters at the end of the dive anyway. I don't have a Shearwater, but their Surface GF display sounds like an excellent idea.

 

[EFX]

This is an important point. Everything about deco science as applied to the human body is subject to false precision errors. Computers can easily calculate a given algorithm to tiny fractions of a second. But what are those algorithms built on? Tissue compartments are approximations, perfusion models are approximations, bubble models are approximations, M-values are approximations, even ideal gas laws are approximations of the real world.

To their credit, computer manufacturers do understand this. Which is why NDL and stops are rounded to minutes rather than the microseconds the algorithm produces. And why they all add fudge factors of various sizes. I hope divers realize this and don't treat the NDL as a binary "safe/unsafe" display.

I would like to address some misunderstandings here for the purpose of giving new divers a better understanding of dive programs. I agree with the first paragraph in that the results of our dive computer programs do not reflect what is actually happening in the human body. The program is a mathematical model that approximates (to a large extent for safety) what happens in the human body.

What the algorithm produces is what the equations for tissue pressure and ceiling depth require. NDL and stop depth and times are dependent on these equations. Tissue pressures are calculated from the Schreiner (or Buhlmann) equation where tissue compartment half-time and time at depth is in minutes. An algorithm is simply a series of instructions to produce a result. Here is the simplified version of an algorithm to calculate NDL time remaining from my spreadsheet:

ceiling ;the ceiling at the current depth has been calculated
time = 0 ;set the time at depth to 0 minutes

Do While (ceiling) < 0)
time = time + 1
Call tc_pressures
Call calc_ceiling
Loop

The time above is passed to a procedure called tc_pressures which calculates the new tissue compartment pressures for the added time at depth. Another procedure, calc_ceiling, calculates the ceiling in feet or meters (units are user selectible). The Loop command returns to the first statement which compares the new calculated ceiling with zero. If deco has not been reached the ceiling will be some negative value. The algorithm keeps looping, adding 1 minute to the time at the current depth until the newly calculated ceiling increases from some negative value (which actually represents a "depth" above the surface) to a value of zero. Zero represents the ceiling at the surface. NDL is defined as the time remaining at the current depth where an ascent directly to the surface can be made. At this point the loop exits and the value in time represents the NDL at the current depth.

The computer is running this loop in microseconds but the results are in minutes because the algorithm is designed to produce minutes. There is no rounding from one precision of time to another. Where rounding does occur is after a ceiling has been calculated. For deco the ceilings will be positive. The algorithm will round up (make deeper) the first ceiling stop which is usually a multiple of 10 ft or 3 m. Once that stop depth is calculated successive shallower stops will be made by subtracting 10 ft or 3 m from the current stop all the way to the surface.

My spreadsheet does not use fudge factors for its base calculations. Since it's designed for educational purposes there is a feature to alter (fudge?) some program constants. For example, the Schreiner equation requires an inspired inert gas pressure (insP) at the current depth. But, when ascending or descending to a new depth what insP do we use? Do we use the starting, median, or final depth? I allow the user to select which of these three choices to use which in turn will alter the program results.

 

[scubadada]

This is an important point. Everything about deco science as applied to the human body is subject to false precision errors. Computers can easily calculate a given algorithm to tiny fractions of a second. But what are those algorithms built on? Tissue compartments are approximations, perfusion models are approximations, bubble models are approximations, M-values are approximations, even ideal gas laws are approximations of the real world.

To their credit, computer manufacturers do understand this. Which is why NDL and stops are rounded to minutes rather than the microseconds the algorithm produces. And why they all add fudge factors of various sizes. I hope divers realize this and don't treat the NDL as a binary "safe/unsafe" display.

Personally I prefer a computer that displays aggressive NDL numbers. I always add a margin of safety by extending time spent in shallower waters at the end of the dive anyway. I don't have a Shearwater, but their Surface GF display sounds like an excellent idea.

Different decompression algorithms, with or without conservation factors applied, do, indeed, give different NDLs for the same exposure. These are often described on a scale from conservative (shorter) to liberal (longer). There is a probabilistic decompression algorithm, SAUL, that has previously been discussed on SB. There is a SAUL recreational dive planner that allows one to calculate the probability of DCS for a known exposure SAUL Recreational Dive Planner | Modern Decompression

Sometimes, it is useful for me to think about the P(DCS) for a range of possible NDLs. For a dive to 60 feet on air, the DSAT NDL is 57 min. The P(DCS) for this dive is 0.35%, 1/286. For a Buhlmann ZH-L16C GF high of 95, the NDL is 51 min and has a risk of 0.27%, 1/370. A GF high of 85 has a NDL of 41 min, and a risk of 0.15%, 1/667. A GF high of 75 has a NDL of 32 min and a risk of 0.06%, 1/1,667, These calculations are all for 1st, clean dive and have a 3 min SS at 15 feet included. These kinds of calculations give me a different perspective on the interpretation of NDLs, another way of thinking about it

Modern dive computers allow us to adjust are own risk. Some of us dive our computers at our chosen GFs, but then spend time at our safety or last deco stop to reduce our risk at the end of the dive. This uses the SurfGF mentioned by my friend, @lowwall above

 

[Gareth J]

This is an important point. Everything about deco science as applied to the human body is subject to false precision errors. Computers can easily calculate a given algorithm to tiny fractions of a second. But what are those algorithms built on? Tissue compartments are approximations, perfusion models are approximations, bubble models are approximations, M-values are approximations, even ideal gas laws are approximations of the real world.

To their credit, computer manufacturers do understand this. Which is why NDL and stops are rounded to minutes rather than the microseconds the algorithm produces. And why they all add fudge factors of various sizes. I hope divers realize this and don't treat the NDL as a binary "safe/unsafe" display.

Personally I prefer a computer that displays aggressive NDL numbers. I always add a margin of safety by extending time spent in shallower waters at the end of the dive anyway. I don't have a Shearwater, but their Surface GF display sounds like an excellent idea.

I have dived with a few people who use two identical dive computers with two different GF setups.
Computer 1 - Normal diving, GF's
Computer 2 - Very aggressive GF's 100/100

Under normal circumstance, the dive is carried out using computer 1.
If there is an incident, gas loss, decompression gas loss, bailout, medical issue, etc. And the diver needs to exit ASAP, i.e. before running out of gas or because of some other emergency. Then computer 2 is followed, with the expectation that a bend is likely.
This follows the rule that "we can fix a bend, we can't fix drowning".
It is also known as the KYAGB (kiss your ass good bye) plan.

 

[NAM001]

I have dived with a few people who use two identical dive computers with two different GF setups.
Computer 1 - Normal diving, GF's
Computer 2 - Very aggressive GF's 100/100

Under normal circumstance, the dive is carried out using computer 1.
If there is an incident, gas loss, decompression gas loss, bailout, medical issue, etc. And the diver needs to exit ASAP, i.e. before running out of gas or because of some other emergency. Then computer 2 is followed, with the expectation that a bend is likely.
This follows the rule that "we can fix a bend, we can't fix drowning".
It is also known as the KYAGB (kiss your ass good bye) plan.

that is done for the most part when you do conservative settings. it says you are at the limit and you are really not,, so in an emergency you may not have actually crossed a deco line that gets you bent. So long as GF data is reliable, one can make that call as to whether to call the chamber man or accept the surf value as the actual valid / acceptable reading.