Are rebreathers getting safer over time?
- Thread starter kr2y5
- Start date
Please register or login
Welcome to ScubaBoard, the world's largest scuba diving community. Registration is not required to read the forums, but we encourage you to join. Joining has its benefits and enables you to participate in the discussions.
Benefits of registering include
- Ability to post and comment on topics and discussions.
- A Free photo gallery to share your dive photos with the world.
- You can make this box go away
Going digital allows complex logic to be applied through software to the analysis of the analog output. A totally errant cell can be ignored, such as one with an internal short. The software can also ignore variances that are within tolerance or indicate cells that exhibit characteristics that are normal when they are nearing the end of their life.
You don’t need all this automation complexity in a hospital or even on a saturation diving system because you can check the cell against a cal gas any time things look questionable.
descent
Contributor
Thanks! Let me try to summarize some of the points, and see if I can get this right. To an unsuspecting passive observer: unless others here, I don't know what I'm talking about... so please, do take this with a 3000 lb block of salt, and then some. I apologize if anything I write turns out to be wrong or even offensive... please attribute that to my ignorance, and kindly enlighten a fellow diver...
First, there seems to be more or less consensus that human error is the root cause of most of the accidents on rebreathers. This would suggest that one should seek potential areas for improvement in whatever can reduce the potential for a human error. One way to classify human errors is as those that happen before the dive, and those that happen during the dive.
With respect to the former, it sounds like not a lot can, or perhaps even not a lot should be done (as discussed below). These are errors that happen during a time while one is not operating under time pressure, and does not have to make real-time decisions. Some people suggest that for a meticulous, very detail-oriented person, a rebreather is not inherently dangerous. I presume this is to say that a very careful person should be able to diligently follow all the pre-dive procedures with 100% accuracy, and that at least at that time, the rebreather is not trying to actively kill them (it can't). Hence, no need for an improvement over state of the art... whatever the technology, it ultimately depends on the person (or not?).
With respect to the latter (errors during the dive), it boils down to awareness and task-loading. Indeed, most of the areas that people seem to have debated, and those where there were either perceived improvements, or a perceived potential for improvement, had to do with real-time monitoring. Presumably, this is the area that one should mostly focus on...
More specifically... what happens before the dive:
1. Servicing. Much user error due to stale cells or scrubber. No change there happened, and neither is expected to happen, since it entirely depends on a human being meticulous. Just as a mind exercise... can the rebreather be capable of a self-test, e.g., only accept cells that report the time they were manufactured and refuse to boot when stale, or run some sort of a scrubber test? One can probably argue that having such capability would only make the system more complex, and added complexity would decrease rather than increase safety. And, that it would promote complacency when the diver grows reliant on such self-tests. So, this is probably a lost battle, it can't (or even shouldn't) possibly ever change...
2. Assembly. Thanks to past incident analysis, better checklists have been developed, which is an improvement, although no amount of technology can make a complacent diver go over the checklists, and everybody is prone to becoming complacent. Just as a mind exercise, again... would it even be desirable for the rebreather itself to run over some parts of these checklists? On one hand, the same argument as above applies: we wouldn't want divers to grow complacent. On the other hand, a shorter checklist probably means a diver is less prone to being lazy, since it's less of a perceived PITA. Is there some tradeoff here potentially to be exploited? And if so, do the state of the art solutions represent a sweet spot? Can more self-testing by unit itself lead to more complacency because divers grow to rely on it, or less complacency because the manual procedure would become shorter?
3. Planning. Some folks mentioned the use of conservative ppO2 and ENDs as a factor that helps mitigate risks. Bailout has also been mentioned, presumably in the context of having enough of it for any contingency... Could better planning, in general, do much to increase rebreather safety record? What other aspects of planning may help here?
Now, on to the stuff that happens during the dive:
4. O2 sensors, taken individually. Everyone seems to agree these are quite unreliable and a real risk, and some suggest that at the moment, this is the biggest potential area for improvement, although no improvement at all has occurred over the past decade. Some point out that the market is much to small to hope that a technology more suitable for diving applications to be developed in the near future (and that the same presumably applies in most of the other areas), so we are doomed to rely on technologies borrowed from other industries.
5. O2 sensors, taken as a system. There has been much improvement, moving from analog to digital, resulting, e.g., in less interference between O2 cells. Voting logic is used to work around the unreliability of the individual cells, although people disagree whether the trend to add more O2 cells actually represents an improvement. On one hand, with more cells, the failure of the system as a whole (the majority of cells bad) is less likely. On the other hand, some people are skeptical whether voting really solves the problem. Some people are suspicious of electronics, presumably treating the voting logic itself as a single point of failure (or so I speculate here, please clarify), and this (presumably) leads them to conclusion that even with multiple cells, at the end of the day one has to be able to manually monitor them all, which leads to more task loading. Overall, this seems to be one of the more controversial points in the discussion.
6. CO2 sensors. More units nowadays come with CO2 sensors, and they have gotten a bit better, although they are still quite unreliable. Presumably, this is another place where better technology could be expected to help... Some folks question how much this really matters, pointing to the fact that new scrubbers make it increasingly less likely to over-breathe the scrubber.
7. CO2 absorbents. Just as it was the case with the O2 cells, it is believed to be one of the areas where the most improvements are needed, but people seem to disagree on whether much improved here over the past decade.
8. Flooding. Everything, from O2 sensors, to electronics, is intolerant of moisture... unclear if any improvements happened, or could happen. Aside from the chances of some part of the unit flooding, is it possible to make the risk of different components flooding uncorrelated? If the O2 cells flood, will they all flood? Would it be possible to achieve more true redundancy, or this another case where trying to make any improvement would inevitably lead to more complexity, and that this would ultimately be detrimental?
9. Electronics. Although it wasn't very clear, I think it was implied that more redundancy, and the ability to self-test, represent an improvement, and that aside from flakey sensors, electronics itself might have gotten a little bit more robust? Is that mostly believed to be true? Also sounds like, not really a major area of concern?
10. Bailout. A few people pointed to bailout as the preferred way to deal with certain categories of failures, and the reluctance to perform bailout (presumably due to the lack of adequate training) as one of the major causes of accidents. Others seemed to disagree. I think it was implied that if bailout is always an option, it helps work around some of the limitations in other areas (e.g., issues with faulty sensors). Have improvements occurred in this area, and are any further improvements possible?
11. Monitoring, in general. Everyone seems to point to situational awareness and task-loading as the key areas that determine the risk of diving on a rebreather. Between monitoring ppO2, loop volume, and other variables, peeking at a buddy's console, etc., there's simply more information to keep track of and process in real-time. It sounds like the general trend has been to throw more information at the user, or make it easier to access: the head-mounted displays, additional O2 sensor outputs, CO2 sensors, etc., all increase the mount of signal that's available to the diver, and at least in principle, if one has the bandwidth to process all that input, it should help to make the rebreather safer. I guess it makes sense that, e.g., the presence of a CO2 sensor, rather than having to guess what the ppCO2 might be based on the workload and other factors, makes monitoring easier... less guesswork, lower potential for errors. This, I'd assume that it's noncontroversial that a rebreather collecting more information is beneficial. Where it appears to get controversial is, whether presenting all that information to the diver (as opposed to the system using some of it internally for processing) is beneficial or not.
With respect to he above, it sounds like there might exist two approaches. One stresses the fact that manually tracking of all information is better, presumably based on the belief that one can't really rely on the rebreather electronics to do it safely, or that the electronics is still a single point of failure, or that not everything can be tracked, and at that end of the day, the diver must be able to remain aware of all the key parameters, without relying on the box. For those, who think this way, more sensors aren't necessarily making things better, since they will have to process all that information anyway, which will add to task loading, and at the end of the day, it's not really helping... This is also the opinion that was expressed by the diver I talked to yesterday. It would seem that if one accepts that the diver must remain aware of all the parameters as a given, then the concern about task loading inevitably leads to the distrust for rebreathers, and little can be done to mitigate that in the future. It's a dead end.
Another approach, I suppose, is to start with the concern for task-loading and information overload, and accept that the diver's ability to process information is bounded. It would follow that as more input becomes available, rather than manually monitoring all input, the diver should rely to some extent on the electronics to digest and summarize at least all signals coming from his or her own unit (not others). Is it possible and desirable at least in theory, for the rebreather to monitor all key variables, and present information to the diver in an aggregate form? Or, are there things a rebreather couldn't (or shouldn't) possibly ever monitor, or where the technology needed to monitor them would be prohibitively expensive? Is it a productive direction or a red herring? Is more automation in information processing potentially helpful here, or is it inherently evil?
Finally, stretching that last point even a bit further, some hinted that there might be areas, in which one could begin to rely on the rebreather entirely (e.g., to scrub CO2 fast enough), presumably because the potential for failure is becoming acceptably low. Regardless of whether you agree with this (most folks I suspect will not), is this something that might be possible/desirable in the near future, with respect to any aspect of a rebreather's mechanics, in any area at all?
Thanks!
First, there seems to be more or less consensus that human error is the root cause of most of the accidents on rebreathers. This would suggest that one should seek potential areas for improvement in whatever can reduce the potential for a human error. One way to classify human errors is as those that happen before the dive, and those that happen during the dive.
With respect to the former, it sounds like not a lot can, or perhaps even not a lot should be done (as discussed below). These are errors that happen during a time while one is not operating under time pressure, and does not have to make real-time decisions. Some people suggest that for a meticulous, very detail-oriented person, a rebreather is not inherently dangerous. I presume this is to say that a very careful person should be able to diligently follow all the pre-dive procedures with 100% accuracy, and that at least at that time, the rebreather is not trying to actively kill them (it can't). Hence, no need for an improvement over state of the art... whatever the technology, it ultimately depends on the person (or not?).
With respect to the latter (errors during the dive), it boils down to awareness and task-loading. Indeed, most of the areas that people seem to have debated, and those where there were either perceived improvements, or a perceived potential for improvement, had to do with real-time monitoring. Presumably, this is the area that one should mostly focus on...
More specifically... what happens before the dive:
1. Servicing. Much user error due to stale cells or scrubber. No change there happened, and neither is expected to happen, since it entirely depends on a human being meticulous. Just as a mind exercise... can the rebreather be capable of a self-test, e.g., only accept cells that report the time they were manufactured and refuse to boot when stale, or run some sort of a scrubber test? One can probably argue that having such capability would only make the system more complex, and added complexity would decrease rather than increase safety. And, that it would promote complacency when the diver grows reliant on such self-tests. So, this is probably a lost battle, it can't (or even shouldn't) possibly ever change...
2. Assembly. Thanks to past incident analysis, better checklists have been developed, which is an improvement, although no amount of technology can make a complacent diver go over the checklists, and everybody is prone to becoming complacent. Just as a mind exercise, again... would it even be desirable for the rebreather itself to run over some parts of these checklists? On one hand, the same argument as above applies: we wouldn't want divers to grow complacent. On the other hand, a shorter checklist probably means a diver is less prone to being lazy, since it's less of a perceived PITA. Is there some tradeoff here potentially to be exploited? And if so, do the state of the art solutions represent a sweet spot? Can more self-testing by unit itself lead to more complacency because divers grow to rely on it, or less complacency because the manual procedure would become shorter?
3. Planning. Some folks mentioned the use of conservative ppO2 and ENDs as a factor that helps mitigate risks. Bailout has also been mentioned, presumably in the context of having enough of it for any contingency... Could better planning, in general, do much to increase rebreather safety record? What other aspects of planning may help here?
Now, on to the stuff that happens during the dive:
4. O2 sensors, taken individually. Everyone seems to agree these are quite unreliable and a real risk, and some suggest that at the moment, this is the biggest potential area for improvement, although no improvement at all has occurred over the past decade. Some point out that the market is much to small to hope that a technology more suitable for diving applications to be developed in the near future (and that the same presumably applies in most of the other areas), so we are doomed to rely on technologies borrowed from other industries.
5. O2 sensors, taken as a system. There has been much improvement, moving from analog to digital, resulting, e.g., in less interference between O2 cells. Voting logic is used to work around the unreliability of the individual cells, although people disagree whether the trend to add more O2 cells actually represents an improvement. On one hand, with more cells, the failure of the system as a whole (the majority of cells bad) is less likely. On the other hand, some people are skeptical whether voting really solves the problem. Some people are suspicious of electronics, presumably treating the voting logic itself as a single point of failure (or so I speculate here, please clarify), and this (presumably) leads them to conclusion that even with multiple cells, at the end of the day one has to be able to manually monitor them all, which leads to more task loading. Overall, this seems to be one of the more controversial points in the discussion.
6. CO2 sensors. More units nowadays come with CO2 sensors, and they have gotten a bit better, although they are still quite unreliable. Presumably, this is another place where better technology could be expected to help... Some folks question how much this really matters, pointing to the fact that new scrubbers make it increasingly less likely to over-breathe the scrubber.
7. CO2 absorbents. Just as it was the case with the O2 cells, it is believed to be one of the areas where the most improvements are needed, but people seem to disagree on whether much improved here over the past decade.
8. Flooding. Everything, from O2 sensors, to electronics, is intolerant of moisture... unclear if any improvements happened, or could happen. Aside from the chances of some part of the unit flooding, is it possible to make the risk of different components flooding uncorrelated? If the O2 cells flood, will they all flood? Would it be possible to achieve more true redundancy, or this another case where trying to make any improvement would inevitably lead to more complexity, and that this would ultimately be detrimental?
9. Electronics. Although it wasn't very clear, I think it was implied that more redundancy, and the ability to self-test, represent an improvement, and that aside from flakey sensors, electronics itself might have gotten a little bit more robust? Is that mostly believed to be true? Also sounds like, not really a major area of concern?
10. Bailout. A few people pointed to bailout as the preferred way to deal with certain categories of failures, and the reluctance to perform bailout (presumably due to the lack of adequate training) as one of the major causes of accidents. Others seemed to disagree. I think it was implied that if bailout is always an option, it helps work around some of the limitations in other areas (e.g., issues with faulty sensors). Have improvements occurred in this area, and are any further improvements possible?
11. Monitoring, in general. Everyone seems to point to situational awareness and task-loading as the key areas that determine the risk of diving on a rebreather. Between monitoring ppO2, loop volume, and other variables, peeking at a buddy's console, etc., there's simply more information to keep track of and process in real-time. It sounds like the general trend has been to throw more information at the user, or make it easier to access: the head-mounted displays, additional O2 sensor outputs, CO2 sensors, etc., all increase the mount of signal that's available to the diver, and at least in principle, if one has the bandwidth to process all that input, it should help to make the rebreather safer. I guess it makes sense that, e.g., the presence of a CO2 sensor, rather than having to guess what the ppCO2 might be based on the workload and other factors, makes monitoring easier... less guesswork, lower potential for errors. This, I'd assume that it's noncontroversial that a rebreather collecting more information is beneficial. Where it appears to get controversial is, whether presenting all that information to the diver (as opposed to the system using some of it internally for processing) is beneficial or not.
With respect to he above, it sounds like there might exist two approaches. One stresses the fact that manually tracking of all information is better, presumably based on the belief that one can't really rely on the rebreather electronics to do it safely, or that the electronics is still a single point of failure, or that not everything can be tracked, and at that end of the day, the diver must be able to remain aware of all the key parameters, without relying on the box. For those, who think this way, more sensors aren't necessarily making things better, since they will have to process all that information anyway, which will add to task loading, and at the end of the day, it's not really helping... This is also the opinion that was expressed by the diver I talked to yesterday. It would seem that if one accepts that the diver must remain aware of all the parameters as a given, then the concern about task loading inevitably leads to the distrust for rebreathers, and little can be done to mitigate that in the future. It's a dead end.
Another approach, I suppose, is to start with the concern for task-loading and information overload, and accept that the diver's ability to process information is bounded. It would follow that as more input becomes available, rather than manually monitoring all input, the diver should rely to some extent on the electronics to digest and summarize at least all signals coming from his or her own unit (not others). Is it possible and desirable at least in theory, for the rebreather to monitor all key variables, and present information to the diver in an aggregate form? Or, are there things a rebreather couldn't (or shouldn't) possibly ever monitor, or where the technology needed to monitor them would be prohibitively expensive? Is it a productive direction or a red herring? Is more automation in information processing potentially helpful here, or is it inherently evil?
Finally, stretching that last point even a bit further, some hinted that there might be areas, in which one could begin to rely on the rebreather entirely (e.g., to scrub CO2 fast enough), presumably because the potential for failure is becoming acceptably low. Regardless of whether you agree with this (most folks I suspect will not), is this something that might be possible/desirable in the near future, with respect to any aspect of a rebreather's mechanics, in any area at all?
Thanks!
No.... Digital (as in can bus) is just a transport protocol. It doesn't add to quality. In the dive computer, the same logic has to dissect the output of can bus, or plain analog cell ADC conversions.
Software has to have a set of rules and follow commands, to ignore this or that, or bias towards this or that. If the logic is sufficient to cater for all situations: planned and unknown, then your OK. But still it has a decision tree. Does it make averages? Or follow a direction as described by TSandM and Andrew Anslie?
Software is very good to watch for irregular situations, and then getting the divers attention, and maybe even a recommended course of action. The diver can then do a human interpretation and act accordingly. But then human error and desire and ignorance of risk creeps in, and disaster follows.
The machine is still dangerous, but the recent change is humans are being forced into a new appreciation of the danger.
In general, digital signal as a transport enables better error correction... so it definitely affects quality. Whether this matters in a rebreather I have no idea, but I think someone here (or elsewhere) hinted at interference between wires as one area where this matters (I may be wrong).
- Messages
- 20,561
- Reaction score
- 14,946
- # of dives
- I'm a Fish!
With digital systems you only have a very short analog run from the cells to their digital conversion. Millimeters really. In an analog system you have the whole run to the handset/HUD, so now you have long skinny wires with low voltage trying to stay accurate across a long distance while they are very close to each other and subject to interference. With digital you have digital signals being sent, it doesn't remove any difference in the cell quality, but it removes a lot of variables in signal quality so the units are now more reliable, that's just a fact, you'll never get anyone who knows what they're talking about saying that the analog electronics are better than digital because they'd just be wrong. Proof of this? Leon Scamahorn switched the Meg to digital..... Game over folks, anyone still building an analog unit is in the stone age for this.
Do we still have cell issues? Of course, I don't agree with AA's voting logic scenario because it just doesn't make sense. Do a dil flush, verify which cells are wrong, ignore the ones that are wrong, and dive off of the ones that are right. The dil flush is the same as running a cal-gas against the sensors, you know that you have 18% O2 in the dil, you dil flush and two cells are reading 18, one is reading 14, why the hell would I run my deco based off of 14? That's stupid. You KNOW that the cell with 14 is wrong, so you ignore it, the software knows it's wrong so it gets voted out, why pay attention to something that you KNOW is wrong? Makes no sense. If we couldn't do dil-flushes, then sure run based off of his logic, but we aren't. I respect AA, I really do, but he's a marketing/business guy, not an engineer. AJ's death about the 2 cells being wrong is an anomaly, but run a dil-flush, you verify that the one cell that is off is the correct one, and you either override the voting logic if you can, or you bailout, it's that simple. It sucks, it is expensive, but you can hop back on the loop at deco if the unit is still functioning correctly and you can manually fly it for better deco, or just stay on OC, that's why you always carry appropriate bailout and while you have to trust the system, you also have to verify it. Any anomaly in the cells at depth requires verification in order to figure out which one you can trust.
dreamdive
Contributor
As you stated in your first sentence.
I like having the Shearwater running 3 cells and my hud running an independent cell or two. For us, heading for the surface may not be the most desired option. I strongly feel that having the ability/option to add cells is an advantage for certain dives. We did not have the option on old units. Likewise, the divecan system allows for my hud to be independent of my controller, independent of my computer....So the chance of going "blind" is reduced.
The paradigm the Poseidon runs on is interesting, but it is not a unit I would do technical dives with - irrespective of what the manufacturer claims.
Similar threads
