Are rebreathers getting safer over time?

[tbone1004]

I do prefer a stick . . . always have, still do, even though I drive an automatic.

It's interesting to me that, despite the fact that it's clear a lot of RB accidents are user error, and the diver is the weak link, most of the folks I know who have been diving RBs safely for a long time have opted for a RB which is as manual as possible. It's a clear lack of trust of electronics and computer processing of information, despite the fact that most of the complex machines we operate nowadays are heavily managed by computer systems (admittedly, not underwater!).

I'm with you on the stick, but I'll never understand why so many divers are adamant they can do it better than the computer can. I see having the option to run it manually in an emergency, but will truly never understand why people want to run these things 100% manual all the time except to beat their chests and say they run mccr because the computers aren't good enough. If they weren't, they'd still be driving carbureted cars, have manual thermostats, no dive computers or bottom timers, analog only gauges, etc etc. The computer still has to spit out the voltage readings on the cells, what makes them any better than the solenoid to keep a set point?

 

[The Chairman]

SeaBear has been testing a prototype solid state sensor in their rebreather for a while. They are very hush hush about it, so I wouldn't expect anything for a few more years.

It hasn't been that 'hush, hush' it seems. Calculated leaks that entice are good for when they finally do release it. However, they aren't the only ones trying to push this envelope. I know of two other endeavors to create this beast and I think it's great. I think we'll see reliable CO analyzers in the next few years as well, although I think that's not nearly as important as the O2 cells. Even with my limited experience on rebreathers, I feel increased CO right away just as I do on open circuit, but no one feels the lack of O2 until they pass away.

It's my humble opinion that check lists are the best thing to increase safety at the moment. We've got to get past the pride that inhibits people to use check lists as well as electronics. I remember Pyle commenting in RB2 that PDCs have improved so much that they are more right than the humans second guessing them. He was even able to quantify that contention.

 

[Akimbo]

I'm with you on the stick, but I'll never understand why so many divers are adamant they can do it better than the computer can…

I’m not promoting one over the other but even mildly complex automated systems are inherently more difficult to accurately diagnose and detect malfunctions quickly. I think it will change but these systems will have to prove themselves over time. I’m also not saying that we are “there” yet.

How many people still carry tables in addition to a primary and backup computer? How many people would bet their life on their smart phone? An errant computer problem on a rebreather isn’t like dropping a call or the battery dying much faster than expected.

 

[KevinNM]

Automated controls without a way to monitor and a well understood and easy to use override result in things like you crashing a brand new airplane when the flight control system decides that you really wanted only one of your four engines to be operating. And an Airbus flight control system almost certainly has a lot better code review and testing then a rebreather control system.

 

[tbone1004]

but the rebreather is only controlling one thing. All the rebreather has to do is know the target mV level of the O2 sensors, monitor them to know what they are at, and decide to fire a solenoid to adjust the mV levels. It determines this level from two inputs, the ambient pressure sensor, and the O2 cells. That's the only control it has, and only one it needs because there is nothing else for it to do. CO2 sensor, all that is is an if/then statement for a threshold and if it meets or exceeds, it fires a warning, and it may give real time output levels, pretty easy. Temp stick, same thing, monitors and outputs the value, but neither of those effect the fundamentals of the system.

If mV of cells is low, fire solenoid until desired mV value is reached. Desired mV value is determined by user input PO2 value, calculated by mV value from cells with the pressure reading from the depth sensor. Desired PO2=user input value. Actual PO2=mV*pressure

Once desired mV value is reached, stop firing until it gets below threshold. Diver should have manual add button and be able to disengage the solenoid in event of an emergency, but you can't compare this to the ECU in your car, or the avionics and engine management in an airplane. They are orders of magnitude more complex. This is the type of stuff that spooks people and makes them not trust the computer. You can choose to not trust it on deco algorithms, fine. You can choose to not trust the temp stick or the CO2 sensor, fine. Not trusting a simple circuits ability to read a voltage level on an O2 sensor, not fine. Not trusting the O2 sensors themselves, shouldn't be diving a rebreather, or nitrox, or anything but air.

Not trusting a very simple coding structure of measure two values, multiple them together, and if that value is not the target, fire the solenoid to adjust is ridiculous. This is one of the most simple machines out there. APD and Shearwater have decided to make them more reliable by removing as much analog transmitting as possible and replacing it with digital. Actual current going across long skinny wires in a bundle across connections is not reliable. 0/1's going down fewer thicker wires is perfectly OK regardless of how many connections because the computers can diagnose the signal and know if it is right or not.


Summary, if you don't trust the rebreather to fire a damned solenoid, you better be the same guys still diving tables and bottom timers and refusing to trust computers because if you aren't you are just being ignorant and hypocritical. It is literally that simple, and I apologize if that offends anyone, but no human can drive a rebreather as accurately as the computers can, it's not possible. I'm not aware of any recent incidents *since the Vision electronics and DiveCan's came out* where the computers were the cause of the accident. It was either O2 sensors, or the idiot diving it. The digital systems remove a little bit of the unreliability to the O2 sensors because now their monitored values can't be wrong. Their readings may be wrong, but the computer will always know exactly what the sensor is reading instead of running the risk of interference/voltage drop over connections etc. That alone is a huge step forward in their safety/reliability because the computers are now always aware of the ACTUAL readings at the cells. Again, those readings may not be accurate, but at least you can trust that the computers are always spitting out the actual values or no values at all. I'd rather have a computer just say "nope, sorry, can't get the signal, I have no idea what the hell is going on, bail out", then start to give me flaky numbers or wrong numbers.

 

[victorzamora]

but the rebreather is only controlling one thing. All the rebreather has to do is know the target mV level of the O2 sensors, monitor them to know what they are at, and decide to fire a solenoid to adjust the mV levels. It determines this level from two inputs, the ambient pressure sensor, and the O2 cells. That's the only control it has, and only one it needs because there is nothing else for it to do. CO2 sensor, all that is is an if/then statement for a threshold and if it meets or exceeds, it fires a warning, and it may give real time output levels, pretty easy. Temp stick, same thing, monitors and outputs the value, but neither of those effect the fundamentals of the system.

If mV of cells is low, fire solenoid until desired mV value is reached. Desired mV value is determined by user input PO2 value, calculated by mV value from cells with the pressure reading from the depth sensor. Desired PO2=user input value. Actual PO2=mV*pressure

Once desired mV value is reached, stop firing until it gets below threshold. Diver should have manual add button and be able to disengage the solenoid in event of an emergency, but you can't compare this to the ECU in your car, or the avionics and engine management in an airplane. They are orders of magnitude more complex. This is the type of stuff that spooks people and makes them not trust the computer. You can choose to not trust it on deco algorithms, fine. You can choose to not trust the temp stick or the CO2 sensor, fine. Not trusting a simple circuits ability to read a voltage level on an O2 sensor, not fine. Not trusting the O2 sensors themselves, shouldn't be diving a rebreather, or nitrox, or anything but air.

Not trusting a very simple coding structure of measure two values, multiple them together, and if that value is not the target, fire the solenoid to adjust is ridiculous. This is one of the most simple machines out there. APD and Shearwater have decided to make them more reliable by removing as much analog transmitting as possible and replacing it with digital. Actual current going across long skinny wires in a bundle across connections is not reliable. 0/1's going down fewer thicker wires is perfectly OK regardless of how many connections because the computers can diagnose the signal and know if it is right or not.


Summary, if you don't trust the rebreather to fire a damned solenoid, you better be the same guys still diving tables and bottom timers and refusing to trust computers because if you aren't you are just being ignorant and hypocritical. It is literally that simple, and I apologize if that offends anyone, but no human can drive a rebreather as accurately as the computers can, it's not possible. I'm not aware of any recent incidents *since the Vision electronics and DiveCan's came out* where the computers were the cause of the accident. It was either O2 sensors, or the idiot diving it. The digital systems remove a little bit of the unreliability to the O2 sensors because now their monitored values can't be wrong. Their readings may be wrong, but the computer will always know exactly what the sensor is reading instead of running the risk of interference/voltage drop over connections etc. That alone is a huge step forward in their safety/reliability because the computers are now always aware of the ACTUAL readings at the cells. Again, those readings may not be accurate, but at least you can trust that the computers are always spitting out the actual values or no values at all. I'd rather have a computer just say "nope, sorry, can't get the signal, I have no idea what the hell is going on, bail out", then start to give me flaky numbers or wrong numbers.

I'm not a rebreather diver, but I do have a lot of experience with computer control algorithms...something you don't. I'll tell you right now that "fire if PO2 is low" isn't as simple as it is, and it's certainly not the whole story. So, firstly: How much firing? It takes a while for the gas to mix and get re-analyzed, so how much O2 do you dump in there? It should vary based on overall difference. So now you need a proportional decision setting, which aren't known for accuracy. How do you fix that? Integral and Differential control methods have to be incorporated, which means a full-on PID system that requires tuning and possibly even machine-learning. I don't know if CCRs have PID systems, but I would certainly hope so. Plus, there's the fact that the unit doesn't know what's about to happen. Imagine it trying to maintain its setpoint just before a large descent. There's NO WAY you could do that dive without "manually overriding" the system. What about it trying to maintain PO2 during ASCENT? No way it could do it. Units aren't good with prediction, they literally can't be. Running an eCCR in semi-Manual mode, to me, is the best of both worlds....you get the "safety net" of a low setpoint being maintained automagically in case of task loading (lost line drill, reel jam, placing a jump, etc) getting in the way of you running it right. Plus, the requirements to fool with it keep you more constantly monitoring it...where I know of a few eCCR divers that had a scary moment when they realized they had lost focus on the unit for too long.

I drive manual cars. I see a hill coming, I drop down a gear to prep for it. With driving, though, your only "task" is driving. If I had to drive, and tie in a reel, and write crap on my slate, and keep an eye on my buddy, and I had to be shifting constantly (which you obviously don't in a car)....I'd drive an automatic with a manual override. Shift for me if I screw up, but let me mostly do the decision making.

 

[tbone1004]

ahh, so point 2 is the important part, point 1 is less important in this instance. Running semi-manual for this types of situations is basically required, but that is a far cry from "I'm going to be at 300ft doing a working dive for half an hour, but now have to add the task loading of manually running the ccr". Semi-manual is OK because humans can predict, the units can't, but the units can maintain settings far more accurately than we can when everything is stable. So let the computer run while you're at one depth and working, and you can manually override for large ascents or descents. I don't think setting it low is the best option, I think setting it where you want it to be is the right answer as long as you can pay attention to it.

Point 1, was obviously over simplified in terms of the actual programming, I know this. I'm not a programming expert, you are. That said, the computer has a helluva lot better chance of being programmed properly than the divers thumb does. That also said, the PID's for O2 addition are very similar to fuel injectors in a car, something that has been done reliably for 20 years with OBDII, and in total for quite a bit longer. I still say there is no way that a human can maintain a setpoint like the computer firing a solenoid can, and thinking that a person can is just foolish. Units have manual O2 add buttons for a reason, they should be used, obviously for ascent. Descent being less of an issue since the ADV's take care of most of that.

Now, regarding your analogy, it is perfect. BUT where I think it should be is you drive in automatic, and you have ability to manually downshift if you're about to go down a hill or want to use engine braking, and ability to upshift if you're about to pass or need the extra HP to boost up a hill etc, but let the computer handle the rest of the shifting for regular use. Also have the ability to run in full manual for funzies if you want to and should know how to run it in full manual due to emergencies and to understand how the computer is doing what it is doing.


Regarding the HUD's. Below is the JJ HUD blinky chart. One LED per sensor, one blink in Orange is 1.0. Each flash in red is .1 below 1.0, each flash in green is .1 above 1.0. It's pretty easy to ignore these when they are all flashing the same sequence. It is very difficult to ignore if they are out of sequence with one sensor malfunctioning.

 

[kr2y5]

I'm not a rebreather diver, but I do have a lot of experience with computer control algorithms...something you don't. I'll tell you right now that "fire if PO2 is low" isn't as simple as it is, and it's certainly not the whole story. So, firstly: How much firing? It takes a while for the gas to mix and get re-analyzed, so how much O2 do you dump in there? It should vary based on overall difference. So now you need a proportional decision setting, which aren't known for accuracy. How do you fix that? Integral and Differential control methods have to be incorporated, which means a full-on PID system that requires tuning and possibly even machine-learning.

To be fair, from what you described so far, we still have not yet necessarily veered outside of the realm of simple counting... and when you refer to things like machine learning, I get a bit skeptical. The methods do sound disproportionately more sophisticated as compared to what the problem being solved appears to be. But, I most likely just do not understand enough about the mechanics of rebreather operation, and there is more to it than what has been described so far in this thread. I would be interested in better understanding where the complexity lies...

I don't know if CCRs have PID systems, but I would certainly hope so. Plus, there's the fact that the unit doesn't know what's about to happen. Imagine it trying to maintain its setpoint just before a large descent. There's NO WAY you could do that dive without "manually overriding" the system. What about it trying to maintain PO2 during ASCENT? No way it could do it.

Unclear why you think there is no way, could you please elaborate?

Units aren't good with prediction, they literally can't be. Running an eCCR in semi-Manual mode, to me, is the best of both worlds....you get the "safety net" of a low setpoint being maintained automagically in case of task loading (lost line drill, reel jam, placing a jump, etc) getting in the way of you running it right. Plus, the requirements to fool with it keep you more constantly monitoring it...where

What you say might very well be true, but those are some pretty strong statements you made about certain things being impossible, and I think it would be fair fto ask for a little more justification. When you say that a diver can predict something, but a rebreather can't, what scenario are you referring to?

I know of a few eCCR divers that had a scary moment when they realized they had lost focus on the unit for too long.

This is interesting, do you happen to know more detail, how the eCCR screwed up when the diver lost focus? Just curious about the limitations of existing systems. In what circumstances do they fail if left unsupervised, and how do they screw up? Do they fail to automatically maintain a setpoint on rapid depth changes, or in some other situation?

I drive manual cars. I see a hill coming, I drop down a gear to prep for it. With driving, though, your only "task" is driving. If I had to drive, and tie in a reel, and write crap on my slate, and keep an eye on my buddy, and I had to be shifting constantly (which you obviously don't in a car)....I'd drive an automatic with a manual override. Shift for me if I screw up, but let me mostly do the decision making.

What you are describing sounds like an analogy to a fully automated rebreather that pretty much runs "on autopilot", but with a manual override that lets you instantly take over... would you be comfortable with that throughout the dive, or only during the times when you are not making depth changes?

 

[Akimbo]

…Desired PO2=user input value. Actual PO2=mV*pressure…

I’m confused here. All the galvanic O2 sensors I have used read PPO2 directly. No pressure calculation was required unless the cell was on the surface and the gas was bled off the chamber or gas stream through a flow meter.

Oxygen monitors inside bells and chambers consisted of a Teledyne sensor, potentiometer for calibration, and an analog millivolt meter with a custom scale… nothing more than that beyond some wire. That is why inside sensors were considered more accurate once calibrated. A typical 0.3 ATA at 1000' would only read 1% when bled off and read on a sensor at the surface.

 

[The Chairman]

Measuring/trimming cells to give a useful figure like PPo2 is pretty simple electronics. I designed a few simple circuits last year that did just this so I could understand the process. I'm certain that there are any number of ways to crack the logic circuits but I'm no electronics wiz so I wouldn't even try.

As for making decisions, descending isn't the real problem. The automatic diluent add should buffer that out to be safe, if not ideal. Going up is the real problem and it's harder for a rebreather to keep the PPo2 in a safe zone during this time. I am always amazed by how the cruise control in my Sprinter keeps the desired speed pretty accurately in spite of going up or down steep hills. It anticipates and responds to those stressors pretty well. After all, that's what we're talking about here: cruise control for your rebreather.