How Will You Try to Kill Me?

Émile Jean-Horace Vernet-The Angel of Death

It’s been over three years since I posted a cautionary tale about oxygen sensors in rebreathers, and the calamities they can cause. Since then, the toll of divers lured to their death has been steadily mounting. In one week alone in April 2016, at almost the same geographical latitude in Northern Florida, there were two diving fatalities involving rebreathers. It is an alarming and continuing trend.

I know a highly experienced diver who starts each dive by looking at his diving equipment, his underwater life support system, and asking it that title question: How will you try to kill me today?

This deep cave diver, equally at home with open circuit scuba and electronic rebreathers, is not a bold cave diver. He is exceptionally cautious, because he is also the U.S. Navy’s diving accident investigator. He has promised me that his diving equipment will never end up in our accident equipment cage.

He and I have seen far too many of the diving follies where underwater life support systems fail their divers. But the crucible in which those fatal failures are often born are errors of commission or omission by the deceased.

Carelessness and an attitude of “it can’t happen to me” seem all too prevalent, even among the best trained divers. Divers are human, and humans make mistakes. Statistically, those accidents happen across all lines of experience: from novice divers, to experienced professional and governmental divers, and even military divers. They all make mistakes that can, and often do, prove fatal.

It is exceedingly rare that a life support system fails all by itself, since by design they are robust, and have either simple, fool-proof designs, or redundancy. In theory a single failure should not bring a diver to his end.
The “head”, triplicate oxygen sensors, oxygen solenoid and wiring leading to the rebreather CPU. Image from


Are oxygen sensors trying to kill you? That depends on how old they are? Are they in date? Ignoring the expiration date on chocolate chip cookies probably won’t kill you, but ignoring the expiration date on oxygen sensors may well prove fatal. Complex systems like rebreathers depend upon critical subsystems that cannot be neglected without placing the diver at risk.

Oxygen sensors are usually found in triplicate, but if one or more are going bad during a dive, the diver and the rebreather can receive false warnings of oxygen content in the gas being breathed.  We have seen a rebreather computer “black box” record two sensor failures, and it’s CPU logic deduced that the single working sensor was the one in error.

The controller’s programmed logic forced it to ignore the good sensor, and thus the controller continued to open the oxygen solenoid and add oxygen in an attempt to make the two dying sensors read an appropriately high O2. Eventually, the diver, ignoring or not understanding various alarms he was being given, went unconscious due to an oxygen-induced seizure. His oxygen level was too high, not too low.

Unlike fuel for a car or airplane, you can have too much oxygen.

Oxygen sensors do not fail high, but they do fail low, due to age. Rebreather manufacturers should add that fact into their decision logic tree before triggering inaccurate alarms. But ultimately, it’s the diver’s responsibility to examine his own oxygen sensor readings and figure out what is happening. The analytical capability of the human brain should far exceed the capability of the rebreather CPU, at least for the foreseeable future.

JAKSA high pressure 6-volt solenoid used in a Megalodon rebreather. NEDU photo.

Oxygen addition solenoids hold back the flow of oxygen from a rebreather oxygen bottle until a voltage pulse from the rebreather controller signals it to open momentarily. The oxygen flow path is normally kept closed by a spring inside the solenoid, holding a plunger down against its seat.

But solenoids can fail on occasion, which means they will not provide life giving oxygen to the diver. The diver must then either manually add oxygen using an addition valve, or switch to bailout gas appropriate for the depth.

Cut-away diagram of a 24-volt Jaksa 200 bar solenoid.

Through either accident or design, divers have been known to invert their solenoid spring and plunger, thereby keeping the gas flow open. In that case, oxygen could not be controlled except by manually turning on and off the valve to the oxygen tank. Of course, knowing when oxygen is too low or too high would depend upon readings from the oxygen sensors.

Suffice it to say that such action would be extremely reckless. And if the oxygen sensors were old, and thus reading lower than the true oxygen partial pressure, the diver would be setting himself up for a fatal oxygen seizure. It has happened.

Assuming a solenoid has not been tampered with, alarms should warn the diver that either the solenoid has failed, or that the partial pressure of oxygen is dropping below tolerance limits.

But as the following figures reveal, if the diver does not react quickly enough to add oxygen manually, or switch to bail out gas, they might not make it to the surface.

The three figures below are screen captures from U.S. Navy software written by this author, that models various types of underwater breathing apparatus, rebreathers and scuba. In the setup of the model, an electronically controlled, constant PO2 rebreather is selected. In the next screen various rebreather parameters are selected, and in this case we model a very small oxygen bottle, simulating an oxygen solenoid failure during a dive. On another screen, a 60 feet sea water for 60 minutes dive is planned, with the swimming diver’s average oxygen consumption rate set at 1.5 standard liters per minute.

Screen shot 3Screen shot 2










On the large screen shot below, we see a black line representing diver depth as a function of time (increasing from the dashed grey line marked 0, to 60 fsw), a gray band of diver mouth pressure, and an all-important blue line showing the partial pressure of inspired oxygen as it initially increases as the diver descends, then overshoots, and finally settles off at the predetermined control level of oxygen partial pressure (in this case 1.3 atmospheres). Broken lines on the very bottom of the graph show automated activation of diluent add valve, oxygen add solenoid, and over pressure relief valve. Long horizontal colored dashes show critical levels of oxygen partial pressure, normal oxygen level (cyan) and the limit of consciousness (red).

Screen shot 1
Screen shot of UBASim results after an ill-fated 60 fsw dive.

The oxygen solenoid fails 53.7 minutes into the dive, no longer adding oxygen. Therefore the diver’s inhaled oxygen level begins to drop. Rather than follow the emergency procedures, or perhaps being oblivious to the emergency, this simulated diver begins an ascent. As ambient pressure drops during the ascent, the drop in oxygen pressure increases.

In this particular example, 62.5 minutes after the dive began, and at a depth of 13.5 feet, the diver loses consciousness. With the loss of consciousness, the diver’s survival depends on many variables; whether he’s wearing a full face mask, whether he sinks or continues to ascend, or is rescued immediately by an attentive boat crew or buddy diver. It’s a crap shoot.

So basically, the rebreather tried to kill the diver, but he would only die if he ignored repeated warnings and neglected emergency procedures.

What about your rebreather’s carbon dioxide scrubber canister? Do you know what the canister duration will be in cold water at high work rates? Do you really know, or are you and the manufacturer guessing? What about the effect of depth, or helium or trimix gas mixes? Where is the data upon which you are betting your life, and how was it acquired?

Scrubber canister and sodalime. NEDU Photo
NEDU photo.












Sadly, few rebreathers have dependable and well calibrated carbon dioxide sensors; which is unfortunate because a depleted or “broken through” scrubber canister can kill you just as dead as a lack of oxygen. The only difference is a matter of speed; carbon dioxide will knock you out relatively slowly, compared to a lack of oxygen.

But if you think coming up from a dive with a headache is normal, then maybe you should rethink that. It could be that your rebreather is trying to kill you.





Computer Simulation as Art — or Rorschach Test

No one has ever confused me for an artist.

I might have been visually gifted as a 3rd-grader, as my parents told it, at least compared to my peers. However, I never seemed to progress beyond that point. I think my progress slowed about the time I saw my first Rorschach test.

I realized then that some people’s art is someone else’s diagnosis. After all, it is no fun to look at an ink blot abstraction, to voice an opinion about it, only to have an authority figure nod his head and write in his notebook as he says, “I see,” when obviously he didn’t.

Clinical trauma aside, I now know that all humanity looks instinctively for visual patterns and searches for meaning in patterns whether they be random or not. There is a survival aspect to that of course; if we detect a tiger’s stripes partly hidden in a confused background of woodland scenery, that offers a potential survival benefit.

Sometimes, even the most mundane things turn out to be “pretty”. Such were the images I saw being formed on my computer screen the other day. The more I looked at them, the more interesting they became. They were like my own Rorschach test, in a very literal way. They were random patterns based on random processes, but my brain refused to look at them that way. They appeared to me as images of natural things, representing anything except what they truly were.

The image to the left, for instance, looked to me like a view through a telescope of a star field with at least one galaxy situated near the center axis.

Or in a very biological way, it might be the view through an immunofluorescence microscope.

The next image looked to me like a view of a placid star seen in ultraviolet light. I could almost feel the blistering heat radiating through space.

Alternatively, it might be a view of a human egg waiting patiently for fertilization, an altogether different interpretation, but like the first, being a necessary component of creation.

The final image looked to me like a cooler star but with clearly visible solar prominences, magnetic storms arcing over the hellish nuclear surface.

I have no idea what others might see in these images, if anything, but I’m guessing each image can be interpreted differently based on one’s own life experiences.

And that after all is the whole point of art, and Rorschach tests.



The above images were created as part of a random, or stochastic, simulation of rebreather scrubber canisters. They are a view of the upstream end of an axial canister, and shows the state of the canister as heat producing carbon dioxide absorption reactions are beginning.

The cooler looking the canister, the less the amount of exhaled carbon dioxide entering the canister.

The simulation tracks chemical reactions and heat and mass transfer processes in an array of 272,000 finite elements making up a simple absorbent canister. Slicer Dicer and 3VO software (PIXOTEC, LLC) were used to visualize the three-dimensional data set acquired during one moment in time shortly after the simulated reactions began.



Another Rebreather Scrubber Thermokinetic Simulation

Compared to the previously posted video of a segment of a rebreather scrubber, this video shows a much larger, and therefore more realistic scrubber with axially aligned, CO2 rich gas flow passing from left to right. Due to the larger size of the simulation space, more widely distributed heat patterns are noticeable, as are fluctuations in heat. The flow of those fluctuations are most noticeable along the simulated boundary of the cylindrical scrubber bed.

The assumptions of this simulation are that CO2 production (diver workload) is constant throughout the simulation run, ventilatory flow through the canister is constant, the surrounding water temperature is constant at 50° F, and the canister was chilled to the water temperature before the “diver” started breathing through it.

The previous simulation conditions were similar except that the canister was toasty warm prior to immersion in frigid water.

To fully appreciate the fine detail of the imagery, click on the video frame then expand the video to full screen size (lower right symbol immediately after “You Tube”) and play back in 1080p High Definition mode.





A Look Inside Rebreather Scrubber Canisters, Part 2

Computer modeling allows you to see things that are invisible in real life.

The previous posting showed the complex thermal profiles generated in a rebreather canister found in closed-circuit underwater breathing apparatus during the CO2 absorption process. But heat generation is just part of the absorption process. Simulation allows you to see how the end product of CO2 absorption, calcium carbonate, gets deposited inside the canister.

To the right is calcite, a form of calcium carbonate. Divers never see crystals of calcite in the scrubber canister because sodalime granules are never completely converted to calcite. Typically, no more than 50% of the granules react completely with exhaled CO2.

The following images show the interior of a a scrubber canister as the sodalime granules begin reacting with exhaled CO2. When sodalime granules first begin to absorb CO2 the image becomes purple. With more CO2 the color turns reddish, and when all binding sites are filled with reacted CO2, the granule color becomes yellow.  

The more carbonate in a particular location in the granule bed, the more yellow the image.

The probability that an exothermic absorption reaction would occur is dependent on the granule temperature, the granule size, the number of granules and the number of sites available for reaction in each granule.

In the second image, CO2 absorption sites in the inlet to the canister were completely filled (thus showing yellow), and small pockets of absorption were extending up the canister walls.

When I saw the third computer-generated image, I was surprised. It showed that in the central portion of the absorbent bed, the moving thermal front seen in the previous post was leaving behind a calcited bed. However, sheets of calcium carbonate were forming on the outer surface of the canister, the coldest portion of the canister.

Initially that result was counter-intuitive. Then I realized that low temperature makes the odds very low that the first granule encountered would absorb CO2. All chemical reaction rates are temperature dependent, therefore exhaled CO2 would be very likely to proceed downstream to the next granule. There again the odds of being absorbed would be low so the CO2 molecule would continue downstream.

However, given enough opportunities, even low probability events eventually occur. That means that along the cold canister walls, carbonate begins to be deposited much further downstream than in the warmest, and most highly reactive portion of the bed.

Unfortunately, the low probability of CO2 absorption in cold granules means that CO2 hugging the cold canister walls is likely to pass completely through the canister, unabsorbed. Chances are also high that the same molecule would be shunted to a different portion of the canister on its second pass through the canister, and therefore would eventually be reabsorbed.

The following link is to a high definition video showing carbonate deposition in a cylindrical scrubber canister as the simulated diver plunges into icy water. For best effect go to full screen and 1080p mode.



Further details about the computer simulation involved in the production of these images and video can be found in the paper “Computer Modeling of the Kinetics of CO2 Absorption in Rebreather Scrubber Canisters”, in MTS/IEEE OCEANS 2001 Conference Proceedings, published by the Marine Technology Society; Institute of Electrical and Electronics Engineers; Oceanic Engineering Society (U.S.); IEEE Xplore (Online service).