Shearwater Perdix AI

[Lorenzoid]

. . .

Imagine: A rec diver gets a new Perdix AI. They are about to do a dive on a reef with a 60' hard bottom. Their NDL is going to be so long that they know they'll have to end the dive because of low gas. Because it's fairly shallow, they decide they are okay with getting back to the boat with 200 psi left. I'm sure I'm not the only one that has seen numerous divers get back on the boat with that much gas left. So, this diver sets the PAI reserve to 200 and jumps in. When his computer hits 0 on the GTR meter, he begins his ascent, following the computer exactly, doing an exact 30 ft/min ascent. He gets to 20' and the computer tells him to do a 3 minute safety stop. Now, because of the way the PAI works, during his safety stop he actually breathes his tank down past 200 psi and suddenly realizes he's still at 20' depth and has just breathed his tank down so low that he can no longer draw a breath. There is clearly a bad choice being made right off the bat, in this scenario. But, the Shearwater implementation for GTR also clearly contributes significantly to the diver going OOA.

Tee hee. I am imagining a conversation a year or so ago in a Shearwater R&D meeting, in which they debated whether to produce an AI Perdix. It wouldn't surprise me if someone mentioned the can of worms it might open, given Shearwater's implicit "you're a big boy, you know what to do" mindset evidenced by user-settable gradient factors and absence of lock-out. So now with AI they need to coddle inexperienced/reckless divers?

Nevertheless, I do agree with your suggestion of how they should modify the GTR feature. If they're going to have such a feature, it clearly should factor in gas consumed during stops.

 

[drrich2]

The 'how often do you check your gas pressure' issue comes up pretty often in A.I. discussions, with the recurrent theme that over time divers tend to settle in and not check so often. Maybe so. I'll offer a minority contrarian view.

On one of my OW dives, in fairly cold water with no hood or gloves, excited, long drive to get there, etc..., I ran out of air at depth. I signaled 'out of air' to my instructor, he was right there and noticed immediately, and we went ahead and did the CESA exercise that was on the agenda anyway. He was good-natured about it, but that episode drilled into my head 'check your gas, check it often.'

I believe this mindset is encouraged in new divers.

Personality/mindset is an issue, too. I'm an absent minded daydreaming introvert. If I'm not directly engaged with the external environment, I soon 'drop out' off and on. On the very rare occasions I need to be on a roof top for some reason (e.g.: clearing debris), I remain fearful the whole time...because there's a power line nearby, and it'd be like me to 'tune out' and blunder into it.

On a deep dive, excited about the reef, snapping pictures, wouldn't be hard to run low. So I check. Often. For someone like me, the convenience of 'at a glance' date confirmation where I'm looking for my depth, time & NDL anyway is nicer than having to unclip an SPG as described by some.

My point is, even if many divers only check their gas supply a few times on a dive, some check more.

Richard.

 

[Kevrumbo]

None of the reasons for my prospective purchase that I posted above are "problems" as such. They are fairly minor items, which when added together, will , I believe, make for a more enjoyable dive for me. I can see the extra dimension added by AI as being of more help to me post dive when looking at my profile and analysing where my consumption gets better or worse. As I say, I am at the point with my diving where I am comfortable doing different types of dives but as to working out a relatively accurate SAC, that would prove difficult as my diving has been so varied.
Examples below (SAC provided via subsurface)
Dives 1-4 OW, depths 9-10m, SAC ranges from 38-24l/m
Dives 5-8 - weekend trip so fun dives but varied shore dive profiles as noted in my post above , depths 8-19m - SAC range 42-24l/m
Dives 9-15 - AOW including PPB (split in two 20 minute dives due to buddies equipment issues) + 1 aborted dive due to poor vis, depths 5-24m - SAC range 41-25l/m.
Dives 16-20 weekend trip - varied sites & currents, depths 10-18m SAC range 32-23l/m
Dive 21-40 various dives on Red Sea LOB - reefs/ & wrecks with night dives, surge, currents etc, depths 10-31m. SAC range 28-23l/m

So for on the fly calcs, I use something between 42 and 23L/m for working out possible duration/consumption. Or I could get an idea from the on the fly SAC that the computer gets me - not so much to ride it to the minute but as a guide. For example the plan for our shore dives is to return with 50 bar in about 40 minutes (doing a roughly square route) so approx 10 minute legs. If I see my SAC climb due to current etc, we can shorten the legs/ cut one off if required(it is fairly common to come back with 100 BAR+)

I can appreciate experienced divers will know what their SAC will be give or take a minor variation. Once I get a few hundred dives in, I will have a body of data that I can draw statistically significant data from - at the moment I have a set of disparate dives where the data is inconsistent due to various circumstances. Drawing conclusions from that data is risky at best.

I think it is similar to a lot of things - do we need computers with integrated compasses? No they are a gimmick - there is nothing wrong with a traditional magnetic compass. Nice to have though? Yes.

@stuartv - the diving world can't even standardise SAC/RMV or whatever else you want to call it (as highlighted in the many threads where it is referred to) so what chance of standardising anything else.

Try this simple exercise on a dive to 18 meters with an AI or SPG:

Use 22 liters/min as a SAC/RMV reference baseline.

Select a dive cylinder: Steel100 tank (12L ); Alu80 tank (11L), etc

Using an Alu80 for example, divide the RMV by this tank's rating factor (11 liters/bar):
22 liters/min divided-by 11 liters/bar equals 2 bar/min per ATA

Surface (1ATA) = 2BAR/min.
10M (2ATA) = 4BAR/min.
20M (3ATA) = 6BAR/min.
30M (4ATA) = 8BAR/min.

Do a nominal dive to 18 meters (2.8 ATA), and you should be around 6bar/min by the above table. In 5 minute time intervals at 6bar/min, you should be consuming 30 bar as a reference baseline for 18m constant depth. In 10 minute time intervals you would use 60 bar as reference, and so on.

Starting with an initial pressure at depth like 190bar for example (less full tank because of BCD inflation, drysuit inflation, momentary exertion getting settled at depth etc), check your SPG or AI every 5 minutes and see if it decreases by 30bar. So after five minutes, your 190bar is 30bar less and you have 160bar remaining; another five minutes and you have 130bar remaining; another five minutes 100bar remaining and so on.

It may be you're so relaxed at 18m depth that only 20 bar per five minute interval is consumed -great! Just keep on using 30 bar delta as reference. Where you have to be concerned is when you start seeing consumed delta values like 35 bar, which means you're working hard with increasing breathing rate or less likely you have a small leak somewhere. Anything like 40bar and above and you already know you're in full stress exertion with CO2 narcosis and a physical breathing rate that's unsustainable (like kicking into a strong current for example) or you have a major leak bubbling.

The point is in order to get a feeling for your pressure consumption rate based on your RMV and tank size, don't focus so much on the dimension of Air Time Remaining in an AI feature, but rather in terms of consumed pressure per five minute time intervals and remaining tank pressure as displayed by the AI for the current depth. When you get good at this, especially in the Metric System, you can do it for multi-level depths as well. . .

 

[Diving Dubai]

But how frequently do I need to check my pressure? Depending on the dive, I may not look at my SPG very often at all. It's there when I need to look at it, clipped to my BC, and I have developed enough muscle memory that I can unclip it, look at, and clip it off again in just a few seconds. I am admittedly no @Kevrumbo when it comes to mental arithmetic, but I leave a large safety buffer of gas, and I have a rough idea based on the depth and time--as confirmed by checking my SPG--when I need to start my ascent. Maybe that's just something developed by doing the same kind of dive many times. Is the marginal cost of an AI computer and transmitter, the inconvenience of batteries, more numbers to read on the display, etc., actually more convenient or less convenient? To you, I guess it is more convenient. Not to me--at least not for the way I presently dive. As I said, we're all different. I try to limit my comments to me.

Let me give you an example. Albeit on a dive that you freely admit is not on you'd do. - and I'm just usign yoru post as an expample

Last year I was on a site, we dive regularly. A pinnacle about 15 mins off shore. When we jumped there were no significant surface currents (the site needs a bit of flow)

My wife and I were happily mooching around at around 30m and had been for some time - we were within out NDL and rock bottom gas. All happy an then out of nowhere we were hit by a vicious down current. We stabilised (grabbed on) at 53m - Although we were on 28% we both swiched to our ponies that had 21% (mine a 40 hers a 30)

I have a 15l (Hp 120 tank), my wife as she breaths less a 12l. On a normal dive our fills are 230bar and we surface in the 90-110 range. We like the extra reserve. This dive was going to test this.

I'm not great at mental arithmetic, in that I have to visualise the numbers written on paper. But normally I know roughly what my contents will be at any point in the dive and can adjust my profile to suit (perhaps had to swim into current)

Having been blown down 20m, we're a little stressed. not in a panic, but not calm. Swimming out isn't looking great as the sea bead is another 50 - 60m below and no idea how big this current is, so we climb. By now our Deco time is clocking up

We both have have an elevated SAC and are both using the data of Time to Surface/Gas time remaining rather than calc it though a narced and stressed brain.

Long story short, we get out of the current having climbed a significant distance (or so it felt) shredding our gloves , bending or metal sticks and leaking a bit of blood (and few mask flood to boot)

From the info displayed we know we can honour our deco obligations and safety stops etc. It was close. We both dug deep into our reserves (checking back I had 15 bar in my back gas and 25 bar in my pony) We were both breathing like steam trains and (it felt at least) that we were screaming through gas

Having the data on my wrist, was a comfort. It told me we had enough gas - and I wasn't' trying to wrestle with the mental calcs when my brain was pretty much concentrating on getting us out.

This is 1 of only 2 times I've relied on my GTR
Our much malinged Suunto RGBM algorithm didn't get us bent, nor give us endless deco
If we followed the general SB wisdom of only carrying AL19 ponys - we would have been surfacing a lot faster

99% of my dives, I don't need the functions of my AI, nor a AL40 pony. But for that 1% I'm sure as heck happy to have it.

 

[stuartv]

Agreed. I suppose I picked up the term "riding the NDL" right here on SB, but I didn't mean it to carry any negative connotation it might have. By "riding," I simply meant following the computer's NDL display (or for that matter, the SPG's reading) to make an on-the-fly decision of when to ascend, instead of any other advance planning. Setting your GF and gas reserve that way, knowing at least roughly how that is going to relate to your bottom time at the planned depth(s), is a form of advance planning.

However, with such conservative setpoints the computer isn't providing that much of a benefit, is it? I mean, if you're cutting it that roughly, you could simply begin your ascent based on estimates. If I'm at X depth with Y PSI remaining, I have a rough idea whether I need to ascend some in order to have enough gas remaining should I need to do an air-sharing ascent with my buddy. If I'm conservative, I don't need a computer to calculate that for me. And I do often surface with 1000 psi or more.

Right. I wasn't accusing you of saying it in the pejorative sense.

And I wasn't proposing those settings as specific ones for people to use. 30/70 and 1000psi was just an example to illustrate that one could "ride the NDL and GTR" and still be diving very conservatively. Riding the computer is not inherently bad. Riding the computer with settings that are too aggressive is what people (especially the less experienced divers) should probably avoid.

Tee hee. I am imagining a conversation a year or so ago in a Shearwater R&D meeting, in which they debated whether to produce an AI Perdix. It wouldn't surprise me if someone mentioned the can of worms it might open, given Shearwater's implicit "you're a big boy, you know what to do" mindset evidenced by user-settable gradient factors and absence of lock-out. So now with AI they need to coddle inexperienced/reckless divers?

Nevertheless, I do agree with your suggestion of how they should modify the GTR feature. If they're going to have such a feature, it clearly should factor in gas consumed during stops.

I'm sure it was discussed a lot. But, I don't think I'm suggesting that anyone needs to be coddled. I'm just suggesting that all of the computer's predictions be based on following everything the computer tells you. You set GF to 30/70 and the computer predicts that if you do the ascent exactly as recommended by the computer, you will hit the surface with a GF99 of 70 or less. Similarly, you set the gas reserve to 500 psi and do the ascent exactly as recommended by the computer, it SHOULD mean that you hit the surface with 500 psi. If the computer recommends a 3 minute safety stop and you do it and then you hit the surface with less than 500 psi, then, in my mind, either the computer predicted incorrectly or your breathing changed. If your breathing didn't change, then the computer was "wrong" (in my mind).

We seem to be totally in agreement on this.

I am really looking forward to seeing how Shearwater does with updating the PAI going forward and if/when they improve the gas management functionality.

I also can't wait to see next year's wristwatch-sized Peregrine model with AI. ;-)

 

[Joneill]

Right. I wasn't accusing you of saying it in the pejorative sense.

And I wasn't proposing those settings as specific ones for people to use. 30/70 and 1000psi was just an example to illustrate that one could "ride the NDL and GTR" and still be diving very conservatively. Riding the computer is not inherently bad. Riding the computer with settings that are too aggressive is what people (especially the less experienced divers) should probably avoid.



I'm sure it was discussed a lot. But, I don't think I'm suggesting that anyone needs to be coddled. I'm just suggesting that all of the computer's predictions be based on following everything the computer tells you. You set GF to 30/70 and the computer predicts that if you do the ascent exactly as recommended by the computer, you will hit the surface with a GF99 of 70 or less. Similarly, you set the gas reserve to 500 psi and do the ascent exactly as recommended by the computer, it SHOULD mean that you hit the surface with 500 psi. If the computer recommends a 3 minute safety stop and you do it and then you hit the surface with less than 500 psi, then, in my mind, either the computer predicted incorrectly or your breathing changed. If your breathing didn't change, then the computer was "wrong" (in my mind).

We seem to be totally in agreement on this.

I am really looking forward to seeing how Shearwater does with updating the PAI going forward and if/when they improve the gas management functionality.

I also can't wait to see next year's wristwatch-sized Peregrine model with AI. ;-)

Agreed - I get why they want to limit GTR to single tank - but it seems like the stops (a safety and deco) could pretty easily be included into the GTR and make it more accurate? As they indicate that part of their rationale was that not all dives require a safety stop, they could just add menu items to include safety stop and/or deco stops in GTR calculations - so user could decide on their own? Instead they seem to suggest to account for this in your reserve setting - that can work for Safety Stops which are predictable - but not for deco stops.

From Perdix AI manual:

 

[dmaziuk]

can you do arithmetic involving multiples of 1bar/min, or at most 2bar/min without AI, or even a calculator app on pre-dive gas planning? Sure you can! All you need is a SPG to confirm what you already know. . .

No, all you need is a tank-pressure-reading instrument to confirm what you already know. Whether that instrument is a bourdon tube on the end of a hose or a newfangled sensor sending 'licktrickery pulses to a shiny oled screen on your wrist is entirely orthogonal.

Personally I'd like to get rid of the hose. I don't give a hoot about GTR or an occasional loss of link because as you say: I should already know what's there anyway. I don't need a new computer now and don't have an unclaimed grand laying around begging to be dropped on a new shiny, but when I do need it and I can budget for it, getting rid of the hose will be at the top of my requirements list. Hopefully there will be OSTC AI by then: I get paid by writing open source code and I'd rather give my money to others who do.

 

[AJ]

By that logic, the SPG is a gimmick too... not all of us are claiming that AI solves a problem - however, it certainly provides a more convenient way to view tank pressure versus the SPG as well as additional and very useful automatic data logging (and GTR as well).

Yes, an SPG is a gimmick too. Agree on that, It's all about awareness, it's about instinctively guestimating how much gas you and your teammates have left without having to be warned by some instrument. It's a sense that can be developed by training and experience, not by trusting technical solutions to do it for you.

By the way, I have seen divers getting into trouble with AI becuase they ignored alarms. They did not have time read numbers on their computer because they were to occupied with other things too much to bother to read their computer (high taskload). So there it is: is AI a soultion? Not to me if you're not aware about your situation. Is AI a usefull addition to your instruments? Yes it can be, just like a SPG. But in the end it's just a tool not the solution to be safe as some claim it to be. It's about how much you're relying on your instruments or on your gut feeling. I know I can trust my gut feeling, but verify my feeling through my instruments.

Is this hard to learn? Not really, I did it within my first 150 dives. I used AI as training wheels for the first 40 or so dives, then switched to SPG. But I only used GTR and such for the first 15 dives because it was off all the time and giving me the scares by it's nonsense warnings. Learned quickly to do the maths myself and trust on them.

 

[stuartv]

Agreed - I get why they want to limit GTR to single tank - but it seems like the stops (a safety and deco) could pretty easily be included into the GTR and make it more accurate? As they indicate that part of their rationale was that not all dives require a safety stop, they could just add menu items to include safety stop and/or deco stops in GTR calculations - so user could decide on their own? Instead they seem to suggest to account for this in your reserve setting - that can work for Safety Stops which are predictable - but not for deco stops.

From Perdix AI manual:

Right. I agree with you. I think their rationale is, well, less helpful than it could be.

If not all dives require a safety stop, so what? If the computer is going to prompt to do a stop (whether it's a mandatory deco stop or an optional safety stop), that time should be included in the GTR. If it's not, then don't include the time. Seems pretty simple to me. If the user sets the computer to not prompt for a safety stop, then the SS time would not be factored into the GTR.

The ONLY issue I could see is if. say, you have the computer set so that it prompts for a safety stop dependent on factors like dive time or depth, and maybe the safety stop time, when prompted, varies. Say, it prompts for 3 minutes if your dive was 80' or less and 5 minutes if your dive was more than 80'. Okay, so what? Your GTR shows a number, then you drop below 80' and the number drops abruptly because you crossed that threshold so your consumption rate "at current depth" just went up, plus your ascent time just immediately jumped because of changing the plan from a 3 minute stop to a 5 minute stop (with the corresponding reduction in GTR). Big deal. It's better (in my opinion) for the number to be as accurate as possible and the user to have to read the manual to understand why it behaved in a certain way than for the number to be inaccurate (and result in your arrival at the surface with less gas than your reserve setting) but the number "behaves" (i.e. doesn't make any abrupt jumps in value).

It seems to me like any concerns about being confusing because the number jumps abruptly could be addressed by having the computer have default settings (for the safety stop) that prevent abrupt jumps from happening. E.g. it could come from the factory set to always prompt for a 3 minute safety stop. Or even better, always prompt for 5 minutes. If the user wants a shorter stop or a stop time that varies based on the dive, they can change the setting themselves and at least be aware that they did it and maybe understand why the GTR now drops abruptly when they drop below 80' or whatever. Or, the user can just leave it set for 5 minutes and ascend after 3 minutes, if they want.

 

[Kevrumbo]

Let me give you an example. Albeit on a dive that you freely admit is not on you'd do. - and I'm just usign yoru post as an expample

Last year I was on a site, we dive regularly. A pinnacle about 15 mins off shore. When we jumped there were no significant surface currents (the site needs a bit of flow)

My wife and I were happily mooching around at around 30m and had been for some time - we were within out NDL and rock bottom gas. All happy an then out of nowhere we were hit by a vicious down current. We stabilised (grabbed on) at 53m - Although we were on 28% we both swiched to our ponies that had 21% (mine a 40 hers a 30)

I have a 15l (Hp 120 tank), my wife as she breaths less a 12l. On a normal dive our fills are 230bar and we surface in the 90-110 range. We like the extra reserve. This dive was going to test this.

I'm not great at mental arithmetic, in that I have to visualise the numbers written on paper. But normally I know roughly what my contents will be at any point in the dive and can adjust my profile to suit (perhaps had to swim into current)

Having been blown down 20m, we're a little stressed. not in a panic, but not calm. Swimming out isn't looking great as the sea bead is another 50 - 60m below and no idea how big this current is, so we climb. By now our Deco time is clocking up

We both have have an elevated SAC and are both using the data of Time to Surface/Gas time remaining rather than calc it though a narced and stressed brain.

Long story short, we get out of the current having climbed a significant distance (or so it felt) shredding our gloves , bending or metal sticks and leaking a bit of blood (and few mask flood to boot)

From the info displayed we know we can honour our deco obligations and safety stops etc. It was close. We both dug deep into our reserves (checking back I had 15 bar in my back gas and 25 bar in my pony) We were both breathing like steam trains and (it felt at least) that we were screaming through gas

Having the data on my wrist, was a comfort. It told me we had enough gas - and I wasn't' trying to wrestle with the mental calcs when my brain was pretty much concentrating on getting us out.

This is 1 of only 2 times I've relied on my GTR
Our much malinged Suunto RGBM algorithm didn't get us bent, nor give us endless deco
If we followed the general SB wisdom of only carrying AL19 ponys - we would have been surfacing a lot faster

99% of my dives, I don't need the functions of my AI, nor a AL40 pony. But for that 1% I'm sure as heck happy to have it.

A very harrowing experience with thankfully a fortunate outcome given that surprise down current (the Nitrox Diver's nightmare) . . .you both were prepared and handled it well. AI/GTR gave you peace of mind in that instant. . .

No, all you need is a tank-pressure-reading instrument to confirm what you already know. Whether that instrument is a bourdon tube on the end of a hose or a newfangled sensor sending 'licktrickery pulses to a shiny oled screen on your wrist is entirely orthogonal.

Personally I'd like to get rid of the hose. I don't give a hoot about GTR or an occasional loss of link because as you say: I should already know what's there anyway. I don't need a new computer now and don't have an unclaimed grand laying around begging to be dropped on a new shiny, but when I do need it and I can budget for it, getting rid of the hose will be at the top of my requirements list. Hopefully there will be OSTC AI by then: I get paid by writing open source code and I'd rather give my money to others who do.

I don't have to wait to synch/pair an analog SPG manometer, or experience the frustration any momentary transmitter signal interruption at an inopportune moment (what if it happened with @Diving Dubai above?), nor deal with batteries and another potential O-ring failure point flood of a battery port.