Shearwater Perdix AI oddities: GF settings and dive planning

[stuartv]

@scubadada just prompted me to play with the Dive Planner on my new Perdix AI to see how it handles some things based on the GF settings. In the process of doing that, I noticed a few things that surprised me.

First, my computer was set to GF50/80 and a gas of EAN27 from my dives yesterday.

The first surprise was when I tried to set it to GF50/95. It wouldn't let me. It displayed this message:

"Lo>50. then Hi<90"

I guess that ">" and "<" really mean ">=" and "<=". It did let me set it to GF90/90. So, apparently, there is some limitation on what the GF Lo can be if the GF Hi is greater than 90.

Once set to GF90/90, I planned a dive to 100' (again, on EAN27). I experimented until I determined that the NDL for that depth and gas is 19 minutes. At 20 minutes, the resulting dive plan would include a 1 min stop at 20' (my Last Stop setting).

Then I changed it to GF25/90. The NDL remained at 19 minutes. And when I planned for 20 minutes, the dive plan was still for just a 1 minute stop at 20'. It did not surprise me that the GF Lo did not affect the NDL - i.e. it is not used unless the ascent will exceed GF Hi. But, it did surprise me that the plan for a 20 minute bottom time gave the same deco as with GF90/90.

Then I decided to compare it to Multi-Deco. I used the same depth (100') and gas (EAN27). I ran my plans in M-D using both Buhlmann ZHL-16B and 16C and the results were the same (between the two algorithms).

But, for the depth and gas, Multi-Deco gave an NDL of 17 minutes, instead of 19, for GF90/90. And when I changed it to GF25/90, the dive for 17 minutes resulted in deco stops of 1 minute each at 40, 30, and 20.

As I have done this kind of thing with M-D before, I was not surprised that it uses the GF Lo value even when GF Hi would not be exceeded. But, I am surprised that M-D and the Perdix Planner give such different results for the NDL (with GF90/90). 17 minutes versus 19 seems big, but I suppose it could essentially just be an issue of how rounding is done during the intermediate calculations.

I am really surprised, though, that when I planned for 20 minutes on the Perdix, with GF25/90, it still only gave the same ascent as with GF90/90 - i.e. 1 minute at 20'. So surprised, that I just ran it all again. I get the exact same results whether I use GF90/90 or GF25/90. The summary says Run: 24 minutes, Deco: 4 minutes, which is 20 minutes bottom time, a 3 minute ascent to 20', and then 1 minute at 20'.

That actually doesn't seem right. My gut says that the SW Planner is doing this wrong and Multi-Deco is doing it correctly. For a dive to 100' (on EAN27), an ascent using GF90/90 that gives 1 min deco at 20' should give considerably more, with deeper stops, when using GF25/90.

I'm going to email Shearwater and ask them about this. I'll post back when I hear something from them.

 

[Kevrumbo]

Run some more comparison deco profile simulations between Multi-Deco and Perdix Dive Planner GF's: try Eanx27 as before at 100', but with a longer bottom times (30, 40 and 50 minutes for example), and see how the deco/TTS compares -and also see if the expected DeepStop appears on the Perdix Dive Planner with GF25/90.

 

[guruboy]

From reading to manual, I got the impression that the in computer dive planner was really basic. Not much good for anything besides finding your no deco time

 

[clownfishsydney]

I ran the same simulations on my HW OSTC 2 and got the same results as you. I know that the algorythm in it is the same as Shearwater's as when my wife dives with me using her Predator, our no deco time is always almost identical and when we do deco dives, it is identical. Therefore it seems correct.

 

[stuartv]

I ran the Dive Simulation on my Seabear H3. It's different than the SW Dive Planner. It actually simulates a dive in 10x speed. You tell it the depth and time you want and it simulates an ascent or descent to that depth (at 33 ft/min ascent or 66 ft/min descent). You can add levels, to simulate a multi-level dive. The whole time it shows you what the computer would actually be showing during a real dive, just running at 10x speed. You can toggle through the dive screens, etc., just like you would be able to during an actual dive. It prompts for gas switches during ascent, and you can confirm them, just like during a real dive.

I set it for GF25/90 and did a dive to 100' for 20 mins. At the 17:00 mark, it goes into deco. At the 20:00 mark, it shows deco stops at 40, 30, and 20. Then I told it to ascend to 20', for 8 minutes. As it simulated the ascent, the 40 and 30 foot stops cleared just as it got to those depths, so no stops required. At 20', there was a stop, but it cleared within 1 minute.

I set it for GF90/90 and repeated. Same thing except it never showed required stops at 40 or 30. A direct ascent to 20' and the deco cleared within 1 minute.

I believe what the Shearwater is doing is it is actually figuring out that the required stops for GF25/90 would clear by the time you got there, so it's not showing them in the calculated plan.

It's cool that both computers agree when it comes to actually comparing the actual dive to be done.

But, I'm a bit surprised that the H3 shows that you're in deco after 17 minutes, when the SW gives 19 minutes of NDL. And I checked the planner on my Oceanic Atom. For the same depth and gas, it gives 26 minutes of NDL. I guess the H3's behavior is just another case of not being smart enough to ignore deco stops that would clear before you get to them. I suppose it's unfortunate because it does result in giving shorter NDLs for the same dive and GF settings, as compared to the Perdix. During the simulated dive, once I got to 100', I could add the elapsed dive time to the currently displayed NDL and it was always coming out to approximately 17:00, where I am pretty sure the Perdix would be displaying an NDL based on 19:00 of total allowed bottom time. Not a big deal, in the grand scheme. Too bad I'll never get a firmware update for the H3....

Anyway, I reckon I'm done experimenting for now.

 

[stuartv]

I got a response back from Shearwater. Here is what they said regarding my question on the GF settings limits and about the ascent plans for GF25/90 versus GF90/90.

Hi Stuart,

I hope I can help with your questions, and also, thank you for the suggestions.

• For planning, I attempted to set the GF to 50/95. It displayed the following message:

“Lo>50. then Hi<90”

…

I might want to use GF60/95 (just for example). Why are you not letting me?

Yes, you are right. It should be >= and <=.

This rule has been in the product for a long time. I’ve talked with the original designer, and there wasn’t a strong basis for its inclusion. The rule fit with the accepted thinking at the time.

We will remove this rule in a future firmware release. We are testing a release now, so unfortunately it won’t fit into the next release. However, the change will be in the next release after that.

• I planned a dive using these settings: Depth 100’, Gas 27/00, GF90/90, Last Stop 20’. It said my NDL was 19 minutes. So far, so good.
  …

Can you explain why the Perdix Planner gives the results that it does?

Desktop planners take the convention that once you reach the depth as set by GF_Low, a minimum stop time is inserted at each stop interval before evaluating whether that stop is clear. The minimum stop time is typically 1 minute. However, once the GF_Low depth is set, the conservatism relaxes and depth decreases, to GF_High at the surface. This relaxing conservatism often results in the first few stops being “instantly clear”. For a dive computer it doesn’t make to try to enforce some arbitrary stop duration when the algorithm says the stop is clear. Note that this is consistent with saying that only GF_High controls the NDL time.


If you have MultiDeco, try an experiment where you change the “Stop Times” setting down to 1 second (and set ascent rate to 33ft/min). When you do this, you will see these “instantly clear” stops change to 19 seconds. One second of this is the inserted minimum stop time. The other 18 seconds are just the time it takes to ascend to the next stop at 33 ft/min, which gets included in the stop time. These 19 seconds stops aren’t really stops at all.
...
Best regards,
Tyler Coen
Engineering Manager
Shearwater Research Inc.

@KenGordon, you may notice the part where this Engineering Manager from Shearwater says that only GF Hi controls the NDL time.

 

[KenGordon]

According to the H3 19 minutes is a deco dive whereas on the Shearwater it is a no stop dive. So the Shearwater implementation and the H3 implementation are different, and a bit different from the planner. Time to take them all in the water and see what happens after 17 minutes.

This illustrates a point I have made in the past when it has claimed that GF is some wonderful open standard.

He is talking about how when you arrive at a first stop the time taken to get there means that it may no longer be a stop In MultiDeco it inserts a stop into the plan according to the ceiling when you leave the bottom. It could do it according to the projected ceiling when you arrive at the ceiling but does not. You can test this by implementing a planner both ways and see which matches. A computer can be different as it only plans as an indication, the actual stops will be generated on the fly when you get to a depth so by default it behaves as though calculating the projected ceiling, ie it takes into account the off gassing on the ascent. If your ascent rate is slow enough then you will never reach the ceiling.Is a dive where you never reach the ceiling but there was a ceiling a no stop dive?

My software is in bits at the moment and I have taxes to sort out but I will try to get some numbers out which show the maximum allowed saturation. There are two ceilings, one from the 25% GF low and another from the 90% GF high. One happens sooner and is deeper than the other. If NDL is when reached when you have a ceiling then clearly the GF low one happens first.

Also you need to ask him a straight question - does the (NDL) planner ignore GF low? Or is he saying that often the GF low does not matter, in planning mode. Or that GF low does not make a difference to an actual dives no stop limit?

 

[KenGordon]

OK, by my calculations the NDL for 30m 27% on 25/90 is something a freediver might think they can try... That lines up with MultiDeco but I get a longer first stop so I want to look harder at why that might be.

 

[mafi]

you may notice the part where this Engineering Manager from Shearwater says that only GF Hi controls the NDL time.

This makes a perfect sense for - GF Hi controls the GF99 you surface with, i.e. limits the time you can spend in the water with the ability to ascend to the surface directly. If you set it lower, your NDL is shorter (lower allowed supersaturation)