SWCloud and Subsurface differences

[Jay]

I know there can be some minor implementation differences when ZHL-16C is implemented.

Many of my dives exhibit large NDL and SurfGF differences, some are within a minute and a few%, but most NDLs are 5-10mins different, and SurfGF are ~20% different. These two values are easy to see in both SWC and SubS (Current GF / GF99 is a little harder to compare).

I was comparing dives where NDLs were meaningful (deeper dives, just prior to ascent generally).

Anyone else see differences?

(in SubS if you dove with different GFHis, you might need to manually alter it, this is for Rec diving (so we're not talking about anchoring GFLow differences))

EDIT - to add: I couldn't get a direct connection between my dive computer and the computer so I exported from SWCloud using UDDF. Everything looks correct.

 

[rsingler]

@atdotde ?
@Shearwater Cloud ?

 

[atdotde]

Could you please post one of your dives as an example?

Honestly, I would not lose too much sleep over differences of those values if all these dives are non-stop dives. Algorithms should be in the same ballpark for the stops they compute but the "amount of non-stopness remaining" is somewhat arbitrary and depends quite a bit on the assumed ascent velocity and if that assumes some sort of safety stop etc.

 

[FreeFlyFreak]

I know there can be some minor implementation differences when ZHL-16C is implemented.

Many of my dives exhibit large NDL and SurfGF differences, some are within a minute and a few%, but most NDLs are 5-10mins different, and SurfGF are ~20% different. These two values are easy to see in both SWC and SubS (Current GF / GF99 is a little harder to compare).

I was comparing dives where NDLs were meaningful (deeper dives, just prior to ascent generally).

Anyone else see differences?

(in SubS if you dove with different GFHis, you might need to manually alter it, this is for Rec diving (so we're not talking about anchoring GFLow differences))

EDIT - to add: I couldn't get a direct connection between my dive computer and the computer so I exported from SWCloud using UDDF. Everything looks correct.

Funny you should post this.
I noticed some significant differences the other day but didn't have time to dig into them.
It was in regard to accidentally exceeding NDL by a few minutes and incurring unplanned deco and the length of said unplanned deco at various GFlo and Hi settings.

At the time I figured I had messed something up because it was late and I was tired, but didn't have time to double check things and haven't dug back into it.

What I thought I saw was significantly higher numbers for deco time in Subsurface.

Its too late here to check now, but you just reminded me to go take a look again and see if I screwed up or they are coming up with significantly different results.

 

[Jay]

Thanks @atdotde

Yes, all non-stop dives.

Please find two dive, #105 and #121. Both UDDFs attached. (please rename and remove the .csv extension - I had to put it there as the system won't allow a .uddf file to be uploaded)

Both GF 99/99, Salt. (I couldn't see where SubS displays/confirms Salt for this imported dive - dives were exported from SWCloud into SubS using UDDF format, unsure if Salt carried on over.)

Dive 105: SubS shows a min NDL of 13min at 15:10, SWC shows 24min. SubS shows a SurfGF (at surface) of 56% and SWC shows 41% <-- some screen shots attached.

Dive 121: SubS shows in deco, SWC shows 9min at 14:40. SubS shows a SurfGF (at surface) of 81% and SWC shows 61%

 

[atdotde]

Here is what I came up with when analysing your dive 105 in Subsurface: First of all, the NDL/TTS calculation is somewhat resource expensive as for each value a full decompression plan has to be calculated. Therefore, Subsurface does not really compute this value for each waypoint of the dive profile but only once every 30s of divetime. Plus it makes some simplifying assumptions like the end of NDL being the time a ceiling appears (which is shorter than the time we could stay at the momentary depth without any stops on the way up as the ceiling might already clear on the way to the first stop).
If you want more accurate data, you should select Log->Edit dive in planner from the menu. There you can remove all profile points after the 15min mark by Ctrl-clicking on the dustbin next to the corresponding segment in the table (Ctrl means delete all further waypoints not just one) and then use the recreational planning mode which is kind of a pimped NDL calculation.
When I first did that I got a maximal stay at that depth of 23min, much closer to what there Shearwater value is. But then I realised that around the interesting time, there is quite some depth variation in the profile: In about 40s, you ascent from 26m to 22m. But of course, for a given tissue saturation, the NDL depends on the current depth. And those 23min were for a depth close to 22m. So I adjusted the depth to 26m and there we are: That makes all of the difference as at that depth, tissue loading is so much faster that you can stay only another 13min without incurring a ceiling. So all the difference is in the depth you use to calculate the NDL. So Subsurface has picked a profile point (remember it does that every 30s) that was significantly deeper than the profile point that Shearwater used to compute the NDL.
Does this make sense?

 

[Jay]

Thank you very much for that analysis and insight.