Planner: Multiple Deco calculations ?

[Guillaume Gardet]

Le 28/08/2014 16:38, Dirk Hohndel a écrit :
> On Thu, Aug 28, 2014 at 04:30:31PM +0200, Guillaume Gardet wrote:
>> For Bühlmann there are multiple models :
>> * number of tissue compartments (8, 12 or 16)
>> * A/B/C variants (at least for ZHL-16): A is theoretical only, B is for tables generation and C is for dive computer (more conservative).
>>
>> So, we could implement those variants.
> The biggest challenge here is "who would implement this".
> We have learned that the discrete implementation of these models is rather
> tricky (as a divelog we only get the samples at the sample rate the dive
> computer returns to us). And because of that, even if we know exactly
> which model a dive computer is designed to implement, we still usually
> don't match the deco data the dive computer shows, as it has access to the
> continuous depth information. Additionally many divecomputers have rather
> under-powered CPUs (to say the least) and they are taking serious
> shortcuts when calculating the compartment values - which also leads to
> different results.
>
> Long story short: it's not clear to me if there is value in implementing
> other variations of the Buhlmann algorithm - it's unlikely we'll ever
> perfectly "match" any dive computer you are diving.

Sure we will never perfectly match logged dives, but my use case would be more in planning mode.

>
> I'd be more interested in a VPM implementation (that is reasonably well
> documented). We had a GSOC proposal to do just that and decided not to
> accept it but it's still something that we are open to if there's a
> developer with the necessary math / physics background willing to tackle
> the problem.

Do you have pointers for implementing VPM? Willem wrote it was totally proprietary.


Guillaume