comparison Common/Drivers/STM32F4xx/Release_Notes.html @ 224:ceecabfddb57 div-fixes-3

Bugfix, deco: fix 2 (small) problems with calculated ceiling This fixes 1 trivial, and 1 not really trivial bug in the calculation of the ceiling. When simulating a bounce dive to 80m, things become clear (tried this on a CCR dive, fixed setpoint 1.2bar, about 15 minutes of bottom time). Closely watch the behavior of the ceiling data. At some point during the ascent, the ceiling begins to decrease in 10cm steps. Then suddenly (while still ascending), the ceiling increases again with 1m, does not change for some time, and then suddenly steps 1.1m less deep. While not very relevant to real deco diving, it is simply wrong. The reason for this is subtle. The algorithm used to find the ceiling is a sort of linear search, stepping down a meter, overshoot the depth, and search back in 10cm steps. It seems some numerical instability. Fixing this, was a bit more computational intensive search by stepping up down in equal steps of 10cm. But, I'm pretty sure that things can be speeded up here, as a ceiling does not change fast, so it should be not that difficult to limit the search space, or use a binary search algorithm instead. The trivial second problem fixed, is that the ceiling ends at the surface and not at 1m depth. This small issue became visible after changing the step down size above. Signed-off-by: Jan Mulder <jlmulder@xs4all.nl>
author Jan Mulder <jlmulder@xs4all.nl>
date Sun, 31 Mar 2019 19:35:51 +0200
parents c78bcbd5deda
children
comparison
equal deleted inserted replaced
223:5f535ef6a3db 224:ceecabfddb57