1.0.2.0
Advanced sensory-data filtering using a history-less first-order low-pass filter with a medium α implemented. Previously, it was an integer-based running-average that required a 3-4 point memory history.
The new algorithm requires < 1.0 floating-point resolution so a custom fixed-point math library with DSP-acceleration is now used plus the actual filter equation was recoded directly in assembly for speed, just for fun.
Functionality that detected large and fast swings in sensor-data temporarily dropped because the filter's medium α duplicates some of the sensitivity.
Also, direct backlight power-state tracking (idea borrowed from Lumos) naively implemented to just avoid sensor-data calculations and adjustments for that poll loop without any visible affect on power-draw. It could affect the ability to adjust to light-levels deterministically, but this is a .0 release so the filter-equation can still use some tweaking.
__________________
Support what I do and donate me a frankfurter.
Last edited by NuShrike; 7th December 2008 at 04:43 AM..
|