A question about thermal regulation:
At the moment, Anduril uses a thermal regulation method which knows nothing about the actual host characteristics or driver. It simply senses “overheating” or “underheating” like a user-adjustable thermostat, with some information about magnitude, like the degrees of excess multiplied by how long it has been in that condition. From this, it attempts to find a happy medium, with the adjustment speed determined by the magnitude. It’s fairly universal and typically works well with no need to hardcode anything about the host itself. Also, the adjustments are so smooth they’re basically invisible without a lux meter.
But.
On such a small and overpowered light, like the FW3A, overheating can happen so fast that it barely even has time to react. By the time it senses that things are trending too high, the light could already by too hot to touch. It works fine on a larger host like a Q8, or at a more reasonable power level like 3 Amps, because sensor lag is much less of a factor there.
So I’m wondering if I should change it. It currently regulates fine from the default ceiling level of ~1100 lm, but it doesn’t drop fast enough from full turbo.
Instead, I could make it drop immediately (or at least very very quickly) to the highest regulated level of ~1100 lm whenever it senses any overheating condition while the FET is active. Even if it’s over by just 1 degree for just a few seconds, it’d trigger a full turbo step-down and shut off the direct drive circuit. Maybe it’d still do the smooth ramp-down, but it would no longer be invisible — it’d be fast enough to see, like 4 seconds. And since there’s thermal lag, it might continue to regulate downward for bit afterward, overshooting its ideal target level, but at least the host wouldn’t get anywhere near as hot as a slower method.
Although dropping from 100% power to ~25% power is a huge change in the amount of heat generated, to the eye it looks more like it’s going from 100% to 80%. Effectively, it looks like dropping only one level.
So.
What do people prefer?
- Smooth but slow universal method which allows full turbo to run longer but may require the user to exercise common sense and turn the brightness down manually sometimes.
- Fast drop to a hardcoded level which is safer but may potentially be a little oversensitive and uses code instead of common sense to enforce “turbo is for brief use only”.
There is also still a muggle mode, which probably couldn’t start a fire even if someone wrapped it in dark-colored insulation with a low ignition temperature. So, use that for the kids. The question is how it should behave in non-muggle mode.