FW3A, a TLF/BLF EDC flashlight - SST-20 available, coupon codes public

Actually, accounting for the future is more like D action. Trying harder when it doesn’t work is integral action. We should do a bit of system identification on the driver and set a properly tuned PID controller and I bet things would work beautifully.

Maybe, but the GT team is swamped right now. The GT group buy is closing so there is tons of stuff going on.

Who knows, maybe The Miller will take over this FW3A GB again. :+1:

What I have now is more of a “PD” sort of system, with a small “I” component used mostly to reduce the effects of noise in the measurements. I tried making the “I” stronger but the results weren’t very good.

Specifically, it measures 4X per second, averages the past few values, and builds a temperature history using the averaged values. From this, it projects trends forward a few seconds to guess where it’ll be in the future. This is the “D” term. It keeps a few seconds worth of projected values, and takes the average, which is where the “I” term comes in. If the average projected value is above the ceiling, it sends an overheating signal to the UI code with the over-threshold magnitude (the “P” term) as a parameter. If the projected value is below the floor, it sends an underheating signal to the UI code (with the under-threshold magnitude as a parameter). Signals are rate-limited to avoid warning too often.

From there, the UI code decides what to do with it. In Anduril, it sets a new target ramp level based on the current level plus or minus the reported magnitude, with some bounds checking to keep things from going too high or too low. Then, on each clock tick, if the current level is different from the target level, it adjusts by the smallest step possible. Time between adjustments depends on how far it needs to go, so it could be anywhere from 0.016 s to 2.0 s per step. This interval is recalculated at each step so it decelerates as it approaches its destination.

In my testing, I’ve found it’s a bit tricky to hit the right balance between adjusting too fast (overshooting, oscillating) and adjusting too slow (overheating enough to hurt someone or cause damage). So most of what I’ve been doing is trying to make it reliably hit the Goldilocks area between.

Quite right.

Please add me for a second

Please add me to the list for one. Thanks.

I think I need to be on the list for one… :person_facepalming:

Please add me for a second unit

First unit: slot 235

I quite like the tapered battery tube, but i find it doesn’t match the not tapered rear end of the battery tube.
I think both ends should be tapered, but we’ll have to find out what tapering angles will look best, and if it should end with a straight section or not.
I’d be happy to photoshop different possibilities, but haven’t found a suitable picture to torture yet.
I guess the tail cap could have a smaller diameter too.

interested please add me

Put me down for one……

Please add me to the list

You are doing a great job.

Please add me on the list for 3 units.

Thanks

Please ad me for 1 light

One please

I’m in

Add me to the list please

Add me to the list please

Add me for a second until please and thanks guys

Put me down for one please