I did both in my proof-of-concept code and reduced the temperature window to 2 degrees. It does really well in the simulation (3rd graph in #1598331).
It depends on the strength of the filtering that the sensor data needs, but assuming no additional filtering is required, the new code should require less program space than the old one.
For the Noctigon K1, I have it displaying realtime voltage data on the colored button. This data was very noisy though, and the button changed rapidly between colors a lot of the time. Basically, if voltage is mapped on a vertical axis and time is on a horizontal axis, the button color looked something like this…
So I added a simple lowpass filter to the voltage values, using extra bits to make extra precision possible. It’s still noisy, but far less than it was before. Now the result looks more like this:
So I’ve been thinking of doing the same with the temperature values. I just haven’t done it yet.
Convincing
"And if you labeled your axes, I could tell you exactly how MUCH better."
Just kidding! 
After flashing revision 463 onto my D4V2 XP-L HI I got this:
… I had to manually turn the light off to prevent damage. Reverting back to 450 fixed it.
Haven’t you tried running the test on 463 with higher temp limit, like 60C?
Everything was at default settings, the only thing I touched was the calibration of the temperature sensor.
Edit: Probably related with this: Flashlight Firmware Repository - #2155 by 0-8-15_User
Thanks for reporting this. This is exactly the sort of thing I thought might happen after merging such a large patch a couple weeks ago, and the reason why I don’t consider those new builds stable yet. Not looking forward to debugging it, but I’ll see if I can find the issue and fix it.
Do you use debugWIRE for debugging?
Isn’t debugWIRE too slow for real-time applications?
This is what the first attempt looks like:
This is what the second attempt looks like:
I just went looking for this.
Does anyone know if the tailcap is interchangeable with older model D4v1?
I might just give it a try regardless, but if it can be confirmed, I will have to include an extra magnetic tailcap for my v1 in my next order.
I’m also using a V1 (maybe from the later model) tailcap on my V2. I do not need a magnet and it makes the light a few mm shorter. The styling looks even more fitting than the native V2 tailcap.
Looks promising. Could you test your code with higher temp limits ( 60C or 65C )?
P.S. In my opinion staying as close as possible to the temp limit shouldn’t be the top priority, steady output that’s what really is important.
I completely agree with that. This is what the simulated target looks like:
Looks pretty good.
At the 60 degree setting, output looks fairly stable at 20% power, which should correspond to approximately 860 lumens with XPL HI.
The v2 tail cap has fitted fine on my D4v1 (from the middle of the production run) and also on a D1. The D4v2 tube fits both of these earlier models, too, although the threads are a bit tight especially in the head of the D4. The colour match of the greys is good but not perfect.
I confirm that the v1 aluminum tailcap fits just fine on any v2 aluminum body. I’ve personally done that with two different sets of tailcaps and body tubes and it works fine.
I prefer the v1’s tailcap as I don’t use a lanyard or magnet. The v1’s tailcap is shorter, weighs less, and is more comfortable in the palm of my hand.
Thank you for the answers. I seem to recall there were minor changes made to the V1 body at some point, and some of the threads might be different, but it sounds likely that V2 tailcaps are compatible with all V1 battery tubes, at least for the aluminum versions.
I would want the regular tailcap for pocket carry, but I’ve realized being able to easily swap for a magnetic tailcap would be handy in the garage and likely a handful of other situations.
you’ve got me mixing tail caps on my V! and V2 lights …… and I don’t want to 