I got an Emisar D1K with an SFT-70 emitter in 5000k. The beam profile is exactly what I expected and the tint is fanastic! The only problem with it is that it gets warm quite quickly. It isn’t scalding hot but it’s still cool out and night and I’d call it uncomfortable at max ramp.
My Convoy M21F (XHP 70.3 hi) isn’t much bigger but seems to handle max output much better. I had thought that the Lume X1 driver would result in M21F like behavior in thermals. Is the Lume X1 just that much less efficient than the Convoy drivers or is this purely due to efficiency differences between the emitters?
I guess maybe I should check and see if I got the right driver. The product page states which emitters are supposed to come with it by default and this emitter is one of them.
So the SFT-70 is considered inefficient then? Unfortunately I did not hear this before now. Although I like the beam profile of this emitter I probably would have just gone with an XHP 70.3 Hi instead.
M21F is quite a bit bigger, it uses more efficient emitter. D1K is smaller and it’s driver harder, if the max ramp mode is above the sustained output it will get hot pretty quickly.
I have the same light with Sft-42R.
I always program my Anduril lights to the highest thermal ceiling which is 70C/158F. This light steps down after 45 seconds to a minute. The outside temperatures were in the mid 40s to low 50s. Fahrenheit
I programmed it over and over and over again to make sure I didn’t make a mistake which I did not. It was only 94° on the outside. Fahrenheit. There’s no way it was another 64 F degrees on the inside of the light. The body never gets hot. The head can get warm but I’ve never had a light step down quickly when it’s programmed to the max thermal setting. What happens is every 20 or 30 seconds I double click to put it back on turbo and it just keeps stepping down.
Other than that which is a big issue to me. I like the output of the light and the beam profile and the tint.
Sometimes I make stupid offers until I really think about it! I offered to buy another one for $35. A $10 discount . He accepted then I thought about it. If this light does the same thing I’m going to have spent $80 for two lights that don’t program correctly. Finally the switch will probably be worn out in 3 months to a year averaging using it 20 times for each battery!
So I use the light and I’m constantly putting it up to Turbo. Light never gets hot at least the handle doesn’t.
Hank is a nice guy. He’s always been fair with me. I rambled on enough.
Lights like the D1K/D4K/DA1K or really just any light in this exact form factor are not meant to be continuously driven on turbo - they simply lack sufficient thermal mass, or said in a less fancy way: they are not big and heavy enough to deal with the heat.
Until someone comes up with way more efficient LEDs or a completely new and efficient way of cooling stuff this is not going to change.
I totally understand that and agree with that. But they shouldn’t step down in 45 seconds or so when you have it set at 70c thermal limit. Especially when the temperature is about 45° Fahrenheit/ 7C.
Heat will build up with continuous running on high and turbo . That is natural with any light small or large. This happened from the very beginning when the light was cool.
Consider a DA1K with XHP 70.3 HI in 5000K or higher, possibly ask for the 70 CRI version (even more efficient) with Lume X1 driver. That’s about as efficient as you can get in 2026.
Still gets hot after a minute, just not quite as quickly.
The thermal PID is set to the driver, not the host, so the driver heats up much faster than the body. I’ve tested many many Anduril/A2 lights that have this, and the thermal throttling only happens once the light has stepped down and running at the regulated level.
With the boost drivers this is much less pronounced than FET drivers, and you can’t compare a Convoy boost driver to a LumeX1 (also very different leds). Thrle Xhp70.3 is much, much more efficient than the SFT-70 (bigger footprint, better power handling, lower thermal resistance).
So question, would an emitter swap to an XHP 70.3 Hi from the SFT-70 get me a light identical to if I’d ordered one with a XHP 70.3 Hi? Or are the drivers configured differently?
The thermal regulation in Anduril is tuned to handle hotrod lights like the original Emisar D4, which could start fires in just a few seconds, and overheat the host very quickly. It can’t wait for the host temperature to reach a limit, or it’ll be way too late. Instead it has to “steer into the turn”, so to speak.
What this means in practice is… if the temperature is rising very quickly, it’s likely to regulate down pretty quickly, even before it hits the configured temperature limit. But if the temperature is rising slowly, it’ll regulate slowly… mostly sticking as close as possible to the configured limit.
It also means that going directly from a cold start to full power will cause the hardest, fastest regulation… because that makes the temperature rise faster, so it has to steer harder. Warming up more slowly allows it to steer more gently.
Basically, it has “current temperature” and “rate of change”. It tries to predict the future by drawing a line forward from the current temperature with a slope matching the rate of change. If that line, traced forward a few seconds, lands above the configured limit… then it regulates down, proportional to how far beyond the limit the prediction was. Or if the prediction lands significantly below the limit, it’ll regulate up… even if it’s currently above the limit.
So, like, a simplified example with numbers:
Let’s say you’re at 30 C and the limit is 60 C, and you’re rising at 10 C per second… Then in 5 seconds, the prediction would be 80 C. 30 + (10 * 5) = 80 C. That’s 20 above the limit, so it’ll regulate down by 20 ramp levels. Ramp 150 → 130. It’s nowhere near the limit, but it’s already regulating down.
Then one second later, you’re at 38 C, and rising at 8 C per second. 38 + (8 * 5) = 78 C predicted, which is 18 C above the limit… so it regulates down 18 more levels. Ramp 130 → 112.
Then one second later, you’re at 42 C, and rising at 4 C per second. 42 + (4 * 5) = 62 C predicted, which is 2 C above the limit… so it regulates down 2 more levels. Ramp 112 → 110.
Then one second later, you’re at 45 C, and rising at 3 C per second. 45 + (3 * 5) = 60 C predicted, which is exactly the limit… so it stays at ramp 110.
It may do further adjustments from here, but they’ll usually be pretty small, ramping up or down by small amounts to stay close to the configured temperature.
I hope that helps explain the thermal regulation behavior.
But also… what everyone else said. There is no substitute for sheer mass. Bigger, heavier lights can run at higher power for longer. More efficient LEDs can also run at higher power for longer. And the temperature measurements are taken inside, right next to the source of the heat… so the outside might not feel hot yet. It’s like how if you sit next to a fire, you may soon be sweating even though someone on the other side of the room is still cold. It takes time to warm up the entire room.
If both LEDs are the same voltage, it would probably work… but I know there are cases where, like… one supported LED is 6V and one is 9V and one is 12V, so they use different sense resistors or something, on the driver. Or cases where one LED’s driver is limited to 2A while one is 5A and one is 8A, so they again have slightly different drivers.
But in this case, I don’t know.
I’d give it maybe a 80% chance of working, but I’m just pulling that number out of my ass. If I recall correctly, they’re both quad-die LEDs with the same package footprint and similar power levels and they both come in 6V or 12V flavors. So if you have the matching voltage, it’s likely to work.
Thank you TK for the explanation on that. I’ve always wondered about that phenomenon in these lights. Is the behavior the same with Anduril2 and Anduril?
I’ve had this experience with other Lume lights and for me it’s typically because the thermal calibration is off. I’ve had drivers that report 35 (all temps in celsius) when it should be 21.
Wait a few hours for the light to cool down and then check the calibration (from off 3C, then 2C). If it’s in the low 20’s, then in my experience it’s just a hotrod that heats up quickly. If it’s much higher, re-set the calibration (7h and then after the first flash click the button 20 times if your room temp is 20 degrees).
Mostly, yes. The thermal regulation was mostly written before Anduril 2, and it’s such a finicky bit of code and takes so long to test… that I’ve been trying to avoid ever having to change it.
The last time the thermal code changed much was about 6 years ago, right before my spine injury. I got it working pretty reliably on torches of basically every size, shape, and power level… and hope it won’t need to be modified any time soon.
If I ever have to mess with it again, I hope I can run it in an AVR emulator and use a thermal simulator instead of having to use real hardware. It’d be a lot easier, a lot faster, and these tests are really hard on both the lights and the batteries, so I burn through hardware while doing it.
It definitely feels like AI has been writing Windows lately. I know someone at Spotify was bragging about how nobody on their team was writing code.
“When I speak to my most senior engineers - the best developers we have - they actually say that they haven’t written a single line of code since December,” Gustav said. “They actually only generate code and supervise it.”