Lume1-FW3X: Constant Current Buck-Boost & FET Driver with Anduril1/2 + RGB Aux

High CRI is my only preference. I have 219B lights, hell I modded a 219B Emisar D18 but you’d think they were like ancient artifacts that need white gloves to handle. If someone kills one with one of these drivers you probably learned a good $9 lesson in modding and can simply buy more and learn from whatever mistake was made.

I can’t help but see it as a distraction. The first thing people want from our hardware and software gurus is a way to bastardize an excellent driver before it has even been tested for what it is.

I’ll end my rant with that though. I know people are very attached to their LEDs and don’t want to risk ruining them it just doesn’t resonate with me. I’ve ruined so many things modding that this seems like a non issue, it comes with the territory.

I haven’t done it yet… that’s a few steps ahead. Here’s how the plan looks at the moment:

  1. LoneOceans created a nice new driver.
  2. LoneOceans got Anduril working on the new driver, using an old branch of the Noctigon K1 code.
  3. I isolated the changes and adapted them to a current version of the FSM branch. (was necessary because the ADC and thermal code have been completely rewritten since the K1 came out)
  4. Test on real hardware to make sure adapted code works. <— we’re here now
  5. Merge changes into upstream FSM branch.
  6. Merge changes into Anduril2 branch.
  7. Test again on real hardware. (optional; changes should be pretty straightforward and low-risk)
  8. Finish Anduril2 (mostly just needs more documentation).
  9. Merge Anduril2 into upstream FSM branch.
  10. Publish builds for enthusiasts to try, and get more widespread testing.
  11. Merge into trunk.
  12. Send builds to manufacturers.

This would probably all be done already, but I’ve had very limited time to work on things lately.

Thanks for the clarification, will work on the updated branch :)


I do have one question that is a bit off topic. Based on your other pure boost drivers if we had triple MCPCB’s which were configured in series could a 9V boost driver have been an easier design with less tradeoffs or does that have negatives of its own?

Actually that would be my preferred design if I were to build an entire flashlight from scratch, but it would be impractical for me to do so if I was going to build one for the community because there's a huge demand for 'turbo direct-drive mode' (which is not quite my preference, though not possible with a boost design), and because MCPCBs are typically not in series. Having series-LEDs would in fact, allow some interesting opportunities, such as using different kinds of LEDs on a single MCPCB with equal current through all of them.

I'd love to have something like a 3 or 4 LED (with mixed tints) with a 15-20W driver, or even better, a double channel boost flashlight with say 5+4 Nicha E21s or even just 2+2 2700K and 5000K LEDs with tint mixing. This could be made small and compact, and Anduril already has support for this.

I dropped a few messages to flashlight manufacturers at the beginning of the year including the examples above, but alas nobody seems to be interested, or had their own projects already, and most of them wanted 'headline' features like maximum turbo brightness. For example, having '4300 lumens maximum' sounds a lot better than '2000 lumens maximum', even if the 4300 lumens lasts for just a few seconds... as opposed to high efficiency.

Maybe tagging up with Clemence and Skilhunt would be interesting.


The market is flooded with triples and quads and FETs and all the lumens and pocket smoldering, skin scorching half baked lights one can imagine. SBT90s in featherweight “EDCs”, 10lb behemoths that don’t fit in a backpack, 87 pieces of XHP70s running off of a single 21700. I don’t desire any of them.

My custom FW1A (soon with a Lume1 once Anduril2 and some spare cash comes around) is currently my most valued light. It probably doesn’t even make 800 lumens on turbo with it’s sliced SST20.

I still desire slimmer hosts (~22mm for 18650, 25 for 21700), variable CCT, dual beam pattern, efficient buck and/or boost drivers, and for the love of all things good… a proper POCKET CLIP! OH and uhhhh if Luminus is listening by chance — SST40 95CRI. Thanks!

I suspected as much, too bad. There are PCBs that have jumpers to allow for either configuration but expecting people to get specialized pcb in addition to driver is a bit more involved than just dropping in a new driver like the FW series.


I can’t seem to get any companies interested in something slim and moderate though, or something with variable CCT or variable beam pattern. And it’s really hard to get anyone to make a good pocket clip.

I’m pretty happy with just one LED, especially if it’s in a host which is barely bigger than the battery. And I find “narrow” is more ergonomic than “short”.

If a light is a quad, I’d prefer to have warm and cool LEDs I can ramp between. Or for a triple, it could have one throwy LED and two floody LEDs with a ramp between… ideally on a headlamp. Like, in a T shape.

I agree with all of the above, I’m looking for a more compact high efficiency option. The FW3A is awesome for wow, but realistically a quality single emitter with a high efficiency driver in a narrower package is my preferred form factor. Apart from led choice and a lack of throw, the zebralight SC64c is the closest thing for me. The FW3X has throw by virtue of sheer output, and the FW1A isnt available in Ti or Cu with a narrower head.

The Lume1 has in my eyes really dealt with the first and biggest issue which is the linear driver, and for that I am really happy. Next is to have a non-fet option, with the UI adjusted to suit.

In the D4 era, BLF convinced a manufacturer to make a 6 lumen, single mode twisty (Sofirn C01). And the lantern was a very novel project, but BLF worked out all the key details for the body, the driver, and the UI.

If we show a sizeable interest list, a clearly defined and stable feature set, and keep the amount of development work reasonable, it seems like we can usually find a manufacturer to build it.

It might not be a good time, though, since the company that has been most accommodating of BLF requests recently seems to have quite a few things on their plate at the moment. Also, it always takes commitment ahead of time to conceive and prove out the concept and manage an interest list before a manufacturer will commit. I know I don’t have to tell you about how these projects can burn members out, so I’ve generally kept to making supportive comments when concepts like these are mentioned, but not pushed too hard.

A 14500 or 16340 based Anduril flashlight would be brilliant. Tint ramping would make it even better, and hopefully work without needing further modifications of the existing lantern firmware. It can be done in a compact host, if it’s ok to be floody - Clemence already offers an MCPCB for a Nichia E17A quadtrix that can be configured for multiple channels, and he recently found a spiral honeycomb TIR that blends the beam well.

Output would be modest, but easily meet my EDC needs - a 15mm, two-sided driver should be able to fit 2 channels of 3x7135 each.

Or keep it a single tint if a no-FET Lume1 variant could fit.

Less EDC-optimized, I also wonder if maybe Hank might try adapting his 5A linear driver to have 2 channels (no FET) for a tint-mixing quad. In fact, if he did that and also offered an alternate bezel option that was a lantern-style diffuser, then that right there would be both a great flashlight, and I think provide every desired feature of the LT1m mini-lantern except USB charging.

I’ve digressed, so to get back on topic, I agree with loneoceans that maximum turbo brightness is not my priority. In fact, one of the reasons I’ve been following the Lume1 development is I presume the better efficiency should provide a slight advantage in sustained output for the FW3A. Maukka measured 280 lumens under thermal regulation in his prototype review when the battery was nearly full, but 420 lumens when it was nearly empty, and the linear driver wasn’t burning off so much excess energy as heat.

That’s 1 of 2 theoretical advantages for sustained output. The stock FW3A driver also PWM’s the LED’s at ~933mA each. Lume1 should be feeding them constant current at around 400-500mA each, so the emitters will be slightly more efficient, too (example data and data source). That should raise the sustained output a hair more - perhaps close to 500 lumens.

Maybe the recent improvements in Anduril’s thermal control will eek yet a little bit more out of it.

I am also +1 about the status of current flashlight situations.

I would like the idea of the 9V (3 LED’s in series) version of the driver, as well. Then, we can use E21 or other domeless and small die LED’s for the best throw with 10507, assuming using the existing components. I would go with 18650 for the diameter of the tube, meaning narrower. With 10507, the height will be shorter, too. For the AUX board, we can maximize the output and put E17 or other small LED’s for the flood and any color combinations. The AUX mode can take care of the moon mode, too, for the better optimization of the main circuit.

It is not my priority to have the maximum TURBO mode, either. Especially, with the configurations of the Lume1 right now, the Turbo mode will be less bright than the maximum regulated mode as the battery voltage gets low.

The wires for the AUX board are coming next week. Hopefully, I can swap the driver with Lume1 without too much difficulty.

Thank you very much for all of the tips and ideas for the assembly.

Is there an interest list for a ToyKeeper/loneoceans designed flashlight. :) I don’t care which one of these ideas gets up they’re all great. I’m ready to place my order and pay a deposit. If Hank Wang can be convinced to manufacture the light it’s guaranteed to be great.

Can anyone recommend a modder to hire for a lume1 and Aux install? I’m spread way too thin to attempt it myself right now. I have an FW3C that I was hoping to have modded by jku2017 but I believe he retired from modding.

It would help to say whereabouts in the world you are.

US. Lower 48. TN

Finally got up enough courage and set aside enough time to install the FW3X-Lume1 driver and Aux Board. Glad I did not rush because this was honestly one of the more taxing installs I’ve tackled. Everything is tight and small.

Thanks to the advice, observations and installation tips of folks on this thread it probably went better than if I had not followed their footsteps. So here too are my experiences.

(1) As received the Aux board emitter holes didn’t align with the installed LEDs (XP-L HI’s). Being a triple, I expected as much. Fortunately, there was enough space around the Aux circuit board that I was able to enlarge the holes and manage to get the Aux board over all three emitters.

[Before Adjustment, Aux wouldn’t site flat because emitter holes didn’t align]

[After some filing, fits flush to MCPCB surface]

(2) My previous mod experience has always involved upsizing wires, so I didn’t have any smaller than 24 gauge for the Aux wiring. I purchased some high temperature 28 gauge for the occasion. Rather than purchase a variety of colors, I got plain white and colored the tips RBG to help with install.

(3) Swapping the teeny tiny resistors was an exercise in patience. At one point I lost one of the resistors and spent 15 minutes searching for something the size of a grain of salt. (Actually, I think salt’s bigger.)
Decided to go -

  • Red 220 Ohm
  • Green 330 Ohm
  • Blue 100 Ohm

I think the light levels came out fairly even across the three colors (same manual camera exposure levels, for comparison purposes)
Red Green
Blue Magenta
Yellow Cyan

And almost forgot, presumable the RGB blended White
White (White. diffused)

Thanks to Loneoceans for sharing the Buck-Boost Driver and Aux designs with us, and thanks to NealsGadgets for making them available.

Edited to add Aux RGB White combo and references to original resistor values.

I’d love this, I got a sofirn IF25 with tint ramping where I replaced the LEDs with sliced LH351d and I really like the ability to choose the color temperature, with the bonus of CCT mixing improving the tint and increasing color saturation. The host and driver aren’t particularly great though, not very efficient and lacking proper thermal regulation, an efficient 2 channels anduril driver would be much better.

JaredM mentionned Clemence and Skilhunt, as far as I know Skilhunt doesn’t chase high lumens with insane turbos and they typically use buck or boost drivers.

So this came out a lot less helpful than I imagined but I recorded it so I might as well upload it. Trying to solder backwards facing the camera was even more difficult than I imagined so there are mistakes. Also, I would suggest going to youtube and making sure it’s in 1080p so you can better see me screw up.

Thanks for the tips, the helps, and the videos for the assembly. I was able to swap the driver with Lume1 along with AUX board. I chose the resistors of 1.8K for Red and Blue and 5.6K for Green, the recommended values. I am happy with these values. The low level is very dim, almost like trit. The high is bright enough. I think the disadvantage with these values is the color mixing to obtain the color somewhere in between like yellow, violet, white, and so on. Those in-between colors are not defined well. I can see red and blue clearly for violet, for example.

Everything seems to be working. I have got several questions, but I will start with those next.

The AUX LED’s have the selections of R, Y, G, C, B, V, W, Rainbow, Volts. I am still learning, but my questions are about Volts. It looks like mine has two different selections for Volts, so it is like R, Y, G, C, B, V, W, Rainbow, Volts No. 1, Volts No. 2. Functionality-wise Volts No. 1 and Volts No. 2 are exactly the same. There are just two of the same selections. At least, that is what I am thinking. What am I doing wrong?

The next question is about Volts, too. When Volts is selected, the color goes to V (violet) regardless of the battery voltage. It does not matter if Volts is selected from OFF or Lockout. There is one more to this situation. Volts is selected with OFF. Then, the main LED’s are turned on. After turning off the main LED, Volts shows the correct color for a while. Eventually, it goes back to violet. It is almost like some capacitor needs to be charged for the Volts function. Indeed, the higher current is, the less duration the main LED’s need to be on to have the correct color. Is it supposed to be this way?

I am new to AUX board. This is the first one ever. I like it very much. I am still learning about it. Indeed, I did some searches, but I could not find the good information.

Hopefully, mine is not broken. I may have broken it during the assembly. It was challenging. What do you think?

So, I did more search and found this from the review of one of the KR4 series.

My light is doing something similar to it. In fact, the color is between violet and blue. When the main LED’s are on for long enough of duration, the voltage of the battery gets lower, too.

The next one is from the same review as well.

There is no violet (purple or magenta) for Volts. Also, the highest voltage should be blue. I am getting violet with the freshly charged battery.

If I remember correctly, Lume1 uses Anduril based on the versions for one of Noctigon lights. (I don’t remember which one.) So, the information from the KR4 review is probably useful.

I understand the explanations from TK in the first quote. However, I am still lost. Particularly, violet is included in Volts.

I read the datasheet for Lume1-FW3X. I don’t think there is much information about it, either.

Hopefully, I can get some help.