Jensen567's Entry for the 5th Annual Old Lumens Scratch Build Contest - Machine Made Category

What Don said. Looks like you are going to finish in plenty of time. :+1:

Yep, coming along fast! You’ll be sitting down having a drink by the time I get mine done. :partying_face:

Well I sort of made the decision to just say forget all the plans and drawings, I’m just gonna build the thing and solve the problems as they come. As a result I’ve spent a lot more time actually making the light and less trying to figure out how I can make it perfect. Along the way I’ve learned a lot, and could probably make a second one much nicer.

I’m hoping to have it done by the end of this week or early next week, then I’m going to enjoy the heck out of it because it’ll have lots of run time and beautiful tint. Plus I may have the only light designed specifically for a 21700 right now.

looking good

Little bit more progress today. Got the tailcap mostly done, all the internal geometry is cut and the threads are cut. I spun the outside down to size to match the head.

Hoping tomorrow I can finish up the outside profile on the back, as it is a little bit long, and I want it to tailstand, so I need to recess the switch boot. After that I just need to make the switch washer and put the brass insert into it. Then the host will be complete!

Making some chips. The inside is already done at this point, just taking down the OD.

Can see how much extra material I left on the back. That small hole on the bottom is what the switch boot goes through.

Light is done! Really happy with it, a nice 4000k color temperature, guessing around 2000 lumens, beautiful beam from the khatod 10 degree quad optic and RMMs quad board, and lots of runtime from that almost 5Ah 21700 cell. The host doesn’t look half bad either. Lots more details with the pics.

Finished up the machining on the tailcap first. Some grooves to match the head and give it a little style. Sand blasted finish of course.

Tailcap assembled. Brass ring was a very tight fit, but not interference fit like the insert in the head, so I used a center punch to peen it into the tailcap around the edges. Won’t be going anywhere. After that it is a standard tailcap assembly, boot, washer, switch. Then soldered in place. Measured it at about 5 milli-ohms at 10A with the big Omten.

Now to take care of the driver. I used a heavily modified H2-C from kaidomain. First I removed the reverse input protection FETs on the bottom and bridged them with lots of copper braid and solder. Should help with heat from the IC too. I have covered this more in my thread on buck and boost converters here.

After that I changed the spring for a convoy spring and bypassed it with solder braid. I also replaced the short wire off the inductor from 24ga to 20ga, and replaced the output wires with nice short 20ga as well. That was the easy stuff. Next, using the schematic I posted (also in the driver thread above), and the MP3428 datasheet I did some resistor mods to change a bunch of parameters.

Once the inductor is off we can see the circuit. The purple circle is the current sense resistor, stock is 2XR100 in parallel, giving about 1.5A on high. I replaced these with an R025, R075, and R500 in parallel to give me about an R018, which I measured at 4A on high. In the green circle we have the high side FB resistor (R1 in my schematic) for setting the output voltage. Stock the driver is setup to run an XHP35, but to help efficiency I wanted to run 2P2S emitters so I needed a 6 volt type output. I swapped the 200k resistor for an 82k, giving me about 7.5V OCV, which is decent for 2S emitters. Finally, in the blue circle is the high side resistor for the battery voltage divider (R11 in my schematic), this is used for driver LVP. Stock is setup to reduce current and eventually cut out at 3.0V, I wanted to get a lower cutoff, so I stacked a 130k resistor on top of the stock 33k resistor to give me a 26.3k resistor for a roughly 2.7V cutoff. Put it all back together and it worked great. 4A on turbo to the emitters, after 60 seconds it starts dropping to about 2.8A over another 60 seconds. Low mode is about 20mA to the emitters, nice and low.

Next up was emitters, since it was a quad I would need 4, and Nichia is my go-to, after using the 90 CRI Nichias, nothing else seems good enough. I used one of RMMs MTN Quad boards which I slightly modified. It is setup for Carclo quad optics, the khatod has the same emitter spacing, but the standoffs are further out, so I turned the optic leg holes into slots, the khatod optics hit the edge of the board to press it to the head once the bezel is tight. Since i am a fan of tint mixing and Vf mixing with boost drivers, I decided to go with 4 separate emitters, and get as much red as I could at 4000k. I went with a 219B SW45k R9080 in parallel with a 219B SW40 R9080, then those were in series with a 219C SM4070e and a 219C SM353 in parallel. The net tint is about 4000k, with a nice rosy hue. The reason I decided to stick the two pairs of emitters in parallel is so that I should be able to keep running if any 2 emitters fail.

Finally time to assemble the head. First I dropped the driver in, perfect fit, and soldered it around the entire ring. Next up was the emitters, some AS5 thermal paste for the MCPCB, solder the driver wires, then just stack on the khatod optics, Kaidomain AR lens, and screw down the bezel. Now there is just one thing left, screw the parts together, pop in a battery, and turn it on. And….

It works perfectly first press of the button!

Hope you all can follow that long post! I really enjoyed building this, and it feels really special to hold a light I made myself from a piece of aluminum and copper rod, and a random bathtub fitting. Time for me to sit down, enjoy my light, and have a nice homebrew hard cider.

That is always a big relief!!


Considering your labor, that is a 8000 dollar light, but one hell-of-an impressive work of art!

You enjoyed building this light and I enjoyed reading about it just as much. I admire your driver modding skills as thats way out of my league. Well done and congrats on the finished working light. :beer:

Awsome job

Thanks guys. Does anyone know if there is another light designed around the 21700 cell out there? I’ve only taken a quick look around Google and didn’t turn up anything. Found a few 20700 and I know you can fit them in some 26650 hosts, but I didn’t find anything made specifically for this cell size.

Nice work and images. I really enjoy seeing how folks work!!! I need more time here. Thanks for Sharing!


So I have identified one mistake I made in my driver modifications. When I changed my LVP value by modifying the resistor, I neglected to think about the fact that this driver can run on 2S cells also, and has LVP for those as well. The result is that with a fully charged single cell the light starts flashing as if the battery is dead, because it thinks it is at the lower voltage for a 2S input.

This goes away once the cell drops below about 4.05V and runs normally down to 2.7V or so. I did all my testing yesterday with a bench supply set to 4V and a cell at 3.7V, so I didn’t catch it.

The solution in this case is to modify the LVP divider again, this time so that I am only using the 2S range, in my case I will swap the stock 33k resistor for a 9.1k resistor, this should give me a cutoff at 2.6-2.7V, and I won’t have any high side problems as there is no 3S LVP value.

All back together and the new resistor mod is working as expected, no more flashing from 4.2V to 4.05V, and cuts off right around 2.7V. I ended up stacking a 13k resistor on top of the stock 33k resistor for 9.326k.

very good build ! i love home build light

Really impressed by how you are able to modify, correct and adapt different situations to your needs to make this final quad 21700 flashlight.
It’s a nice combination of all your current skills and knowledges you’ve accumulated so far.

The light may be done, but I’m not! Decided to make some accessories. Some of these are 3D printed items, and I did draw and print them myself, but a 3D printer is not exactly a machine shop tool, so exclude them from judging in the competiton as I don’t think they meet the rules. Still want to share regardless though!

First up I made a lanyard hole in the tailcap. I never use a lanyard with my lights, but I like the hole as it does serve another purpose which will become apparent in a moment.

Next up I 3D printed a diffuser with some Taulman 910 filament, it starts clear but looks translucent white when printed. The surface finish is terrible because the filament I have is fairly old, and has absorbed a lot of water. The water actually boils out when extruded and causes bubbles to form in the plastic. Thankfully for this application, that’s not a problem. The design is similar to the diffuser Convoy sells for the M1 on the end.

Here is the real reason for the lanyard hole. I used a leftover piece of 1/16” stainless rod I had from another project to fashion a small hook, that happens to fit perfectly in the lanyard hole. The idea is that I can just hang the light with the diffuser on it out on a tree branch for a camp lantern, or at home on a ceiling fan if the power goes out.

Works well, already used it to do some work under my desk.

Finally I printed up some battery spacers for 20700 (this one could really just be a piece of paper), 18650, and 16650, so I can throw whatever battery I happen to have ready in the tube. I could have machined these, but it is so much cheaper to just print them, and they aren’t anything groundbreaking, just a size I didn’t have yet.

I’ve also learned some things about this light since I have been carrying it as my EDC. Mainly, that the sandblasted surface finish is fairly delicate and scratches pretty easily, which is not surprising really. There are already battle scars, doesn’t effect the function at all though, so just adds character. Otherwise it has been functioning as expected, I used it Saturday night in ceiling bounce for almost 3 hours on the second mode, no hiccups and the light got warm but not hot.

I built up another H2-C driver to test, oddly using the same sense resistor combination gave me slightly less current, 3.9A instead of 4.0A, but close enough, probably just resistor tolerances. I was really most interested in efficiency.

Also worth mentioning that I would take the measurements in Turbo with a grain of salt, my electronic load showed the voltage bouncing around and around 6.9V, whereas the meter measuring voltage on the driver board was a steady 6.36, all other modes were solid on both with the load showing lower as expected, so not sure what was happening there.

Either way, very happy with these results. Maintains great regulation through the entire discharge (Barring turbo), and efficiency in the modes that matter is basically 90+.

Why is the input current at the turbo mode dropping, when the input voltage decreases? The current should get higher, the lower the input voltage is. Output current is dropping as well. Why is this happening? Is this because the input voltage drops too low and the internal resistance limits the current? But then the internal resistance would have to be like 0.5 Ohm, so way too much.