I’m always looking for a better way to throw a lot of lumens a long distance from a compact light. I have been a fan of the Ledil CUTE 3 optic with dedomed XPL or XPG2 emitters for its ability to pack a lot of lumens in the beam with some decent throw and for its relatively easy integration into lights with the XP32 MCPCB. I also like the beam profile of the TIRs. With this light I wanted to improve upon the CUTE 3 setup in output and throw while keeping the head size compact.
I have had good experience with using the 20mm narrow beam TIR from LEDDNA in my S2+ and in my 7xXPL convoy L2 mod, so I planned to use them in this mod as well. I used dedomed XPG2 S4 2B emitters instead of dedomed XPLs to keep the runtime and heat a bit more manageable in a light this size. I also bored out the tube to fit the new Sanyo 20700 cells. I chose to go with the higher capacity “B” cell which will give this light a nice runtime boost. The 20700 cell hasn’t arrived yet, so I will update the post when it arrives tomorrow.
I’m pretty happy with the result. It draws about 10A with a fresh 30Q cell, measured using a 10mOhm shunt in place of the tail cap. I was hoping/expecting an amp or two more, but the output is good nonetheless. It outputs approximately 3200 lumens at start and approximately 3000 lumens at 15s, measured by ceiling bounce. At 30s it does 90Kcd. This is with an AR lens. This is a nice improvement over the CUTE 3 dedomed XPG2 which does ~2300 lumens and ~50Kcd. I think this quad setup provides a nice combination of output and throw that is hard to beat in a similarly sized light. One could also use dedomed XPLs instead of XPG2s and this would increase the output and decrease the throw some, but in my opinion that load would be almost too much for a single cell light this size (as far as heat production and runtime).
The driver is a MTN FET+1 with ToyKeeper’s Crescendo FW, which I love.
I started by making the aluminum spacer on a lathe.
It is a tight fit in the head. I had to trim the MCPCBs in order for them to fit (trimmed using cut-off wheel and dremel). I also had to do quite a bit of shaping of the TIRs (by filing) to get them to fit. The bezel blocks some light from the outer sections of the TIRs, so I bored out the bezel some to reduce the blockage.
This is a picture comparing the fronts of the quad C8 and a CUTE 3 optic. Taken from about 1.5ft.
This is a comparison beamshot between a stock reflector EE X6 with XPL HI V2 1A (left) and the quad XPG2 (right). Lights were not on full so this is just to compare beam sizes, and this photo is a good representation of this. As with most TIRs, this light has a larger hotspot than same-sized traditional reflectors at the expense of a more narrow spill.
So, compared to a (stock reflector) EE X6 with dedomed XPL V6 1A, which will do about 94Kcd, this quad has similar throw with a nice increase in hotspot size. Compared to a C8 with dedomed XPL V6 1A, which will do 155Kcd, this quad has significantly less throw, but a much larger hotspot.
I received the 20700 cell today. With this NCR20700B cell it pulls 9.7A instead of 10A with the 30Q. The tube is just long enough to accommodate the longer cell.
I am confused about something now, though. I measured the output again and the measurements are inconsistent with previous measurements. When I first built the light I took a measurement before I bored out the bezel or received the AR lens. With a 30Q this initial measurement was 2880 lumens at 15s and 77Kcd at 30s. Then I bored out the bezel, installed the AR lens, and adjusted the focus of the TIRs some because I noticed the focus was not optimized. I then did a measurement with the same 30Q cell (after recharging) and got 3040 lumens at 15s and 90Kcd at 30s, the numbers I quoted in the first post. The output increased by 6%, which is consistent with the change to the AR lens. The throw increased by more, which is consistent with the additional focus adjustment which would increase the throw but not the output.
Today, with nothing changed in the light I did another measurement with the same 30Q and got numbers very similar to the lower initial measurements. The measured current is the same as before. I tried another tailcap/switch just in case something changed in the switch to lower the current with the tailcap installed, but it didn’t change anything. So, I can’t explain why the measurement is lower now. Measuring other lights suggests my meter is reading fine now.
Using a boring bar tool on a lathe. Like I noted above, the tube wall is pretty thin now in some places. It is definitely not possible for lots of 18650 tubes. It turns out I didn’t have to bore the tube out as much as I did, though. The 20700 cell rattles just a little bit now.
It takes a bit of time but a drill bit wrapped with tape and 180-220 sandpaper will work just fine. Wrap the tube with tape for a better grip and wear gloves, mask, and eye protection as it gets hot with fine aluminum dust. You want a snug enough fit that you can just get the tube over the sandpaper and hold it still against the torque. Add tape under the paper as the bore increases and figure on using most of a sheet.
I replaced the dedomed XPG2s with sliced SST20s, hoping to get some more output. It pulls 15.3A (with VTC5A) measured using a 10mOhm shunt replacing the tail cap. It does around 3900 lumens at start and 3700 lumens at 5s and 92kcd at 10s.
Right after I changed the emitters I measured the current and it was only 12A. Based on the emitter forward voltage and estimated circuit resistances I estimated it should pull around 16A, so I suspected there was some extra resistance somewhere, around 16mOhms. Sure enough I found it the first place I checked, between the driver ground and the head/tube threads; 15mOhms. This driver is one with the thinner PCBs from mtnelectronics. I had put solder blobs on the ground ring so it made good contact with the retaining ring, and it seemed like it was making good contact, but apparently it wasn’t. I beefed up these blobs with pieces of copper wire and got the resistance down to 1mOhm. I think the bad ground ring connection must have been the reason for the inconsistent measurements I described in earlier posts. Anyway I am very happy to have boosted the output of this light and to have figured out the source of the low/inconsistent performance I saw before.
The oxide layer that forms on most solder is likely a cause for the increased contact resistance. For this reason, I’ve stopped using solder blobs for contacts wherever possible. Instead, I’ll solder on little brass or copper discs/bb’s to make the contact surface less oxidation/abrasion sensitive. DeOxit D100 works great as well.