Exactly. Diodes in the power path and small traces are the biggest culprits here, I’m sure of it.
Try bypassing the driver completely, full direct drive.
Are you using flattop or button top cells? It seem that those + bumps on the driver board are set up for flat tops. How do button tops react when they encounter those bumps? I was thinking about creating a whole new board with a large copper disc soldered down for the cell contact, but then you are pretty much committed to button tops. It’ll have to be one or the other.
That's my plan is to first try DD. I got bronchitis, so it's been slowing me down, on antibiotics since yesterday, I'm home but don't have access to my workshop right now - mother-n-law up for the holidays and my shop/office is downstairs.
Those CaroNite cells are a total unknown but can't be high performance, actually might be pretty bad.
Update: started on the mods - will go with a XM-L2 U4 1C initially (later a de-domed XP-G2 S2 2B). Least this way I'll know what I can expect in max amps, hopefully not fry the LED. I'll be staying with the stock driver MCB for it's cell brass contacts. Board traces are all wide. I was suspect of the thru-holes in the grnd ring -- not many and small, but when I did amp tests, I jammed one of the DMM leads under the driver (MCU side) and didn't see any noticeable difference, so I'm thinking it's ok.
Just changed the switch to 6A and linked the contacts with 18 AWG wires, an increase output of approx. 10%.
Then tried LG HG2 (used in my M43), not fully charged, got even lower:(
And finally changed the 18x 7135 driver, get slightly higher at high mode than the stock driver, about 5%, but ceiling bounds still lower than TK61. This method is quite fair because I zoomed the whole light spots to get similar diameter for all flashlights I tested.
Warning: my big springs cause a stiffer screw-in and Ouch! My CaroNite batteries got scratches around the edge of the flat head…… will change the button to copper ring then…… Sh*t!
Tom, your health is way more important. Get well first before you lay you magic fingers on BOSS1 again. There is no need to rush for this project actually. We will wait for your findings patiently 
. Don't forget to enjoy your holidays too!
OK, thanks freeme - just got your advice, but sorry to say it's too late 
.
The light is a wide hold for sure, and awkward. The pic with the Coke can shows the body of the light to be 8.3% wider, but in real life measured, it's 9.75% wider . It's the only light I'm thinking of leaving the handle on - SRK size lights make no sense to have a handle on it, but this one does for sure. On the handle, it feels well balanced. The handle is better than I thought it would be - it does have a cheap looking rubber covering, but it feels pretty darn good - nice and soft but rigged and balanced.
There's a few things I did:
- XM-L2 U4 1C/20 mm Noctigon - nice fit, sanded down the 20 mm shelf center under the MCPCB to 1000 GRIT paper (smoothed out machined rings), GC Extreme under the MCPCB
- 20 AWG LED wires - not the min size, little longer, shorter than stock 22 AWG wires though
- removed the 5 diodes, replaced with 20 AWG Teflon coated wires (on furthest away pads, ran a longer wire)
- wired 'LED-' wire to a direct ground pad (no driver, true DD)
Did not bypass springs or mod the tail PCB at all - left as-is stock. Did not secure or add thermal grease for the pill top/shelf, also left as-is.
Amps: 5.15A across the switch
lumens: 1870 @start, 1812 @30 secs
throw: 296 kcd, measured at 5m (pretty solid reading, not even noticeable drop over 20-40 secs or so)
This is pretty darn good for not even bypassing the springs or adding jumpers on the tail PCB. One EFEST 26650 4200 cell measured 5.05A direct on the driver, so I think 4 of those cells in parallel help.
I'd say a FET DD driver should work out pretty well in here. I don't think the driver board traces are hurting much, but more could be done to reduce resistance.
For the lumens measurement, for me it's expected, since I measured just over 1900 lumens in a recent Yezl Y3 mod with a 5.30A tail reading, plus a UCLp lens. Possible I could measure 2000 in this light with a UCLp in there, with the springs bypassed. UCLp's have been giving me at least 5.0% gains, maybe up to 7%. For kcd, a 5% bump in lumens could be a 10% bump in kcd - it seems to be always higher, I've seen double many times.
So, lots of possibilities with this light. My raw kcd measurement could be 5-10% low (possible) because the shorter distance of 5 meters. A de-domed XM-L2 U4 1C could hit a min of 600 kcd, maybe much higher. I saw Racer86 reported a ~250 kcd on a TK61 mod with dome, higher amps but on a U2 XM-L2. I think this light has more potential than a TK61.
More pics for the mods:
Stock light beamshot on ceiling:
Stock LED and wires (22 AWG). The scratch across the plate is my doing - tool slipped when prying it off. A coupe of taps on a hammer to the edge of the MCPCB loosened it up good, the screw driver type tool to pry it up.
Retaining ring, and 'filler' ring below the driver:
Stock driver:
Closer look at the rings:
One thread shown here - I'd consider this almost fragile, careful not to mangle them. I'm afraid it would not last over many uses, hopefuly you don't need many though:
You can see the trace routes here. Must be a simple 2 layer board, like our OSHPark drivers:
Before sanding:
After sanding:
Diodes removed:
LED re-flowed and mounted:
22 AWG Teflon jumpers added:
So, what’s the main biggest difference? Going direct drive and bypassing the driver? If you put the emitters ground lead back into the driver circuit after your mods, what do you get?
Seriously, a light like this only needs one mode, you don’t use it read the road map in the front seat of the car, right? Not a thing wrong with true direct drive.
So the cat is out of the bag, pop a dome on an XP-G2 S2 0D and let’s see what the K’s game is! :bigsmile:
This type of thrower with the heat sync'ing might just be a perfect candidate for no driver.
I'm debating - I got wight 044 drivers on the OSHPark queue, which Mitko has had great results with on XP-G2's de-domed.
I got like 8 XP-G2 S2 2B's left. One came out really well de-domed in a Convoy C8, at 5.01 amps on a 30Q @4.22v. I didn't order the UCLp yet - think Chris would cut me one at 90 mm, just have to confirm the size.
I got my SR-90 UCLp from Chris, think it was in stock, looks to be about 94mm measuring from the outside, 91mm in the clear showing through the bezel.
Edit, yep stock lens at 96mm, I cut it down to fit the SR-90 using a belt sander. $12.25 and worth every penny.
Those are really nice numbers for a domed emitter and original lens. I’m looking forward to more results!
Chris just sent me 3 UCLp custom cut sizes for the Convoy L2, UranusFire 818, and a Roche LS01. The Roche Ls01 has an XHP-70, zener mod FET driver, and got a 5.5% gain in lumens, where it supposedly had an AR lens in it. And the 5.5% is at least that, because I did before/after on the same cells and on a 10A drain, HE2 button tops can get knocked down in a few seconds.
Hi Tom, just to confirm, the ‘pill’ overlaps with the body by about 2mm? Does the sides of the pill contact the body as well or is just resting on the shelf? Thanks for the preview 
EDIT: freeme, I had initial ideas of using this light as a host for an MTG or XHP70 LED or even a tripple XHP50 but the shelf design has me hesitant. As a single XML XPL thrower, even DD, it seems sufficient though.
Well, I'd call it a plate, shelf, or pill top, not pill. Where the shelf overlaps the body, the shelf is 2.0 mm thick, but the overlap is about 4-6 mm I'm guessing - it's assembled right now. Not sure if maybe CWK mentioned that dimension earlier? Maybe not.
Post #111, first pic, shows the lip the LED shelf rests on - looks pretty wide in the pic. One could easily drill/tap holes there for setting screws, for example, or could just coat with thin layer of thermal grease.
Hahahaha 
you made me laugh, will have to put this into my sig 
exactly, you dont need lungs to mod a flashlight
Hope you get well soon!
Almost 300kcd with a domed XM-L2 and uncoated glass? 
And yeah this light doesn’t need modes, with 4 26650 that’s 20Ah of capacity, enough to run 4 hours straight at turbo! If one needs lower modes while carrying this monster, just switch to a smaller flashlight.
Yes, I’m thinking thermal grease is the ‘only’ option for us peeps without tools. Whenever you can, if possible, can you check if a C8 reflector or better yet, three of them can fit in the head? IS a triple with C8 reflectors possible? Thanks
hmmm
Well, post #108 has all the dimensions you need - plenty of depth, for width you'd have to figure whether an 89.7 mm diam reflector has the space for 3 C8's. You'll have a lot of vertical space to plug up. If it does, it's close. A Shocker is effectively 3 C8's reflectors, and all focused to boot.
Spring bypass's: added 22 AWG bypass's on all 5, including the center. Actually the center one is pretty cool for running on one 26650 cell only.
Now I got 5.0A on one cell in the center (measured across he switch), and 5.36A with all 4 cells @4.17v. Throw is now measured @5m for 308 kcd, again an XM-L2 U4 1C with dome.
Tried hot dedoming a XP-G2 -- no luck 
. Got one in the gas for 1.5 hrs so far - will take a while.