I modded Convoy S2+ with triple KD Nichia 4000K MCPCB, Kiriba-ru CU spacer soldered into the original brass pill and Qlite driver, which gives really nice Moonlight mod and around 1000 lumens on High mod. I had also sended down a legs of triple optic (KD one, not Carclo) make the lenght optimal. Nichia gives amazing slightly warm neutral tint, more like 4500-4700K.
In spite of the Samsung LH351D being thought of as a low Vf emitter, in a single emitter single cell light it’s difficult to really push one to it’s limits. I’ve seen around 1600-1700 lumens in such a set-up with a top cell and quality FET. Used em in an Eagle Eye X6, Emisar D1S, and in the GT mini. I tried pushing the GT mini even further by boring the battery tube and using an 20700A cell, but can’t get it to do 2000 regardless. I even put a solid copper disc with copper based spring and a bypass in the tail. Still, it’s a neat look in the beam profile and color, even if ~1800 is all I can get from it.
That’s because the lowest VF emitters are XPL2s and Nichia 219Cs.
The LH351D 70/80CRI only has a bit lower forward voltage than an XPL-HD, at 3,40V/3A for the LH351D vs 3,47V/3A for the XPL-HD.
Only the LH351D 90CRI has a considerably lower forward voltage at 3,28V/3A, but that is still higher than the XP-L2 70CRI at 3,20V/3A.
LH351D 70/80CRI = 3,40V/3A
LH351D 90CRI = 3,28V/3A
XP-L HD = 3,47V/3A
XP-L2 = 3,20V/3A
In a FET config, that means the LH351D is between the XP-L2 and XP-L HD in terms of DD brightness potential, but considerably behind the XP-L2 regardless.
HOWEVER, it is the most efficient 3V 3535 LED if put in a regulated buck/boost driver config in terms of light output.
On my second Sofirn Q8, I have made some progress with the stock light. Out of the box it was making 5775.3 lumens on 30Q cells, it now makes 6379.05 lumens. This gain was through a solid copper disc in the tail end replacing the pcb, with Blue’s large springs and a 20ga bypass. While I have flashed Anduril to the driver, no other changes have been made and the top end of the light still hasn’t been opened up.
I also re-threaded the holes that hold the tail spring board down, used 8-32 machine screws 3/4” long…
Added a screw in marine grade bronze heat sink to help the final build once that’s done. 
Couple of looks inside, showing the Blue springs….
The bronze heat sink aids in the hold and also locates the switch, so whether or not it actually helps with thermal management it looks cool and serves other functions so it works for me.
Sorry i expressed it wrong. Because of the thin material of the H03 (a positiv thing for a headlamp because it makes it light) and the more Amps pulling LH351D it would get way to hot. And i would go from XM-L2 to LH351D 90CRI.
Maybe the led would not get into crispy state, but my forehead would for sure.
The Cu disc looks nice.
Wonder how much slop the springs need to take up. Else instead of using springs, use something like Belleville washers instead, whether BeCu for springiness or something else if better.
Then youse could all go crazy with it!
Since the driver has a brass “donut” for contact, the cells kinda have to be button top. Different cell lengths, protected cells, that’s why the springs in the tail are critical. These large heavy duty copper based springs from Blue’s group buy fit the bill nicely. 
I only gained 151 lumens with the bypass.
I’ll go all crazy with it when the replacement emitters arrive. 
Few more low Vf LEDs:
Luxeon V= 3,10V@3A,
Luxeon MZ= 2,93V@3A.
XHP35 is more efficient. E2 does > 2000 Texas Lumens while still being over 100 TLm/W. E4 should be close to 2200. LH351D is 1750.
Well, I now precised that it is the most efficient 3V 3535 LED :+1:
& Luminus SST-40 3.0v @ 3A.
Beam shot Armytek Tiara A1 v2 XM-L2 (White) vs Samsung LH351D 5000k 90CRI
Firefly mode the Samsung LH351D 5000k 90CRI is a tad warmer than stock light
Stock LED on turbo
Samsung LH351D 5000k 90CRI on turbo
GIF
Hey TheOnlyDocc I use 14mm aluminium (i know copper is better but Im yet to find one with 3535 footprint). ” purchase link”:http://kaidomain.com/S008185-14mm-Aluminum-Base-Plate-Heatsink-Board-for-Cree-XP-E-5-pcs?search=14mm%20Aluminum%20Base%20Plate%20Heatsink%20Board%20for%20Cree%20XP-E%20
4x high-CRI 4000K SST-20 from Kaidomain into my formerly Nichia 219c D4S, followed by re-flashing with the full-power Rampingios firmware.
I haven’t had a chance to try it at night since reflashing, but the beam is definitely less floody, which is one of the goals of this mod for. I’m hoping for a some more output, too.
Please report back, as I want to change my BLF Q8 to a SST-20 or a LH351D.
If the output is acceptable, maybe the slightly-higher CRI SST-20 will get a chance.
I consider about 4A ideal for the SST-20 (80% of maximum output), and in the BLF Q8 (which is a low-resistent direct drive flashlight) they will certainly draw much more current. Do not do spring bypasses, and you may want to use thin ledwires (22 AWG?).
Pretty much what I expected. Thanks for confirming!
Matryoshka :-) 
Cree XP-L HD V6 3D emitters on 20mm noctigon MCPCB
MTN 17mm driver
XP-L HD Modified 10507 Carclo optic
ARC glass
Aluminium & brass heatsink
Brass bezel & switch inserts
Forward clicky tail switch
Tritium vial in switch
Magnetic tail
Aluminium Jacket with CF inserts
Length – 78mm
Width – 25mm
Weight – 108g (with 18350 cell)
Max – 2320lm
Two piece aluminium casing is held on by friction. Simply pull it apart to reveal the inner core carbon fibre & brass flashlight.
Dyneema wound mid section.
Length – 73mm
Width – 22mm
Weight – 80g (with 18350 cell)
Max – 2320lm
Sibs.
.
This light was previously 18650 compatible but I shortened it and made the new aluminium casing.
Amazing idea and a stunning execution :+1:
Swapped a 5600K Yuji led into an old Led Lenser keychain light and it became attractive again, so nice even that it made my keychain again after 6 years of absence.
Output on 4xAG3 button cells: 12 lumen
CCT of the spot: 5070K, on BBL
CRI: 95.3
