The efficiency can drop, but not from 90-95% to 50%. I think something would catch fire if that happened.
I do think the efficiency does plummet to around 70-75% or below when modded.
The driver gets ridiculously hot, and the current draw from the cells seem excessive.
From the video, some quick calculations:
Emitter power: 6.0-6.5A x 4.0-4.2V = 24 - 27.3W
Input power: @0:33 3.65A x 10.0V = 36.5W
Efficiency: (24 to 27.3)/36.5W = 65.8 - 74.8%
Oh boy I know this too well… this is what fried my TK61 driver running only 5.75A. Getting a new 10A driver from Richard but only pushing it to 7.5A.
I understand that as voltage sag increases (or batteries deplete) the amps will go up to compensate [watts = volts * amps, and the power wants to go as high as possible limited only by driver or power input], but I still don’t understand why efficiency would drop that low.
EDIT:
I guess as supercooling decreases resistance in wires considerably I guess the opposite is very well possible. That makes a lot of sense now.
I remember from year 11 physics it has something to do with the atoms vibrating and flying in all directions causing collisions with each other within the wire, and by cooling them down they flow in a straight line. It would make sense that by increasing the heat energy within the wire it would create more atom collisions resulting in a higher resistance.
Hmm couldn’t hurt to keep it cool. That’s what they do in CPU overclocking - liquid nitrogen cooling but purely for setting records.
Sorry about the TK61. The driver does seem very finicky compared to the K40/K50/TN31.
The heating effect may be the cause of the decrease in efficiency.
If the driver goes from 25C to 125C, the resistance of all copper components will increase by 39.3%. This goes on a positive feedback loop as the driver tries to pull more and more current from the cells.
The components on the driver were chosen to work at the original output level with a bit of safety margin thrown in, no doubt none of it was designed to still work like stock after the output's been doubled.
The K50 with the higher 4S input voltage to start with will have more room to grow than the other lights with only 3S.
Just thinking out loud here…
Now how do we cool down the driver and wires… The only thing I can think of is using larger wires for less resistance and then potting and filling up the driver cavity completely with a good conductive material. It has to be easily removable otherwise you’d have to get a new driver every time you want to make a change. Of course the only pro to this is extended battery life by potentially 20 if it even works, but I doubt it.
Well I jumped in at the very last minute. I had no intentions of even buying the light but out of pure curiosity I sent OL a PM to see what the price was, next thing I knew I was on the list and had been sent payment information haha so naturally the easiest thing to do was buy it!
Anyway, I haven’t seen one of these pulled apart. Are the drivers similar looking to the TN31, K40, TN30 etc? Meaning we can just do the normal replacing of the sense resistors with solder and upgrading of the other bits and pieces?
Which driver did you got from RMM?
The TN31 and K40 have identical driver layouts. The TN30 has a nominal output voltage of 9.0V for the three XM-L2’s in series. Not so sure about the K50, I’m curious as to how far we can push it. I do have a gut feeling that bridging the sense resistors will be too much for the LED. (> 9.0A)
He hasn’t put it on the site yet, but it’s a new driver. He’s putting it in for me 
Christmas came earlier…
Not sure if this is already common knowledge or not. But I emailed Bella and asked about whether the light was glued together. She said it isn’t.
Lovely. Thanks for the info, I was getting curious.
I also concern about the focusing of the beam after dedoming the LED. Until now there is no any specific method or guideline to do this yet AFAIK, and you have to keep doing it on the trial-n-error basis and by judging the beam pattern on the wall or by lux meter reading.
Pretty much trial and error. I’m not sure if I’m going to physically modify the reflector but Vinh has suggested that the reflector base needs to be push outwards to achieve a good focus. (see OP)
I read somewhere that if there is a donut in the beam when you shine the light close to something its not focused properly. This happened when I modded my TN31. So I sanded down the centering ring until it was extremely thin but was just enough to locate the reflector still. The beam only has a donut with my hand right in front of the light now. Seems to throw pretty well haha
Have no idea if that is actually how it is done.
Do you have lux measurements of before and after? Most big throwers have donuts until a few meters away. Look at beamshots with defined beams, you can see the outside of the beam is much more intense than the middle part (which would create a donut shining on nearby things).
Definitely black holes are part of the focusing process. You should get rid of the black hole/donut under 1 meter or so. My latest Shocker (best performer) nailed the donut test, like under 10-12 inches or so -- best result I ever got on a de-domed Shocker. Usually when you get a black hole, you have to get the LED more into the reflector.
Slewflash, I’m talking about an actual black spot inside the hot spot on a TN31 beam when shining in a lower mode at a white wall. Unfortunately I don’t have any measuring equipment so no official results. But after sanding down the centering ring as much as possible I only got a black hole starting to form at 18cm ~7 inches when shining on a low mode at a white wall.