Keep that train of thought going. It looks like you almost got there, but not quite. I think your post ends up more misleading than enlightening.
If 255 is always the maximum, then 50% of that is always 50% of that. Same thing applies to 2% (~5/255). We can call it a percentage or a fraction, but they mean the same thing. A percentage of the available whole, where the available whole is whatever the battery can currently deliver.
And the engineer went on walkabout.
I see where you’re coming from though. This constant variable is one of the reasons the whole “lumens” race is so out of place. The standard is to measure lumens at 30 seconds, and for large cells that is fairly reasonable. There is a climb when the light first comes on then it finds stability for just a moment, the lightbox shows some funny things. Constant variation. Voltage sag, thermal sag, Vf rise, driver inefficiency, all kinds of things affect it and it’s different in every light, from cell to cell.
But yes, I see the thought process, even if it’s flawed. 255 is 100, but 127.5 is not necessarily 50. It’s not a linear curve and it’s constantly variable due to all those issues. If you want to set 3 modes, Level 3 being 100% and each level below that a halving of intensity. It might stand to reason that you could set it up at 255, 127.5, 63.25 (assuming it were actually possible to use fractional settings) It might appear by the numbers to be an equal halving of power, or a doubling as you go up from the off position towards 100. But it doesn’t work that way. Try it and see. The lightbox will show you that it’s not a linear scale. You might set the UI to be 25 of the whole, but the actual output will not follow that logic. Inefficiency of the PWM cycle, driver, whatever the reason.
It COULD happen that 127 might at some test point give you 50% of the full output, but it would be a point on a curve, not reliably duplicable. And this is why book sense is not always or even seldom good common sense. The devil is in the details.
And it just makes it all the more confusing for people like me that don’t have the background in electronics, no base understanding as it were as to why any of this works.
50% duty cycle. Not 50% output.
Is duty cycle “on” 50% of the time? It’s all so misleading and confusing. lol
So this new current regulated LD1 driver is much much better because it’s not using PWM, but instead actually using a lower current to reduce output, right? So the emitter is seeing just the lower current, not a cycle time of full power, thereby much more efficient at lower levels. Right?
Which is, of course, a whole nother can of worms.
So why would medium, at a 39 of 255 PWM, have given ToyKeeper the best efficiency?
That’s all correct.
Good question. There are probably a lot of factors. There may be something inductive happening, smoothing drive current a little. I really don’t know the answer though.
I guess this all wouldn’t be so much fun if it were easier, huh? 
A sword with two edges you say? :-p
Someone sharpened the handle on mine!
So on time, duty cycle, and actual photons being emitted are closely related but do vary.
On-time is normally measured from the time current begins to flow to the time the current stops flowing. That is to say whether light is being emitter or not, if current flows, the circuit is “on”. With PWM, there is rise time and fall time. Part of that time, the current is less than required to emit photons but energy is being utilized.
Duty cycle calculates the total energy being consumed in relation to some fixed “maximum”, normally high mode or direct drive. Most of our lights never truly run direct with PWM for all intent and purpose here, whatever current you draw in high mode is considered 100% duty cycle for that particular torch. When the light puts out less lumens, then the current draw is less and can be used to calculate total runtime. This is easy with constant modes but a lot more challenging to maintain efficiency with blinkie modes. You need a way to calculate the average current draw to do the calculations, or just run the light until the torch goes into foldback (the driver protection kicks in).
When does the emitter actually emit photons? The diode part of the LED is active a voltage is presented to it and current flows. But you can have on a few milliamps flow and generate a little warmth, but no detectable light. This is worse in blinkie mode as parts of the rise and fall time will be in this range. If you have a moonlight glow, that means you are dipping into the non-emitting range with each pulse of the PWM.
ToyKeeper’s number above are interesting. Exactly what I would expect. a lot of “dark” current in the low modes. The most efficient modes are near the LED’s intended range. Intended range here I mean heat saturation. If the pull were supercooled, the output lumens wouldn’t begin to sag that much as the output was increased. You can back-calculate the performance of the heatsink with good lumen curves like this as well.
The analogy is much like the stove. It can burn a lot of energy and never “look” hot. But you know you can pump more power into the coil, and it will glow red. Led’s are not much different. They generate heat… and photons as a byproduct.
the convoy triple nichia 219b is impressive!
Lumen measured in my integration sphere :)
Convoy S3 3x XPG2 R4 5A1 5.8a
1414@ turn on
1321@ 30 sec
462 med
59 low
3 firefly
Convoy S3 3x Nichia 219b 5.8a
1401@ turn on
1279@ 30 sec
448 med
57 low
3 firefly
NightSpy;
Some useless info too. The turn-on of an LED is MUCH faster than a incandescent lamp. The 60 Hz PWM of a standard AC line (50 Hz for most of the world) is unnoticeable when driving an incandescent light but should be if driving a LED. I presume that LED bulbs for house lighting have some filter circuitry added to reduce the AC blink. The NHRA had to modify their criteria and timing for what constituted a “red light” start when they shifted from incandescent to LED start lights for drag racing due to the faster response of the LED lights.
I note “High Current” unprotected 18650 batteries are listed for anywhere from the new 10 Amp Panasonics to batteries listed as 25 amps discharge capable without adverse effects. Is there any advantage as far as light brightness in a light such as the RMM full house modified M6 or the RMM S3 to using the 25 amp batteries? I see that the highest current capable batteries have less total watt hours of capacity.
I note from reports that the current draw of the M6 in max mode is circa 4.75 amps per battery and the S3 can pull 5.8 amps from a fresh cell. A light with fully regulated output presumably has a rising current draw as voltage drops too until the voltage drops below the regulation range so a modded TK75 will presumably have a continuously rising current draw.
Any one know of a good discussion on the above that they can link to?
Sorry, no link I can think of at the moment. Linear regulated drivers (7135 based drivers are the popular ones here) will not increase current draw as voltage drops, but buck or boost drivers will. The TK75 you mentioned will have a buck or boost driver, I forget which.
For DD and linear lights especially it pays to look at the actual discharge curves since maintaining a high voltage at the battery is so critical. Take a look at HKJ’s website to compare batteries. http://lygte-info.dk/review/batteries2012/Common18650comparator.php Also note that he’s got comprehensive reviews on there in addition to the comparator.
I don’t have hotrod M6/SRK or any high drain cells, but for 5A draw per cell I’m sure it makes no sense. The 10A Panasonics seem to keep up with IMR cells up to 10A, so they still make the most sense for the application you asked about (I think).
I assume the current function is only true with a boost/buck driver where the total wattage is regulated. If the regulation circuit is only the 7135’s, then the extra voltage is lost across this device and therefore, the current remains constant and the wattage is higher with fresh cells and reduces until the cell was no longer able to maintain the Vf required for the emitter to pull the current. This is true for voltage regulators as well… the more V you put in, the more the device has to work (wattage the device has to cope with). It really amazes me how robust the 7135’s are.
The hot rod M6 is an FET direct drive light. So the better high discharge cells are going to allow the highest output. I believe I have Samsung 25R button tops in mine. This type of light with 4 cells and a big contact board favor the button top cells, I’ve seen flat tops short out on those.
The Panasonics do well, but don’t give the highest output in the M6. Voltage sag under load comes into play with this type of light.
Thanks for the reality check DBCstm. Dunno what I was thinking.
The most output with no other consideration or very very good output with longer run time?
Mine hits 4700+ lumens with the right cells, I believe Richard has seen over 5000 lumens with his. The Panasonics are more sensible, making less total output but also less heat and more run time.
Just found my notes from when I built mine. Samsung 25R button tops gave 4740 OTF, Panasonic PF’s gave 4457, Sanyo UR18650FJ gave 4044. The Pan B’s would be somewhere between those lesser 2 if I’m not mistaken, but with longer run time.
Thanks! I’ve been wondering what those measurements would be for months! 
Also, it looks like the 219B is much better at handling high current than the 219A was. The max output is almost twice as bright as the previously-quoted 750 lumens from a triple.
What battery seeing as how there is now an ongoing discussion here about the effect of different batteries on output?