I’m anxiously awaiting these questions’ answers as well. I fully intend to get a triple S3 (glad I missed the last one) in XP-L running at… whatever amps Richard will sell me.
Something with a 45 second turbo, medium and low would be sufficient.
Waiting on optics; maybe he’s contacted Carclo and has information.
Rich, my M6 has 7 levels that are spaced at 0.1725, 4.45, 24, 110, 438, 1059, and 4740 lumens. With a reversing UI it’s really a very versatile light. Love it!
DBCstm;
Thanks! I knew there was quite a jump from level 1 to 2 but did not realize that it was that large. Just shows what a poor Lux meter the human eye is. Too much automatic compensation for changing light levels! Still a very versatile light UI IMO. What batteries were you using during your measurement and were they fully charged?
Rich, I think this is the UI set I used 0,22,30,40,50,80,140,255. So the second level could easily be brought further down and as far as that goes the lows could all be nudged down for better dispersal in the low end. With 3 emitters 22 is pretty much the very bottom figure that can be reliably used.
I use Samsung 25R button tops in it. And yes, they were at 4.21V fresh off the charger.
I can probably just measure this with a DMM when I get back to a place where I can do so… but does the current typically scale more with the PWM levels or with the lumen output?
I mean, if the levels are 22 (0.17 lm) and 30 (4.45 lm), is the second level 1.36X as much power as the first, or is it 25.5X as much power? Or somewhere in-between, allowing for MCU overhead (perhaps 24X instead of 25.5X)?
I’m just not sure if PWM-based moon levels are incredibly inefficient, or if the PWM levels are merely a poor indicator of the power used.
ToyKeeper, I think it’s not a linear scale so it’d be difficult to break it down by the UI numbers vs actual output. Figuratively, if a setting of 100 produces 1000 lumens, it’s not going to also mean that a setting of 50 produces 500 lumens. And to complicate it further, the capacity and current capability of the cell will also alter the formula to the best of my knowledge.
Richard would probably be the one that could confirm this and enlighten us on how it works.
I’ll just measure it after I get back to my tools. 
Maybe even make a simple firmware which goes through each PWM level slowly enough to measure it, for mapping out the results. But even though the firmware would be easy, the measurement sounds like an awful lot of manual work and it’d be easy to lose track of which level it’s on. It’d be a lot more useful with a data-logging DMM and data-logging lux meter, neither of which I have. And probably a power supply other than a battery, which I also don’t have.
Yeah, this kind of light is a bit difficult to get tail amperage readings on. When I tried going between the battery tube and contact board the most I got was just over 13A, but it’s got to be closer to 18A to make the lumens it makes. Tried going at it from the back side through the spring end of the battery tube but couldn’t get a more reliable reading there either.
If I can find someone to make a joining ring, I’ll join 2 battery tubes together to utilize pairs of cells and run triple MT-G2’s. I have the spare battery tube, no way to cut threads in a joining ring.
+1
Okay, I got some measurements.
On a host with 5x7135 (380mA), or 1900mA total… with a Nichia 219B and a reflector which rides too high (loses like 20% of its output due to that)… I got the following readings:
|. Mode |. Lumens |. PWM |. Current |. Runtime on 3100mAh cell |. Efficiency |
| Moon: |>. 0.14 lm |>. 6 |>. 2.6mA |>. 49.6 days |>. 54 lm / A |
| Low: |>. 7.3 lm |>. 14 |>. 36mA |>. 86 hours |>. 203 lm / A |
| Med: |>. 42 lm |>. 39 |>. 187mA |>. 16.5 hours |>. 225 lm / A |
| High: |>. 155 lm |>. 120 |>. 745mA |>. 4.16 hours |>. 208 lm / A |
| Higher: |>. 342 lm |>. 255 |>. ~1800mA |>. 1.7 hours |>. 190 lm / A |
So, comparing low to moon… PWM of 14 / 6 = 2.333. Lumens of 7.3 / 0.14 = 52. Current of 36mA / 2.6mA = 13.8. Which means the answer is solidly in-between any of the guesses in my earlier post.
The “moon” mode is by far the least efficient. OTOH, 49 days isn’t bad. I think a comparable Zebralight gets about 70 days at a similar level, so I can’t really complain. The other modes though… ZL is getting at least twice as much runtime on every other level. A non-PWM driver would probably help a lot.
Twice as much runtime excluding the “higher” / 255 mode I assume? If the ZL gets twice the runtime in that mode too then the problem is certainly not PWM. If there’s a link to comparable or related ZL data that would be useful to me. I’m not well versed in the ZL products. (read: I know nothing about them)
Right, sorry. I was looking at the spec page for the ZL SC62d:
http://www.zebralight.com/SC62d-High-CRI-Daylight-tint-18650-Flashlight_p_135.html
It was the most similar model I could find. Of course, it also doesn’t have the problem mine has with losing ~20% due to a too-high reflector and another 5% due to diffuser film and probably another 5% due to lower-quality optics. At this power level, I should be getting 450+ lm OTF instead of 340. (I have another 219B with better optics getting 400 lm at 1.5A, so this 1.9A light should get at least 450 lm)
The ZL gets almost twice as much runtime on its highest mode though… even accounting for mine being like 500 emitter lumens and ZL’s being like 350, it’s still more efficient. I think they might do something clever with a boost/buck driver to convert extra volts into more light instead of just letting it burn off as heat.
That is absolutely insane!
Hearing all this talk of triple xpg2’s creating to much heat, well that’s nothing compared to this!
While using a set-up like this in standard temperature and conditions (25 degrees C) how long could this remain in high for before heating up to unbearable temperatures? 15 seconds? By the looks of HJK’s discharge graph the 20r would only be able to sustain that current for 2 minutes at best, down to 3.6v.
My main goal right now is to have the brightest, smallest light one can possibly come up with, even at the expense of practicality and real world usability, and it looks like this is it. I may possibly be ordering parts for a build like this in the coming days
Basic maths you say?
- 2,000 is 1/3 of 6000.
Either something not very basic is in those efficiencies you mention, or you meant 20 mins? Or I missed the tongue in cheek…
Id also like to know how the low modes on the BLD17DD are determined. For example the lowest mode is 2. Is that 2 of whatever the max amp draw is? For example if the driver is giving 11A on turbo than the lowest mode would be 220ma? Or are the low modes pre-determined by a set value?
Thanks for the link.
Can they really attain 3hr at full current? That seems like it would be ~13.44Wh at the LED based on info from djozz. Allowing for 90% efficiency you get 14.93Wh. That’s more than any 18650 holds I think.
Oops. When I wrote the above I thought the flashlight used a Nichia 219B. That could be a major hickup in this comparison. I don’t see an 85CRI part at a glance (probably looking at the wrong datasheet), but I do see an 80CRI part that can supposedly do 300L at 1.0A at 2.86Vf. I suspect that using Luxeon T has as much to do with their runtime figures as a fancy buck/boost driver might. FWIW if Vf is as low as Philips says then ZL would use a straight up Buck driver!
Looking again, I do not think ZL is correct about using a Luxeon T. More likely they use Luxeon TX # L1T2-5085000000000 since that matches the spec they give. Vf is still low, they’re still running at close to 1A (a little over it seems) - an efficient buck circuit should give them the runtime they claim?
Shortly after initial turn-on with your light you should be looking at 80+ efficiency. Efficiency gets higher as Vbat drops w/ the linear driver. So like 90 before an NCR18650B is half empty. Since this comparison is a long ways from apples-to-apples it’s hard to say how they compare. I would expect the linear setup to do well.
Lol, wondered when someone would catch that!
It’s been a rough week, nothing is basic anymore. My on-the-fly math is seriously discombobulated.
But you DO get the picture. 
If I understand correctly, the BLF17DD uses PWM to control its output. So, the 2% should be a 2% duty cycle at full power, meaning it would actually be 2% of the maximum lumens, assuming perfect heat sinking.
We don’t have perfect heat sinking though, especially at 11A, so in reality the 2% value is 2% of something the light won’t actually be able to do —because the 100% mode won’t stay at 100% for more than a split second before it starts to droop. The modes should have linear performance though, until it gets hot enough for thermal sag or battery sag to kick in.
In any case, if the maximum mode is 3500 lumens at start (even just for a split second), the 2% mode should be 70 lumens. A 25% mode would then be 875 lumens. And the 100% mode would start at 3500 lumens then drop like a rock. In a 1x18650 tube light, even 25% of 11A might have pretty noticeable thermal sag. A maximum of 11A is a lot more appropriate in a big host with lots of heat sinking.
If it were a true current-controlled driver like led4power has been making, the performance would not be linear. 2% power would probably produce more like 5% brightness, according to the output curves measured by match and djozz. With true current control, overall efficiency is generally highest on the medium-to-low modes, and the high/turbo modes and moon modes get less and less efficient the closer they get to maximum or minimum power. This effect is also somewhat true with PWM-based lights, but it’s a lot closer to linear than a current-controlled light.
That’s a solid explanation. If you aren’t already familiar with the concepts you may have to take a step back and read it twice, but I think you won’t get a better explanation than that.
EDIT: I guess the only thing not explicitly covered in ToyKeeper’s explanation is that the DD drivers are not regulated and what affect that has on the lower modes. If a fresh cell gives you something akin to 220mA in 2% mode (all made up numbers), a half discharged cell will give you less in 2% mode. The driver is literally doing it’s best to connect the battery directly to the LED 2% of the time, it doesn’t care what results are produced by doing this. So as the battery approaches empty the lowest mode will be considerably dimmer than when the battery was fresh.
I mean, I understand how the PWM works, but we don’t set the driver to work in percentages. We set by 0-255. Percentages are something we RELATE to the result.
Edit: The setting of 2 in the UI will always produce the level of 2 in the 0-255 scope of things. The Percentage will be on a constant fluctuation as the cell drops and output falls accordingly. So thinking of it in terms of percentages is misleading, as the percentage (except 255 or 100) will be on a constant curve. Even that 100 will be dropping, but 255 will always be the maximum setting, whatever is available. This is why it’s difficult to nail down a set number. There isn’t one.