I am attempting my first mod on a cheap, single emitter lantern from Amazon before I mod my S2+. I’d rather screw this up and learn on the lantern than the flashlight. I’m looking for max runtime over lumens. My question is, in theory, with everything else being equal, would a 7135x8 linear driver at 50% have the same output/runtime as a 7135x4 at 100%? Would there be any advantage using one over the other? (Besides obviously the 7135x8 would be much brighter, shorter runtimes at higher modes, etc.) I guess I’m asking about efficiency, heat, stuff like that. Or if you have any other suggestions for a driver for a lantern. Thanks!
If a 7135×4 driver at 100% will give you enough light output, and you don’t feel the need to have the extra output of a 7135×8 driver at 100% in reserve, then I’d use the 7135×4 driver. 7135×3 drivers are available to buy as well; if one of those would be enough, then that’s what I’d use.
LEDs tend to be less efficient at higher current. 7135 drivers run at a given current and pulse that current to get lower output levels.
A standard 7135×4 driver at 100% will run constantly at 1400mA. The 7135×8 at 50% will pulse a current of 2800mA to be on half the time and off half the time (the technical term is “50% duty cycle”) for an average of 1400mA. Pulsing a higher current is less efficient than running an otherwise-equivalent lower current constantly.
For example, a 7135x4 driving an XM-L2 U2-1A LED at 100% (1400mA) will produce about 620 LED lumens. A 7135×8 running the same LED at 100% (2800mA) will produce about 1060 LED lumens, so you’d expect 50% to give you about 530 LED lumens - a loss of nearly 15% compared to the 7135×4 driver at 100%.
There are other factors in play, though. For example, LED output drops as they get hotter. That’s a big part of the reason that 2800mA doesn’t just give you twice the output of 1400mA.
However, running 2800mA at a 50% duty cycle allows the LED to run cooler, so some of that loss won’t happen after all, and your light output will actually be somewhere between 530 and 620 LED lumens. It’s hard to predict where it’ll land in practice, but whatever happens, the 7135×4 will still produce a bit more light for the same power.
This is exactly the information I was seeking. Thank you so much for this detailed answer.
If you’re modding and you want the most efficiency possible while also keeping a wide range of outputs available, look at Mike C’s driver(s) which use multiple channels to feed different banks of 7135s for different output levels. Only one 7135 ever gets PWM’d and the rest are grouped so that you get modes that use an exact number of 7135s for each level. It’s hard to describe in such general terms. You should just look at his thread and see for yourself how he did it.
Thanks, DavidEF, for the additional info. I will check them out.
Mike C’s drivers are very elegant, and optimised for great efficiency at whatever setting, but he designs them for his personal applications, I don’t think he sells them, and I don’t think shares his firmware much.
I think you would be better using a generic e.g. Nanjg 105 driver and choosing only sufficient 7135s to run your LED at peak efficiency at maximum level.
Choose a firmware that suits you, start off-the-shelf, and learn about flashing others etc. only if you need to.
Study djozz’s LED tests and the transfer characteristics of the LEDs, i.e. BLF temptation is to push them up to maximum current levels, where perhaps 50% more current only gives 10% more output, and require exotic DTP copper MCPCBs and other good thermal management.
For an efficient and usable lantern this should not be necessary, I’d guess that a driver with only 3 or 4 7135s will be sufficient. By all means buy a x8 7135 driver, but these are quite powerful and get hot quickly. As long as you know how to de-solder the chips until you find the right number to set the maximum LED current to an efficient maximum level, without PWM.
A properly balanced driver, with only sufficient 7135s to run the LED at peak efficiency, will always be much better than stuffing more on, then PWMing them to turn it down.
AFAIK pThe moonlight special uses 2 channels.
Okay, Mike C drivers aside, it’s still going to be true that a multi-channel driver offers the best efficiency in most (if not all ) modes. The more channels, the better. Avoiding PWM is what you want to do, as Phlogiston explained already. All I was trying to do is add that you don’t have to give up higher output necessarily, if you can get a driver that has middle modes that are efficient. Single-channel, and even dual-channel PWM drivers generally don’t give you that option. But multi-channel drivers can. AFAIK, the Mike C drivers have the most available channels, so you get the most available lighting levels to choose from that are as efficient as possible (no PWM). That’s why I mentioned those. There are other multi-channel drivers around as well.
Multiple channels are not necessary if the LED is only ever driven by them at it’s peak efficiency current, not over-driven.
Lower levels can be fast-PWMd.
Moonlight or firefly levels, if required, may need a separate channel, which might be a single 7135, or just a series resistor from maybe a 20 mA MCU output pin, which can be PWMd too (probably better results than trying to PWM a 350 mA 7135 of unknown provenance).
If you look at each review you can find a graph of the efficiency.
https://lygte-info.dk/info/indexLedDrivers%20UK.html
Avoid PWM, get a constant current driver, and then put many LEDs in parallel so that each LED only draws a fraction of the driver’s current.
This will give the LEDs much higher efficiency for the power they use.
For the best efficiency you will want the LEDs to get about a hundred mA each: Using power LED at low current levels | Candle Power Flashlight Forum
Tom Tom, it is specifically the PWM of the driver that ‘eats’ efficiency. Read Phlogiston’s post.
You gave me an interesting question to think about, so I’m happy to know that you found my answer helpful. Good luck with the lantern build 
I read it, but disagree.
If the LED (s) are operated at an efficient operating point (current, and the according forward voltage), then PWMing them for lower brightness will have absolutely no reduction of efficiency (lumens/watt).
Neither, as has been suggested, will PWMing them at higher, less efficient, current levels somehow make them run cooler, far from it.
Real world simple linear current-mode drivers such as a few 7135s burn off the extra voltage as heat, which works very well in practice, from a single Liion cell.
Take a look at data such as for the Cree XPG3, at http://www.cree.com/led-components/media/documents/dsXPG3.pdf
Specifically pages 15 and 16. These show the transfer characteristics (forward voltage vs. current, and current vs. luminous flux).
You have to combine these two graphs to get an efficiency curve of relative luminous flux per watt of input power. But that is not achievable except with e.g. a buck driver, which introduces other inefficiencies in the more complicated circuit.
My advice is to select an LED, and a simple fixed current linear driver, set to give desired maximum output, and PWM them for lower levels.
Variable current linear drivers are rare. Efficient variable buck switching drivers also. I don’t think these more sophisticated drivers are necessary for a lantern application, where a decent match of LED to e.g. single channel xN 7135 driver is an economical and sufficiently efficient solution.
No, you missed the point about PWM. A constant current of X will always result in greater efficiency (more light output) at the LED than a 50% PWM at a current of 2X. It is because the LED itself is more efficient at lower current. This is not something you get to just “disagree” with. It is a known, documented, tested, proven fact. Take a look at the CREE PCT and compare any LED at any current value to the same LED at a lower current drive. It shows clearly that there is greater efficiency at lower current.
TA also shows the lm/W for all his LED tests:
One LED at 1A will be more efficient than having that LED run at 2A for 50% of the time.
Also why I mentioned earlier that having multiple LEDs at a lower current each is also beneficial.
Ten LEDs at 0.1A each will produce more lumens (and therefore be more efficient) than one LED at 1A.
Isn’t that what I said ? Set things up for efficient current and lumens/watt at the highest level that you want to use.
PWM it down below that. Or if you can wind down a linear driver, or buck, by all means.
Unless you you have some mysterious buck driver that can ramp up and down maintaining peak efficiency at all levels (where can I buy one ?)
Taken to the limit, with your advice, we’d still be buying those 100 LED torches, which are not good.
A real torch needs only one LED, or a few, in a decent reflector. All this theoretical talk about efficiency is nonsense, for any practical thing, never-mind a simple lantern. The basics are obvious, but seem not to be well understood. Never mind by anyone who has actually used and compared a few. Peak lumens figures with full cells are one thing, practical run times and overall output (integrated over time) are quite different.
I’m out of this now.
Good background information, Enderman
Nicely put :+1:
That’s entirely true. In fact, the reduction in average LED temperature that follows from the reduction in average power dissipation can even allow slightly greater efficiency at the lower brightness.
That’s not quite the point I was making. My point was merely that 50% PWM of 2800mA would run cooler than constant operation at 2800mA, which needed to be taken into account when properly comparing the options GreenLights wanted to choose between.
We all agree that 50% PWM of 2800mA will certainly run hotter than a constant 1400mA.
I would love to see a real life test comparing efficiency for constant operation at 2800mA, 50% PWM of 2800mA and constant operation at 1400mA, just to know where the 50% PWM of 2800mA would actually land between the other two.
We arrive at the same conclusion 
That’s true for a flashlight where you want to send a beam in a given direction.
For lanterns, however, using many LEDs to obtain greater efficiency at lower power levels per LED can make a great deal of sense. It can even be an advantage in obtaining the wider, more even light distribution that a lantern calls for, because you can orient the multiple LEDs in different directions instead of reflecting and / or diffusing the light from a single LED into all the directions you want to cover.
It’s actually more subtle, yes, in isolation, the LEDs deliver their best efficiency (lumen/Watt) at lower current levels, primarily because they have lower Vf at lower currents. Light output vs. current is pretty linear until severely over-driving them (as we sometimes do).
However when driven from a Liion cell through a linear driver, the benefit of the lower Vf is pretty much irrelevant, the surplus volts from the cell are just burned off as heat in the driver.
Take a look at the curves in the datasheet I referenced at http://www.cree.com/led-components/media/documents/dsXPG3.pdf ,pages 15 and 16.
Then take two scenarios:
A) LEDs run at their normalised (100%) “relative luminous flux” i.e. 375 mA, Vf 2.725V (my interpolations)
B) LEDs run at 400% flux, i.e. 1800 mA, Vf 3.03V
Assume cell voltage of 3.8V
Now, forget about these Vfs, because in reality all that matters is the cell voltage, which gets burned off either in the LED, or the driver.
So in A) we get 100% flux from (0.375 x 3.8) = 1.425 Watts.
And in B) we get 400% flux from (1.8 x 3.8) = 6.84 Watts.
So yes, with the LED at the 100% point the system is (1.8 / 0.375 / 4) = 20% more efficient than the LED run at 400. But in absolute terms it is hardly an efficient system, (3.8 –2.725)/3.8 = 28 of the cell energy is wasted as heat in the driver.
And this is a gross simplification, the cell voltage will vary from say 4.2V to 3.5V as it discharges. Counter-intuitively the efficiency increases as the cell discharges.
If you want the extra brightness from the 400% LED, or just prefer the simpler packaging, and cost, instead of fitting four LEDs run at 100%, the efficiency hit may be worthwhile.
This is why an efficient buck driver is so desirable for this sort of application, where the excess voltage of the cell compared to the LED is not wasted. However I’m not sure where best to get one.
If you’re looking for the best efficiency, you might want to consider using a boost driver, along with an LED with a high forward voltage (like an XHP35). That way, you don’t end up throwing away excess voltage as heat, like a 7135 does. Though I suppose 7135’s are a lot more simple, and are usually good enough when it comes to efficiency.
You can “wind down” a linear driver by simply having fewer 7135s on it.
This will give you less current without reducing the efficiency.
If you built a proper 100 led flashlight, not a cheap potato from china, you would definitely get much higher efficiency.
The reason those flashlights are not good has nothing to do with how many LEDs they have, it is simply the bad quality of the entire light.
Having a constant current driver power many LEDs at a low current each will give more lumens than having the same driver power a single LED with all that current.
The only down side of this is that your light is coming from multiple sources, so it will give less throw compared to a single LED and optic.
In case you didn’t notice, OP’s entire topic is about a lantern, not a thrower, so the fact that the light is coming from many LEDs instead of just one is not an issue.
In fact, it probably is better for this application because you can get more uniform light coverage by placing all the LEDs around the lantern.
Since the goal of higher efficiency in this case is likely longer run time rather than more lumens for the same amount of power, then OP should use a driver with fewer 7135s to deliver lower current.
Then, having multiple LEDs in parallel at that lower current will make each LED run at a higher efficiency, providing the same amount of lumens as a single LED but while using less power.