What is the absolute most efficient driver for low currents?
I want to drive a single N219C with a single 18650. I want about 100 lumens, and the most important thing is runtime. Hence the most efficient driver question.
Probably a single 7135.
In theory you can make the most power-stingy converter with a buck converter, but in practice, copper losses, iron losses, circuitry losses, etc., all add up.
Plus, buck converters have to adjust to everything from right out of the charger (4.2V or so) to just about the Vf of the LED, whereas linear regs just drop enough voltage to maintain the specced current through the LED.
A fleapower buck converter which is limited to, say, 300mA-400mA or so, can be designed to use a smaller inductor, lighter-weight switching elements, etc., and “tuned” to low-current apps, increasing efficiency.
But, you gotta find/design such a critter.
One of the members here, HKJ, has tested a quite a few drivers and posted his results here.
You should be able to learn a quite a lot by looking up his very detailed technical reviews.
I have wondered this before… I did some research and built some lights that I am quite happy with…
I ended up buying a batch of single mode 1.5a drivers, and resister modding them down to the output I want. It turns out the drivers were being over driven anyway, so the power I wanted from them was probably much closer or even inside their efficiency window.
These were and are some of my favorite lights. My goal was 200 lumens for 8 hours. I dont have an IS, but they compare well to other 200 lumen lights, and they definitely run for 8 hours…
I dont have the exact driver I went with, but it looked similar to this. I remember it didn’t have a chip which looked like an MCU like on the atiny boards…
http://www.kaidomain.com/p/S010093.Cree-XM-L-T6-1-Mode-Circuit-Board-5pcs
http://www.kaidomain.com/p/S008846.A9-Buck-Current-Regulated-3_0V-8_4V-1_3A-Flashlight-PCB
This driver in HKJ’s review looks familiar, maybe this is what I used:
219c has a low Vf curve and we’re talking about a point laying low on it. Linear won’t do great…
all depends from cell capacity curve
My understanding of exactly how Vf works is admittedly incomplete. I thought low Vf would be preferable in a situation like this - for example we always say that LEDs with low Vf will stay in regulation longer. Is that not related to efficiency?
I guess, can you explain further?
Would an N219B work better?
Again, my technical expertise on the Electrical Engineering side is lacking. I thought that linear regulators like 7135s were good to keep the light output Regulated, and compared to just a current limiting resistor they would be considered efficient. But don’t they still produce a fair amount of heat at the chip itself (energy being burnt off as heat = energy lost). Or am I incorrect on that?
Here what I understand / know, correct me if I’m wrong.
Vf is the LED forward voltage, this relates to the current going through the LED. Higher current, higher Vf. This relation is non-linear though. It rises fast on the very low currents and then levels out a bit until you get to the really high currents (weird things happen).
A linear regulator “burns off” extra voltage as heat. Thus a fully charged cell sitting at 4.2V driving a low Vf LED at 1A having a Vf of 2.9V (219C approx) will burn off 1.3V at 1A. Thus you would waste 1.3W of power in the driver.
The upside is that the battery voltage can drop to something like 3.2V or so before the linear regulator does not have enough overhead voltage to keep regulation. A higher Vf needs a higher Vbatt to maintain that margin.
And here is where a well designed buck regulator will win, it will not “burn off” the excess power, instead it will use less current from the cell when the voltage is higher than Vf. Of course, some power is lost in the buck converter, but it will be no where near the loss of a linear driver. (If well designed.)
Yeah, for low Vf LED, efficiency is lost with Li-Ion full-charge voltage. Better to use a LiFePO4 or Li-Po cell for those, since those chemistries have lower full-charge voltage. Maybe even better, if designing the entire thing from scratch, to use a high quality AA boost driver with Eneloop and low Vf emitter?
So I guess we are deep enough in at this point that I should give the purpose for this project…
I’m trying to design an ultralight system for lighting a group (4-6 people) campsite when backpacking for up to 10 days. Every gram counts, and the reason that efficiency counts is because it enables me to carry less battery weight (by extracting more runtime).
The first idea I tried was four 12inch lengths of 12V LED strip lighting (with lengths of hook-up wire between them so they could be spaced around the campsite), powered by 4x18650 and a DC-DC cc/cv buck convertor like such:
The runtime was somewhat disappointing, the overall weight was high, and having all the lights wired to a single battery made light placement restrictive.
My idea for version 2.0 was single LED, with single battery and driver. This would eliminate wires, increase light placement options, and potentially reduce weight too.
For better efficiency you need multi-led (or 4-die led).
Forget about recharhable li-ion. Use primary cells.
Sounds like a neat project, I look forward to see what you come up with…
How does multi-led increase efficiency?
I was interested in DavidEF’s comment about AA cells and boost drivers, especially from a weight perspective. Lithium primaries would work better in cold also.
Check Nichia 219C D260 9050 90+ CRI Emitter output test by Texas_Ace
| 1.0 A | 353 lm | 122.0 lm/W | 353 lm/A |
| 0.5 A | 198 lm | 141.8 lm/W | 396 lm/A |
| 0.25A | 111 lm | 163.2 lm/W | 444 lm/A |
| 0.15A | 72 lm | 178.8 lm/W | 480 lm/A |
If you run multiple leds the current on each one will be smaller, then it will be more efficient.
Not so much. You can check many led tests made by djozz and Texac_Ace and find how much power (current for 7135) is necessary for your target (100lm). Big multi-die leds can have enormous efficiency at tiny current rates, like 200lm/w. For several leds (4) you can take you target (100lm), divide by led quantity (100/4=25) to get each led target (power for buck or boost and current for 7135) and check how total power looks in comparation with single 100lm led config.
I dont now how to make a build with this parts but probably 2x rcr cells with 30mm L6 driver (was tested by Jensen567) powering xhp70.2 (probably target output can be limited by res mod) will give you best hour/weight ratio from parts that can be easily found.
Exactly right. But there’s a tradeoff.
The 7135 will saturate at 0.1V, so if at the rated current (each chip’s share being 350mA), if the Vf of the LED is 2.9V, then it’ll stay in regulation until the cell drops to 3.0V. But right out of the charger when the cell’s at 4.2V, it’ll be dissipating 350mA•1.3V. So that can be quite a bit.
Thing is, buck drivers have their own losses, especially if designed to deliver multiple amps (lower currents being PWMed high-current spikes). So there are copper losses in the wiring (even in the inductor), iron losses (magnetic hysteresis), silicon losses (all the circuitry involved), current through sense resistors, in the feedback loop, etc.
So for higher currents, always running in “turbo” (100%), the buck driver will be more efficient, as wasting 1W of power when dumping 10W, 15W, or more into the LED isn’t all that much. But at lower currents, it’s like using a 600hp Corvette engine in a riding lawnmower. Yeah, it’ll work, but it’ll hardly be efficient.
Now, if you design a buck regulator to be as power-stingy as possible, knowing you won’t be dumping 15W into the load but only 1W or so, scaling down the inductor, “tuning” everything to be as efficient as possible, no added baggage like a µC to do dozens of modes, etc., you can make the buck reg pretty efficient. That brings down the tradeoff point.
But even so, by the time the cell drops to, say, 3.5V, the linear reg will still be more efficient. By 3.2V or so, no contest, the linear reg will always always always be more efficient.
So you gotta figure efficiency (and the tradeoff point) over the whole battery discharge curve, not just the worse-case condition (biggest voltage difference).
Why didn’t you say so from the beginning? 
LFP (LiFePO4) cell, crowbarred right across a higher-Vf LED.
Get a nice NW XP-G/-G2 LED and a bunch of LFP cells. Your “regulator” is the power switch. Maybe a small value resistor to slightly limit the current (“low mode”).
Those solar garden-lights use that. Simple, stoopit, and it just works.
Ah, yes I see now. I did know that LEDs were more efficient at lower currents. So for example (fake numbers) four dies could produce 100lm at 0.04A/ea (0.16A total) and is more efficient than one die at 0.2A making the same 100lm. Makes sense. I had not considered XHP70.2, good idea (although more expensive, but it seems higher performance/lighter weight always is more expensive, lol.)
Is the L6 driver known to be efficient, or is it just the first one to come to mind? Would it even work with two CR123a primaries (6.0V)?