Achieving high efficiency at low currents

NYCaver

Newly Enlightened
Joined
Oct 27, 2009
Messages
49
Hi all,
This is my first post here. I've been lurking for a while now, searching and reading whatever I can get my hands on that's useful. My intention is to build a headlamp for caving use and I have a few technical requirements:

- Needs to have the ability to drive LED's at both high and quite low levels while remaining as efficient as possible.
- Needs to have separate very wide and more medium beam, preferably independently controlled. This will require a minimum of 2 LED's obviously.
- Single, li-ion battery source of at least 4000mAh and 7V

I have more ideas in mind but that is the general idea.

My main concern at the moment has to do with efficiency of drivers while dimming to low current levels. I was looking at a number of cost effective and well regarded drivers on DealExtreme and Kai until I found out that at least some of them have efficiencies as low as <50% when driving at their low settings. This is simply not acceptable for a light that will probably mostly be powered at low levels. And so I come to you, the knowledgable. Humble and waiting to be enlightened by your extensive knowledge.

From what I have gathered PWM dimming has very poor efficiency at low levels due to the fact that it is driving the LED at it's full rated current, at a certain frequency. Is this the sole cause of the abysmal low level efficiency of the cheaper drivers I'm looking at or is there more to it? Is it simply impossible to dim an LED to low levels (with a driver) and still remain reasonably efficient? From what I've read CC or constant current/current control dimming is much better at low levels. (Newbie's test results have confirmed this in the past)

What drivers out there use this dimming method and why is it so poorly documented whether certain dimming drivers are PWM or otherwise? I noticed that TaskLED only publishes efficiency data for their drivers while in full current mode, not while dimming. This seems to be a big lack of info if you're at all interested in it.

Anyway, thanks for reading and hopefully someone has some good suggestions!
 

ifor powell

Enlightened
Joined
Nov 18, 2007
Messages
230
Location
Bristol UK
I noticed that TaskLED only publishes efficiency data for their drivers while in full current mode, not while dimming. This seems to be a big lack of info if you're at all interested in it.

I just dug out an old spreadsheet with the numbers for an nflex driving 4 leds off a 6 li-ion cell pack. For the upper 4 drive levels on the 1000ma setting the driver efficentcy was > 90% the lowest level was down at 62% but I think this was an artifact of the low quality meter I was using, the curents I have recorded are 0.03 batery and 0.04 led. I need a meter with at least one more digit of acuracy to be able to measure these low curents acuratly. I have a new meter so will atempt to take similar mesurments with my next light.

The issue with PWM dimming is that the led efficentcy is lower than CC dimming. This is different to the driver efficentcy. For your 60% figure if this is driver efficentcy I think you have to factor in the fact that the leds are being driven at a low efficentcy as well.

What realy matters is system efficentcy e.g. light out compared to battery power in but that means measureing the light out. Which for real lumens needs specialst equipment. For a given optic setup it would be possible to just take a lux reading and use that to get relative efficentcys at different drive levels. I may consiuder trying this for my next light, they may be interesting numbers.
 

ifor powell

Enlightened
Joined
Nov 18, 2007
Messages
230
Location
Bristol UK
OK so I got my old nflex driving 4 Xre R2 light out which can be dismanteled to get measurments from and hooked it up to the bench power suply and a load of meters. I had a fan blowing on it as it is meant to be a running / biking light so needs some airflow to keep it cool.

Code:
Level V In I in P in  V out I out P out   Dr % Lux Lux / Watt
1 20.800 0.033 0.695 11.420 0.039 0.445 64.109 31.200 44.910
2 20.800 0.091 1.882 12.180 0.125 1.523 80.881 92.200 48.980
3 20.800 0.170 3.536 12.830 0.259 3.323 93.975 175.000 49.491
4 20.800 0.330 6.864 13.460 0.475 6.394 93.145 279.000 40.647
5 20.800 0.730 15.184 14.230 0.957 13.618 89.687 456.000 30.032

Lux was just taking the brightest reading from across the room somthing like 3 or 4 meaters but at a consistant distance.

So the nflex is less efficent at the very low levels, I belive the microcontroller and other subsiduary stuff is putting in quite an overhead here. The overall efficentcy given by the lux/watt shows I am better off with the 259 ma drive level, the driver is very efficent and the leds efficentcy has not started to fall off a lot. The fact that the high level is only 3/5 the most effiecent level is interesting. I expect the 7 up XPG light I am working on to be better for this. I will try and get the same numbers for it before I seal everything up.
 
Last edited:

georges80

Flashlight Enthusiast
Joined
Oct 23, 2002
Messages
1,262
Location
Sunnyvale, CA
At low output currents (assuming a current regulated versus PWM based LED driver) efficiencies drop due to losses in the switching IC. Most switching IC's maintain a constant frequency and at low current these switching losses become a larger proportion of overall power transfer and so you see an efficiency drop.

Some switcher IC's that are intended to be used in low current applications (voltage regulating switchers) have a feature where they will go into a different 'mode' at low currents to either drop frequency or do a pulse skipping trick to reduce switching losses.

Anyhow, it's a fact of life that if you have a LED driver that can handle 1A outputs it is unlikely to be 90% efficient at low output currents.

You can find some published efficiency curves of switcher IC's and you will see the efficiency peaks around the IC's target market output current. At lower currents (even 50mA) the efficiencies can easily drop to near 50%.

cheers,
george.
 

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
Just out of interest, why are you going for a 2-cell voltage (nominal 7.4V) battery solution?
When you talk about high output levels, what do you mean by high?

Given the low Vf of many LEDs even at fairly high drive currents, you might be able to do a lot with a single-cell-voltage (3.7V) battery, and it would be easy to run fairly efficiently at low currents from that even using simple AMC7135-based drivers.
Also, with a cell voltage close to the Vf, it does potentially open up some doors for adding simple redundancy into your drive circuit.
 

TorchBoy

Flashlight Enthusiast
Joined
Jan 15, 2007
Messages
4,486
Location
New Zealand
A 20 ohm resistor in series with your two LEDs run from 7 V will give you 50 mA and efficiency of about 85%. :p
 

TorchBoy

Flashlight Enthusiast
Joined
Jan 15, 2007
Messages
4,486
Location
New Zealand
XR-E Q5 = 120 lm/W at 50 mA.
XR-E Q5 = 100 lm/W at 250 mA.
XR-E Q5 = 68 lm/W at 1,000 mA.

Driver efficiency... um. That's a bit harder to say. I haven't got much test data at low drive currents. For some generalisations, old drivers with PT4105 do quite well at low current (I've seen a claim from Tobias B of 80% for 30 mA) but drop off above 750 mA, while the replacement AX2002 does very well at high currents, and with 5 V or more very well as low as 300 mA.

Anyway, you would have a trade-off between driver efficiency at either low currents or at high currents with PWM, and LED efficiency. I also suspect a putting a capacitor on a PWM driver's output might not be quite straight forward; even the size of the chosen capacitor might change efficiencies unexpectedly.

FWIW I use a resistor for the lowest two modes in a couple of own headlamps - I figure that tightly regulating such a small current is probably overkill - but that only uses 3x AA cells, so its efficiency is not too bad.
 

gillestugan

Enlightened
Joined
Nov 20, 2007
Messages
242
Location
Nora, Sweden
Hi.
There are extensive information in how to build a caving lamp with your specifications in a thread in the heaslamps section :"replacement of carbide lamps by led". Especcially a lot of information regarding efficiency and finding the optimal reflectors for different caving situations. There is also a building instruction for a very nice lap with three leds with different beam widths designed by Tobias Bosset.
 

NYCaver

Newly Enlightened
Joined
Oct 27, 2009
Messages
49
ifor powell, thanks a lot for the info. Extremely useful stuff!

George, thanks that explains a lot. Out of curiosity, since you said that essentially the higher rated an IC is, the less efficiency it will have at lower currents, does that mean that the bFlex is likely to be more efficient at 50ma than the hipFlex for example? Or do they use the same IC?

UKCaver, my design is not finalized but the 2-cell 7 volt solution was to get close to the voltage of 2 LED's in series. Now that my design is changing I'm looking at an 11.1v 3-cell solution to keep 2 parallel strings of 2 series LED's in constant regulation. 7.4 volts is cutting it close for a buck driver. I'll post more details about my headlamp as I finalize the design.

gillestugan, I've extensively searched these and other forums on info about homemade headlamps and have read Tobias' thread in depth. He has a lot of interesting info particularly in regards to optics that are more useful for us cavers and lower light levels, also useful to us. His circuitry is a little more advanced than I'd planned on coming up with for my light but who knows... I'm still designing.

I'm surprised there isn't more of a community of cave-light builders like there is for flashlights modding and bike/dive lights. I would have thought that the caving community is very much DIY in spirit.
 

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
I'm surprised there isn't more of a community of cave-light builders like there is for flashlights modding and bike/dive lights. I would have thought that the caving community is very much DIY in spirit.
I guess flashlight modders are often coming from a different angle, with serious power and aesthetics being fairly key factors to an extent they may not be for cavers
For cavers, though there are a few people who want to build massively bright lights, the numbers don't seem that great. Maybe partly that's down to there being a downside to having lights *way* brighter than your mates?

As for perfectly machined casings - someone might do that for a caving light if they just enjoyed the pure act of machining, but a light that is going to get used is going to get scraped, and a caver probably wouldn't build a shelf queen.
 

NYCaver

Newly Enlightened
Joined
Oct 27, 2009
Messages
49
Well, for my part I intend to plug the benefits of DIY cave lamps to all of my caving friends, hopefully recruiting new DIY'ers that can make use of the wealth of info here. Headlamps can be so much better than what's commercially available! Even the high end headlamps from Scurion, Sten, etc are lacking in certain areas, and are way too expensive for most.
 

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
I hope I didn't sound too negative before. There's certainly a lot of fun that can be got from building your own light, but it doesn't seem something that too many people do.

I think quite a few people are put off by the electronics side, even though to start off with, it's possible to do a lot without knowing much about electronics, just some simple wiring.

It is really great when people get enthusiastic - earlier this year someone was asking about making a light in an existing mining-lamp headset, so I sent her a little 'starter kit' - a Seoul P4 + optic + driver, and some 5mm LEDs + resistors for a backup beam, and then her group happened to be staying at my club's hut a couple of months later and it was really nice to meet up and see how keen she was.

Only a couple of days ago, she was posting up a list of stuff she'd learned to try and help out someone else who was wondering about trying a homebuild.
 
Top