[Frs1661]

I'm looking into DIYing a LED helmet light; thanks for all the great information that is out there!

My question is on the drivers; Basically our goal is to create a constant current source for the LEDs, correct? Everyone seems to use either the buckpuck, fatman, or n/bFlex dirvers. My question is, could we not produce the same result with something much simpler? i.e. an LM317 configured to drive constant current, or a transistor and a MOSFET.

The first answer that comes to mind is efficiency, but if the battery pack voltage was close to the drive voltage of the LEDs, the MOSFET or LM317 wouldn't be dropping much voltage.
Figure 3.75v each cree XR-E, I'd run 3 of them, so 11.25V, plus 1V of overhead for the LM317, I could run a 12v or 14V pack without TOO much loss in the LM317...


Anybody know how efficient the commonly used drivers are?

 

[Frs1661]

This is the sort of thing I had in mind:
http://www.instructables.com/id/Super-simple-high-power-LED-driver/?ALLSTEPS

Basically, the resistor is converting your drive current*1.25v into heat. so driving at 1 amp you're wasting 1.25W.

3x XR-E at 1 amp would consume about 11.25W. Using a 42 W-h pack like this on run time with 100% efficient use of power would be 42W-h/11.25W right? ~3.7 hrs?

With the resistor in there burning up some electricity as heat run time would drop to ~3.3hr. Seems pretty acceptable to me. However I feel like I'm missing something; the LM-317 must consume some power in its regulation duty. I'm not sure how to estimate it's effect.

Am I off base? Any thoughts from someone with more experience?

 

[KevinB]

Efficiencies of around 85% are considered pretty good.

I think you'll need to figure in a dropout (overhead) of 2V for the LM317 to drive a 1A load. (Take a look at http://www.national.com/ds.cgi/LM/LM117.pdf and look at the Dropout Voltage graph.) That bumps you up to 13.25V minimum for Vin. You can, of course, go below this, but you'll no longer be driving your load at 1A and your light will get dimmer as the battery voltage decreases.

I know from experience with using other linear regulators that they get quite hot when there's a large voltage differential between Vin and Vout. Thus, I think you can expect fairly poor efficiency the higher you make Vin. I'd expect your circuit to become more efficient the closer you get to 13.25V. I have no idea what the efficiency would be though.

If you take a look at the LM317's example applications, you'll find that it can be used to build a switching regulator too. Of course, the example given is for a switching voltage regulator and not a switching current regulator, but perhaps the circuit will give you some ideas. I've build a couple of switching (voltage) regulators from LM317s and they turned out quite well. (The LM2575 and the like is even easier to use though...)

FWIW, I've been playing around with building constant current boost converters using a PIC12F683 (for the PWM), an inductor, and an N-channel MOSFET (and a few other components too) to drive a number of discrete LEDs in series. I'm not using these circuits for bike lights yet - they're not bright enough, but it's been a fun and educational project.

 

[KevinB]

I know from experience with using other linear regulators that they get quite hot when there's a large voltage differential between Vin and Vout. Thus, I think you can expect fairly poor efficiency the higher you make Vin. I'd expect your circuit to become more efficient the closer you get to 13.25V. I have no idea what the efficiency would be though.

I was curious enough about this that I decided to breadboard it and take some measurements. I used an LM317T with a 1.2 ohm resistor connected between Vout and ADJ. I used that to drive an 11.3 ohm load (two 4 ohm and one 3.3 ohm wire wound resistors connected in series.)

I took current measurements for input voltages ranging from 9V to 16V, measuring both Iin and Iload. The current readings were identical. I was kind of surprised by this at first, but then remembered that the current measured at any point in a series circuit should be the same, so this makes sense.

Here are my measurements and calculated efficiencies given those measurements:

 Vin   Iload   Iin   Efficiency
----   -----   ---   ----------
 9.0   0.57   0.57   71.6%
 9.5   0.61   0.61   72.6%
10.0   0.65   0.65   73.4%
10.5   0.69   0.69   74.3%
11.0   0.73   0.73   75.0%
11.5   0.77   0.77   75.7%
12.0   0.81   0.81   76.3%
12.5   0.85   0.85   76.8%
13.0   0.88   0.88   76.5%
13.5   0.92   0.92   77.0%
14.0   0.97   0.97   78.3%
14.5   1.00   1.00   77.9%
15.0   1.03   1.03   77.6%
15.5   1.03   1.03   75.1%
16.0   1.03   1.03   72.7%

At 15V the LM317 gets pretty warm and at 16V it gets quite hot. In fact, it appears that it may even be going into a sort of partial thermal shutdown at this voltage. If I leave it running for too long at around 16V, the current levels start to drop a bit. I think you'd need a heat sink for stable operation at or above 16V.

 

[Frs1661]

Thanks for the replies Kevin! I was lacking the high wattage resistors to do the test myself; thanks!

So it looks like best case is about 78%... not terrible, but not great. If I can assume the buckpuck style stuff is ~85% the lm317 might be a decent option. I'll also look into the switching regulator circuit you mention.

Nice that with the Vf of 3 XR-E's the sweet spot lands right near a standard Li-ion pack voltage.

If I give it a shot I'll be sure to write it up.

 

[christexan]

Don't use an LM317T

It is too lossy, for simple, cheap, and the exact same circuit, use an LM108x family regulator (LM1086 for example)... pin compatible and identical format to LM317T, but only around 0.5v dropout (at 1A) or less at lower currents. With this you can get efficiencies closer to 90% for instance 3 Crees at 3.6V at 1A current (10.8V) and an LM1086 at 0.5V for 11.3V consumed, from a 12V source... or you can even reasonably run a LiIon at 11.1V, and still get enough current for many Cree's (my 3-Cree Q2 lights tend to run around 10.5Vf at 1A, although the Q5 bin is reportedly a little higher Vf on average). If you can run the 11.1 LiIon, then your only loss is the regulator at 0.5V. Cost is under $1 at Mouser.com.

 

[Frs1661]

I brought up the 317 because I happen to have acouple laying around; however I'll need to buy some 1W resistors anyway so perhaps I'll be placing an order anyway.

Thanks for the tip

 

[KevinB]

use an LM108x family regulator (LM1086 for example)... pin compatible and identical format to LM317T, but only around 0.5v dropout (at 1A) or less at lower currents.

Thanks for the pointer to the LM1086! I've just looked over the data sheet: http://www.national.com/ds.cgi/LM/LM1086.pdf

My reading of the data sheet indicates a dropout voltage of between 1V and 1.2V for a 1A output. This is certainly superior to that of the LM317T, but not quite as good as the expectations that you set. Do you perhaps know of another part with an even lower dropout?

 

[Frs1661]

A 1V dropout would put it at about 83% efficiency, based on KevinB's numbers for the LM317. That's not too shabby

Anybody got a buckpuck or other driver floating around they'd be willing to do an efficiency test on for us?

 

[ettore]

I use an LM317 in my older lightset; I use a modified relay to switch between two different resistors for "brightness". It also has a 555 timer in there, from what I remember, for a blinking mode but I never really implemented it.

The coolest thing about the LM317 is that it's basically free, and it's small; I glued mine to a small heatsink but it barely gets warm. Efficiencies are the only concern, since you have to have a convenient voltage to make it run efficient (ie: lowest battery voltage needs to be just above 1v more than the LEDs for the given drive).

My NEWER light I got the MaxFlex ... mostly because I house the regulator in a small circuit between my lights and the battery. I can use the same driver with just about any battery and LED combination ... nice and modular.

 

[Frs1661]

What are your thoughts on the Maxflex vs. the LM317?

Apples and oranges?

Also for reference, the sandwich shop lists the eff. of its regulators:

Downboy: 75-85% 500mA max [Step down converter] Here
BadBoy: 80-91% 1A Max [Boost converter] Here
GD1000: 85-93% 1A max [Buck/Boost Converter] Here

So, as long as you choose your battery wisely it looks like linear regulators compare pretty favorably to commercial offerings.

 

[ettore]

It really IS apples to oranges. My "older" lightset uses a 8.4v battery (if I remember correctly) and the two Luxeon III stars running at 900mA (I wanted 1A, stupid crappy resistors) were somewhere in the 7.2v range. So, as convenience happened, the driver is using just over 1v so all was good.

With the MaxFlex, I wanted a slightly nicer interface (button presses and holding) ... plus I have a large array of different setups I want to drive off the same unit (mostly Cree XR-Es in different combinations)... the MaxFlex worked well. My main complaint of the MaxFlex was it's programming; mine works very poorly ... a few modes not being able to be attained (700mA, namely). My LM317 "can't" not work ... so it wins for simplicity.

 

[christexan]

LM108x regulators

I've found they run on the "low" side of their max voltage consumption at a given current, at 1A, it's below 1V in my testing, YMMV, but the spec sheet is worst case and based as far as I can tell on "original formula", as time goes on all semiconductor specs improve (LEDs being the perfect example). A reasonable assumption might be 0.7V but none of the ST Micro LM108x that I've gotten have gone much over 0.5V. Worst case still better than an LM317 and still under $1 (3 times more expensive, but worth it for the extra 1-2V available in the circuit in my book).
I don't use them anymore because even in a "dead-bug" layout, they are roughly the same time as the DX AMC regulators that are only a couple of bucks, and saves me the hassle of assembly except soldering 2 wires.
Have fun!

 

[KevinB]

Searching at Mouser for "LM1086" got me no hits. I was able to find it at Newark, Jameco, and Digikey though. After reading your most recent post, I searched Mouser again, this time using just "1086". That produced a lot of hits; the adjustable version made by STMicroelectronics is LD1086V. The Mouser part number is 511-LD1086V. They're 74 cents a piece for quantities of 10, and only a penny more if you only want one.

I think I'll get some to experiment with...

Thanks again,

Kevin