Using a voltage regulator to control an LED?

I'm planning a bike light project, part of which is a rear light using a red Cree (2-3Vf), and I'd like to drive it just by regulating the voltage going to it. I know LEDs are supposed to be driven by current regulation, but I think that it can be OK to drive them with voltage regulation. With some experimentation, I've found that I can direct-drive strings of LEDs in series with varying numbers of different types of batteries, and based on the voltage I apply, I get varying current draw (and thus output).

I'll be using a battery pack of 6 Li-Ions in series, for a battery voltage range of 18-25V, and an adjustable switching buck regulator, so I'd like to just regulate the voltage to somewhere that my meter says is a reasonable amount of current draw (and a satisfactory light output) when direct-driving the LED, and leave it at that.

Are there any problems with this setup, or anything else I should be aware of? How about regulating the voltage to within a "close" level and then using a series resistor to trim it slightly? I don't know of any other cheap way to drive 1 red Cree from the relatively high battery voltage I'll be using - any other suggestions on that?

Alex

I'll assume you will want to drive it at specifications like 350mA or possibly more.

Given that statement the Vf with temperature will be a problem. Vf drops as the LED heats up. To compensate for this you can burn a little power with a series resistor and that will help minimize the Vf LED variations with temperature. If you can afford to drop 0.5V or 1V across the resistor that will help stabilize the circuit. You can get by with as little as a 0.1V drop across the resistor, but, the ratio of resistor voltage to Vf means that the LED Vf variation will dominate the LED current.

You might also want to look around for a buck LED driver like the Downboy, SOB or drivers from dealextreme and elsewhere. Then you can get true CC regulation and not worry about Vf variations with temperature.

Yeah, temperature. That's the vague warning I'd forgotten. Does Vf go up with temperature, then, thus increasing the output as it heats up? I can see how that can be a real danger if left unchecked with inadequate cooling. I'll be mounting this in a copper end cap, mounted on my bicycle, so as long as I'm moving I'll have excellent cooling, but I do want to have a relatively constant output, not a light that gets brighter and dimmer with temperature.

I'm planning to drive the LED at whatever level seems appropriately bright for a super bike taillight (with an elliptical optic). The LED is rated for 82lm @ 700mA, but I expect I'll have sufficient brightness somewhere between 200-400mA. I did some calculations, and with a nominal Vf of 2.4V and a Vin of 3V, a little 2-ohm resistor will give me 300mA and burn 200mW, which is an acceptable loss. If I wanted to have the resistor do more work and thus keep the circuit more stable, 4Vin with a 5ohm resistor will give 300mA also, burning 500mA, but that's almost as much as the LED will be using at that level.

I had dismissed all the LED drivers I've seen because the ones that can take higher Vin (like the Downboy, SOB, CC1W, etc) are too expensive (my buck regulator cost $1.50, so $15 for an LED driver seems more than I need) and the DX cheapos can't take more than 4-9Vin. But I hadn't thought of using them after my buck regulator, thus getting fairly efficient current-based LED driving for cheap. This one looks like it would work, although I'd have to make sure to jump quickly through the high mode since the LED has a maximum continuous current rating of 700mA, not 1000mA. Anybody know of a way of "turning down" the max output of a cheapo driver like that? It would be nice to use current-based regulating and to have the various levels available.

Alex

Temperature bites you because as the LED heats up the Vf goes down and with a constant voltage the current goes which in turns makes more heat which turns... And so the vicious runaway condition occurs.

Not driving the LED very hard is less problematic and under 350mA the thermal runaway condition is not something you need to worry about too much. It's still there and depending on how you mount the LED and how well it is thermally managed will also determine how much the temperature on the LED will change. The larger the heat sink the less problem and of course the smaller the heat sink the more it's an issue.


You have the opportunity to use switches and resistors and or change the voltage to change the LED brightness level if you so desire and a voltage buck regulator is probably the way to go.

I would recommend dropping 1V or so across the resistor. The resistor is then setting the LED current by the formula ( Vout - Vf ) / R = Current

If you are a math wizard you can evaluate this equation and determine the effect of variations of Vf using differential equations or just plug this into an excel spreadsheet and plot the Vf range you think you will see and look at how much the LED current will change. You can then dork around with more or less resistance to optimize the trade off of power loss in the resistor vs LED current over the temperature range swing. The Vf change with temperature on the die can be found in the datasheet for the LED and for white luxeon the number is something like -2mV/C which means increasing the die temperate of 1C the Vf drops 2mV. This is not the ambient temperate. This is the die temperature and even if the heat sink is at ambient the die can still be higher than ambient.

A rough estimate would be assuming 1W of power. That makes calculations easier and is more current that you will be using, but, still a viable calculations. Assume the die to slug temperature number is 10C/W and you have glued the LED to a heat sink and that glue joint is another 3C/W and that the heat sink is perfect and remains at ambient all the time. That's 13C/W and since we are using 1W of power equates to 13C change over ambient.

At 25C the die would increase 13C or on a white LED some negative 26mV.

That's no biggie. You can plug that into the excel and equate that.

Now, let's look at the worse case. On the worse possible hot day it's 50C. That's another 25C that needs to be factored in. 25 + 13 = 76mV.

That should give you an idea of how to look at the effect of the series resistor and the change in LED current over temperature. You'll need to use the actual tempco of the actual LED and the die to slug and other numbers from the datasheet.

Another way to look at this is 76mV in this example over a 1V drop on the resistor is less than a 10% change. So, if the LED current was 300mA the maximum change with temperature would be an increase of 10% on a hot day.

Hope this helps.

Wayne

Yes, that's extremely helpful. Vf going down as temp goes up makes more sense now that you explain it. And I do see that number in the datasheet which I didn't know how to use before; it's -3.2 to -3.0 mV/C.

The thermal conditions for the LED as I'll have it mounted are close to ideal: the slug is soldered to a star MCPCB which is thermal-greased and pressed into the flat back end of a copper pipe end cap, which is mounted facing backwards on the bike rack, so the slug will be facing into the wind. There'll be considerable airflow directly on the part of the cap with the shortest thermal path to the die, and the cap is big enough to provide a bit of thermal mass as well as being a heatsink. So, as you say, running a 700mA part around 300mA in that kind of environment should result in very small changes in die temperature, at least while I'm moving.

Given your quick calculations, I don't think it's necessary to do the complex spreadsheet you mention to find exact numbers for current draw as I think your 10% change sounds like a reasonable amount. Also, given the thermal environment and the low power levels, I'll have considerable headroom for thermal runaway before anything bad starts happening.

Still, I think I'll run a little resistor in there just to have some safety. Using the wonderful LED Pro program, I can calculate various resistances, and using a variable resistor and my meter in real life tests, I can come up with some nicely acceptable value. I'll be able to set the adjustable buck regulator to output any voltage I want, so I can set it higher and use a larger series resistor for more safety, or set it lower and use a smaller series resistor for more efficiency. I think some experimentation will provide some good results.

It'll be at least 2 weeks before I have all the parts collected and can start building, but if anyone's interested, I'll post my progress and questions over in the bicycle forum. Wayne, thanks a lot for your excellent help!

Alex