High Power LED help.
October 23, 2011 8:23 AM   Subscribe

Electronics help. I'm looking for a good informational/tutorial web site for working with high power LEDs 5 Watts and up.

I want to build a UV light source using high power LEDs like these.. I have a pretty good basic knowledge of electronics, and have built LED circuits with ordinary low power LEDs in the past. I'm looking for information about circuits, using drivers, multiple LEDs in a circuit, and heat sinks.

To head off any derails, yes, I know UV light is potentially harmful, so no humans or creatures will be exposed to it. And, yes, I've looked at other alternatives for UV light sources, but LEDs seem like an elegant solution for my application.
posted by Eekacat to Technology (11 answers total) 2 users marked this as a favorite
I researched doing a project with high-power white light LEDs, though I never got around to it, but as I recall I found good information on aquarium sites.
posted by XMLicious at 8:36 AM on October 23, 2011

I've used these (self link), the Phillips Luxeons. You can buy high power drivers these days (e.g.), which might be a fine idea if you need to drive just a few LEDs. In general, the two strategies employed will be pulse-width modulation (varying the duty-cycle of pulsed current source), and constant current supply. Constant-voltage operation isn't recommended because you will always be sinking a lot of power into a limiting resistor.

I built constant current supplies for my 1A/3W LEDs no more complicated than is diagrammed in Horowitz and Hill. Briefly, you place a precision ~0.1 Ohm resistor in series with the LED, compare the voltage dropped across it to a setpoint voltage (a potentiometer connected across a voltage reference) using a differential op-amp. The op amp drives the gate of a power FET (or bank of power FETs) also in series with the LED to complete the feedback. Keep everything on the drain side of the n-FETs (so the source is grounded) so there's no funny business with the gate voltage requirement. I added little MOSFET switches to enable digital on/off control as well.

Heat sinking? The application notes for the LED will have lots of useful information, including calculations for taking a rated heatsink (specified °C/W) and estimating the LED junction temperature, and any required current de-rating. You may face some challenges if you intend to densely pack the LEDs and run them all at maximum power.

The catch is that any heatsink you employ may not get all that hot but is still vitally important... a few Watts just isn't much compared to the power transistors for which they were made. However, you still need great thermal conductance from the LED to ambient because the internal thermal resistance from the LED to the mounting structure is very high. The light emitting part of the LED is just too small, too poorly coupled to the case (compared to a big honking power transistor, say), and surrounded on one side by a nice insulator. So, you can work the thermal equations and find that a heatsink might rise from 20°C to 23°C while the LED junction (the important bit to keep cool) gets to 125°C and beyond. And doubling your heatsink might reduce that 23°C to 21.5°C but barely changes the 125°.
posted by fatllama at 9:30 AM on October 23, 2011 [3 favorites]

What you want to do is go to the web sites of the major vendors for LED drivers and look around. They will provide application notes and reference designs to use their devices.

Texas Instruments
National Semiconductor (now TI)
Linear Tech

These are the biggies in analog devices. There are others if you google "LED driver."
posted by JackFlash at 10:48 AM on October 23, 2011

Analog Devices is also a pretty big name in analog devices :)

One thing a friend ran into a while back is that some high-power LEDs are very fragile when it comes to overcurrent— most LEDs can handle overcurrent for tens or hundreds of milliseconds as long as the average power is OK, but some high power LEDs will fail much faster than that, so you need to be pretty careful about overshoot in your regulator if you're using one of those.
posted by hattifattener at 12:57 PM on October 23, 2011

http://www.candlepowerforums.com/vb/ has a great DIY LED section

https://www.icmag.com (cannabis site) has a LOT of good posts about UV/UV LED and High power LED- unfortunately the search engine is not so good.
posted by plumberonkarst at 2:44 PM on October 23, 2011

fatllama: why not just use a linear regulator? I've set up dozens of high-current LEDs using nothing more than an LM317 and a calibrating resistor.
posted by Mars Saxman at 6:39 PM on October 23, 2011

I've driven a number of high wattage LEDs using this in conjunction with an Arduino.

When you get to the point of designing the circuit, be sure to check this handy calculator out, it gives you resistor values based on the power source and led specifications.
posted by tip120 at 7:57 PM on October 23, 2011

why not just use a linear regulator? I've set up dozens of high-current LEDs using nothing more than an LM317 and a calibrating resistor.

You can use a linear regulator as a current limiter if you really need precision current and don't have a regulated input voltage, but otherwise the regulator is just an expensive resistor and you waste a lot of heat and power in the regulator, particularly for a 5W LED.
posted by JackFlash at 9:26 PM on October 23, 2011

fatllama: why not just use a linear regulator?

It'll work, of course. It's going to waste a lot of power and heat up both the regulator and the resistor. But, say you want to change the brightness with a pot or analog signal, how is that going to work? The pot itself can't really be the limiting resistor... that way lies pain, a hot pot, and dead LEDs. Seems like the only good way (to be able to control the brightness) is with a transistor current control of some sort, so why not put that sucker in feedback and get rid of all the nonlinearity?

Consider also that a constant current supply allows one to string as many LEDs in series as one has the maximum voltage for with no changes in circuitry. Because each LED sees the same current, they will be emit the same brightness (the consistency between LEDs is higher w/r/t current than voltage).

Also, in this series circuit topology, you can turn individual LEDs off by having MOSFET switches "short out" individual lights... in constant current mode, none of the other lights or supply is really affected until they're all off, in which case you should "open" a MOSFET switch in series with the whole string so as to not short out the supply. You can't play this trick with a constant voltage and fixed limiting resistor.

Don't get me wrong, I've lit LEDs (and powered relay coils, etc.) using just the method you describe when it was just a red blinkenlight on a panel. When the LEDs cost $10 each, though, it's a good excuse to build something nice.
posted by fatllama at 7:13 AM on October 25, 2011

Isn't your MOSFET operating in the linear region, though, and burning the same power that the LM317 in current-regulator mode would be?

You could turn it into a switching regulator by replacing the opamp with a Schmitt trigger and adding an inductor and a diode, but at that point you're just reinventing the wheel and might as well use a chip made for the purpose (eg, eg).
posted by hattifattener at 12:09 AM on October 27, 2011

Isn't your MOSFET operating in the linear region, though, and burning the same power that the LM317 in current-regulator mode would be?

Ok, I had in mind being able to actively lower the voltage at the drain of the FET (imagine a slow feedback loop that dialed down the pulsed output of a switch-mode to keep just the minimum voltage available to the FET), but in truth you could do this to the input of a linear regulator with maybe just a little more "drop out", making the point about power loss in the transistor mostly a red herring. There's still loss in the fixed limiting R, though.

And there's still the issue that a fixed limited R with a constant voltage source can't adapt to changes in the number of lights on the fly. Lights that don't blink might as well not be lights at all.
posted by fatllama at 11:42 AM on October 29, 2011

« Older Recommendation for an affordab...   |  Landlord possibly going into f... Newer »
This thread is closed to new comments.