Driving LEDs

[**DONOTDELETE**]

I'm experimenting with driving several (up to 12)white LEDs in series from a regulated voltage source, without a current limiting resistor. My voltage is a multiple of the LED forward voltage (in my case 3.6V), and I was expecting the current to be the stated forward current for those LEDs at that forward voltage. It isn't. It's much higher and increases with time. In fact, I cant even put 3.6 V across just one LED and get a steady forward current (stated as 20mA). I know a lot of designs use multiple LEDs in parallel from a constsnt voltage source without a sries resistor. What am I doing wrong? Is the increasing current caused by thermal run away? will it ever stabalise?

 

[Jonathan]

Search online for an LED datasheet, and look for a current versus voltage curve. What you will find is that LEDs are very unlike ordinary resistive loads.

With an ordinary resistive load, the current flowing through the load increases pretty much linearly with the voltage applied to the load. With an LED, the current flow is pretty much zero, and stays zero, until you reach a threshold voltage, and then the current starts going up _very_ quickly. What this means is that around the proper drive voltage, the current changes dramatically. For example, if the proper drive voltage is 3.5V, then at 3.4V you might see only 1/4 the desired current, and at 3.6V the device would be severely overdriven.

The problem is that the proper drive voltage changes with numerous factors, including device production lot, age, temperature, heat sinking, etc. Your nominal 3.6V device might really need 3.5V or 3.7V, and the only thing that you can do is determine this once you have the device in hand.

There are really only two good ways to drive an LED. One is with some sort of current regulated power supply, which constantly adjusts the output voltage in order to maintain the proper drive current. The other is to place a resistor in series with the LED in order to make the current versus voltage characteristic less sensitive to small changes in LED voltage. Driving with a stiff voltage source is exactly wrong

One trick that is sometimes used is to drive the LED with a soft voltage source, meaning one in which the voltage drops as current increases...this is essentially a voltage source in series with a resistance (one of the basic techniques described above), but this resistance might (for example) be the internal resistance of a lithium cell.

-Jon

 

[remuen]

The increase in the current you have seen is caused by the thermal runnaway.

As Jonathan already mentioned each LED is different from the other. Here a comparision of the currents (in milliamps) I've done a couple of months ago:

Vin LED1 LED2 LED3
3.4 21.9 13.5 09.9
3.5 33.6 20.2 12.1
3.6 53.5 31.2 14.8
3.7 87.5 47.3 17.6

These figures show two things:
- The dramatical change in the current on only a small change of the voltage
- The currents of these LED's vary very much even they have the same voltage

What does this mean for any application with LED's?
<UL TYPE=SQUARE><LI>There is only one really correct way to drive LED's: Using some kind of current controller. A resistor is a very simple but not an effective and in most case not a efficent way of controlling the current. It is more a current limiter. Better would be a circuit with a Zetex ZHSC300 chip or with a LT1932 from Linear Technology. Use the CPF search function to find more about these chips and circuits in the LED forum (use ZLT and LT1932 as search items).

<LI>Switching a couple of LEDs in parallel is a bit dangerous. The above figures show: If one LED is still underdriven another one could already be overdriven! A better solution would be to switch the LEDs in series. With 12 LEDs you could make eg. 3 or 4 chains with 3 or 4 LEDs each in series and then connect these chains in parallel. Note: All chains must have the same amount of LED's!

To drive a couple of LEDs in series you will certainly need a higher voltage. Most current controller are booster circuits (step up regulators) which are able to deliver a output voltage which is higher than the input voltage.
[/list]

Btw, a current controlled booster does also control the thermal runnaway of the LEDs because it keeps the current through the LEDs constant.

 

[Evan]

I thought thermal runaway was the process that actually destroys the LED after the heat of overdriving accumulates to a fatal degree.

This does raise an interresting question: is anybody using a thermal sensor on the LED to feed back to the power source and cause the power to be adjusted to achieve a specific case temperature rather than a certain voltage or current? This would seem to be the way to maximise LED output without thermal runnaway since the limiting factor in driving the LED is the junction temperature.

 

[dat2zip]

Evan,

From what I've seen and I think I read on the Luxeon Star datasheet is that the LED has a negative Voltage coefficient.

What this translate to is if you apply a fixed voltage say 3.3Volts on a white LS and it draws 0.35A. As the LED heats up its Voltage drops and the current increases which heats the LED up even more and the voltage drops and the current increase and so on. If this is left unchecked it ends up running away overheating and cooking the LED.

The Nichia and for that matter I think any LED has this behavior.

Adding a small resistor in series with the LED helps stabilize the high sensitivity to small voltage change VS current changes of the LED. One might want to run 3 LEDs in series with a 10 ohm resistor and make 4 strands like this and power all the strands in parallel.

You can start with a voltage approximately 3.5 * 3 and then measure the voltage drop across each of the 10 ohm resistors. Increase the voltage till the voltage drop across the 10 ohm resistors gives you 20mA.

200mV / 10 ohms = 20mA.

With this method you will find the 4 strings will all have similar currents flowing through them.

If you want a more precise method you can make a current source regulator instead of a voltage source. Look at the LM317T series 3 terminal regulators as a possible IC for use.

 

[Evan]

Yes, I've used an LM317 as a constant current source. But still, regulating current is just staying down in the safe zone. If you really want to overdrive an LED to the limit without the heat turning its fine crystal into amorphous crud, it seems the ideal would be to monitor the junction temperature and keep pouring on the current until the junction was just below 125 degrees C or so, controlling maximum dissipation rather than limiting current.

Of course there are complications because you can't normally get the junction temperature, you get the case temperature and must compute the junction temperature using the thermal resistance for the case and room temperature. Plus the case takes time to react to the heat coming off the junction, so there is a time delay. Still, if you wanted to drive an LED for all it is worth, you would need a circuit than ups the current a smidge every few seconds as long as the case temperature is below, say, 50 - room temp, and cuts the current perhaps a bit more rapidly when the case gets above 50 - room temp, so that the junction takes all the current it can but never quite gets hot enough for thermal runnaway.

 

[Evan]

How did you choose the LT1618? Why do you prefer it over the LM1932? What kind of efficiency are you getting?

With regard to pushing LEDs real hard: I have noticed that the voltage across some LEDs I've tested drops as they warm up. So with constant current drive, the input power should drop as the LED gets hotter and you get some of the temperature regulation I was after. The place where I think this would matter is where LEDs are used in applications where there are temperature extreemes, such as in a car. Is the self-regulating nature of the LED at constant current enough to keep it from becoming toast as, say, a marker light on a black truck in an Arizona Summer? But I guess if you'll settle for whatever the LED can handle on the hottest day, there isn't that much advantage to being able to be a lot brighter on the coldest day. And yes, by other threads the gentler you drive the LED, the more efficient, and of course the longer its life.

 

[Jonathan]

I chose the 1968 rather than the 1932 because I wanted to operate at substantially higher output currents with a higher input voltage. The 1932 is limited to 10V in and about 40mA out.

The Superflux LEDs are rated for 70mA, and I have three parallel strings, for a nominal max current of 210mA. You select output current by selecting a sense resistor which develops 50mV at the desired current.

With an input voltage of 12V, I push 2W through the part with no problem, but it starts getting hot when I push 4W through it. Not bad for an itsy bitsy little thing.

-Jon

 

[Evan]

2 to 4 watts! For just the tailight? What are you running it on?

 

[Jonathan]

You might be able to use the LED itself to measure the junction temperature and get rid of the time delay. If you measure the forward voltage drop with a fixed current, then you would be able to see the temperature changes at the junction quite rapidly.

You would need to jump through several hoops to calibrate this control to the particular diode in question, eg. measuring the room temperature forward voltage drop at a fixed _low_ current prior to powering the LED with you regulated (but changing under regulation) drive current, or using an external temperature sensor as an outer control loop to tell the inner control loop what to do.

But an important question is 'why push a single LED so hard'. LED efficiency and lifetime drop as the junction gets hotter, and the chance that you will toast the LED instantly becomes very real. If you need a small, low efficiency, high power light source, then use a halogen lamp

-Jon

 

[Jonathan]

But getting back to the original question for this thread. I've just reached a near finished prototype for a bicycle tail light, which uses the Lumileds Superflux LEDs.

Lumileds is the company that makes the Luxeon high power LEDs. The Superflux LEDs are much lower power devices, operating at 70mA rather than 350. They are also only available in Red and Amber (Vishay sells similar LEDs in other colors, but they are much less efficient). However they are also much less expensive, costing about 50cents each. They are perfect when you want a distributed array of emitters rather than a single extremely bright point source.

I am driving these LEDs in series, in strings of 8 LEDs. Because these are red LEDs, the string voltage is between 17 and 19 volts depending upon current.

To drive these LEDs, I am using a current regulating boost circuit based upon the Linear Tech LT1618 chip. Basically using the datasheet circuit, I take an input of 3 volts up to the LED threshold voltage, and boost it until the current through the LEDs is at the regulation point. If the LEDs heat up and the voltage drop changes, then the chip automatically changes the switching to get an appropriate output voltage, and maintain the proper LED current.

Somewhen I'll be writing up a complete description of the circuit, but the above pointers should help anyone interested in building their own.

-Jon

 

[Jonathan]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by Evan:
2 to 4 watts! For just the tailight? What are you running it on?<HR></BLOCKQUOTE>

Well...2 to 4 watts is what the LED array and boost converter are capable of; I anticipate running them substantially dimmed down most of the time, but I wanted something that would work as a daytime visible brake light, and an amber light that would work as a daytime running light....and after I built the LED arrays I found that I'd overdesigned a bit.

Power is from the main NiMH pack which runs the halogen headlamp, and it has loads of watt hours available.

-Jon

 

[**DONOTDELETE**]

Thanks for all the posts in reply.
My origional question (driving LEDs from a constatnt voltage source with no current limiting) was prompted by a design for a caving torch driving 21 white LEDs in parallel from a 3.1V voltage source derived from a DC to DC step up LM2621 with no visible means of current regulation. (www.resurgentsoftware.com/Perfect_LED_Light.html)
How does this design 'get away with it?' or does it just accept different LEDs will be driven by varying currents depending on individual LED variations.)
In terms of my own project, (driving many LEDs in series), all the replys assumed I was using a battery source of limited volts (Not surprising i guess, given the name of the forum!). In actual fact, the unit will be mains powered (240V around here), so step up current limited drivers are not really necessary. I'm currently experimenting with regular 3 terminal regulators (7812 etc) in a constant current configuration. Any better ideas? I need to push about 40mA through at least 12 white series LEDs (approx 48volts forward voltage), and eventually be able to dim the whole lot using pulse width modulation.

 

[dat2zip]

Chissers,

Running LEDs off the mains (120VAC or 240VAC) needs some special attention.

I hope you are planning to use a transformer to step down the 240VAC to 48V DC (33.9VAC) or something thereabouts.

If you were to rectify and put a large filter cap on the rectified 240VAC you would have 339V DC. Subtracting the 48V of the LEDS leaves the difference (339V - 48) that the 7812 would have to acommodate.

First of all most linear regulators cannot handle that large Voltage input and two the regulator would have to dissipate some 11 Watts of power (291V * 0.04).

A better solution is to use a transformer to step down the AC voltage to something close to the desired voltage to minimize the power the linear regulator would have to dissipate.

(www.resurgentsoftware.com/Perfect_LED_Light.html) has been updated if you have not seen it lately. There is a revised version which does regulate current.

The LM7812 (LM340) series has a MAX Vin of 30-35 Volts. If the GND pin is connected to GND then you will not be able apply more than 35V to the input. If the LMXXX is wired as a constant current source where the GND pin has a resistor to the output to set a constant current. Then I think the 48V applies to the difference seen by the IC between the input and output pins. In this configuration the IC can float at a higher voltage than 35V.

It might be much easier to do an array of series parallel LEDs with a small resistor in each leg to balance the strings.

Good luck with your project

 

[**DONOTDELETE**]

Thanks for the reply.
I shold quickly add that my design does include a transformer, to produce a 48VDC supply from which the constant current source is powered.

 

[Jonathan]

I recall that there was a CPF thread on directly powering an LED string from the mains, using a capacitor for current limiting...such is what I would call a scary design.

In terms of a simple way to run a long series string of LEDs, I think that the plan of using a step down transformer followed by a constant current linear regulator makes good sense. It is simple and will get the job done.

I'd suggest taking the output of the transformer, rectifying and filtering it to get unregulated DC, and then using a three terminal regulator in current mode to drive the LEDs.

As has been noted, you will have to carefully check the voltage ratings of the three terminal regulator that you use, and you may find that you are better off using a 24V secondary and two shorter strings of LEDs in order to drop the voltage a bit.

In terms of PWM, a simple logic level fet driven by your PWM generator should do a fine job. Select a fet with sufficient max Vds to interrupt the current through the LEDs.

-Jon