I'm looking for good data on the forward voltage vs junction temperature characteristics of medium (~50mA-150mA) to high power (>=350mA) LEDs at a fixed current or any other data detailing the operating characteristics of LEDs at diferrent juction temps, eg power vs Tj would be OK depending on the measuring cct.......(If v Tj at a fixed Vf wouldn't make much sense, Vf will vary regardless, however If can be controlled).
Anyway, looking to design a cct that can predict the most reliable current/voltage to drive an LED or LED array under variable thermal environments (eg sometimes bad heat sinks...sometimes good....sometimes excessive ambient heat, sometimes good convection cooling etc). I'm guessing the characteristic curves will show some sort of indication of when an LED is driven too hard, some sort of sweet spot.....I hope it's not too linear!
I've attempted to bracket the thermal coefficeint of the Vf for Luxeon LEDs. I used 10 samples from the same manufacturing tree, all of the same bin. I used a peltier to cool the devices to 3 different temperatures, and they were held there for quite a while and operated at different current levels.
What I got were wildly varying results, from 2mV/C all the way to 6mv/C. I don't think that Vf is a reliable way to measure Tj in any reasonable fashion. Remember that that coefficient is a typical value, just as Vf and Luminous output are, and we see how much Vf and Luminous output vary on an individual basis, the same is very likely true for the thermal coefficient of Vf.
Thanks for the info. Actually individual variation isn't too much of a problem, I'm hoping the cct will adapt to whatever the characteristics of the LED are....in particular variations in installation (which are out of my control).
Mainly I'm looking for whatever electrical characteristics of the LED that can be measured to determine where the LED should be operating at, for example, how do manufacturers determine that a particular LED must be operated at a given junction temperature.....and whether the LED Vf/If characteristics will show any characteristic curves to determine that operating point.
I know that Vf falls with increasing temp (hence a resistor limited LED may experience thermal runaway as the current increases)......but not how or why it falls. Even if Vf could be measured at initial switch on (ie Tj at ambient) then the cct will limit current when Vf falls below a certain % will be OK.
It would probably be easier to simply measure the temperature of the slug. You can use a thermistor, which has a known resistance vs. temperature curve. If you're using a constant current diver already, you would simply use the thermistor as part of the feedback system.
It's easy to estimate Tj by knowing the slug temp and power in the device. Since all devices have a similar Vf range (lets say 3.3-3.6V at a given current), you're only going to be off by 10% at most in the power disspation area, and off by even less in terms of Tj.
Lets say for instance, that the slug is measured at 40C, and you're running a Lux III at 700mA. If you assume 3.6V for the Vf at that current, then you're estimating 40 + (15 * 0.7 * 3.6) = 78C for Tj. If the actual Luxeon really has a Vf of 4V, then the actual Tj is 40 + (15 * 0.7 * 4) = 82C. If the Luxeon has a Vf of 3.3V, then the actual Tj is 40 + (15 * 0.7 * 3.3) = 75C.
So by simply measuring the slug temp, and making an assumption about Vf, you'll only have a margin of error on Tj of +/- 4C. I seriously doubt you could achieve the same accuracy by trying to measure a the Vf drop due to an increase in Tj.
mmmm....OK thanks, I thought about a thermometer feedback, but the thing gets installation specific then...difficult to make into a general driver...which is what I'm aiming for. I can see your point about accuracy.....
Do you (or anyone else) know why LED spec sheets specify X max current or Y max junction temperature etc.....basically I'm asking the question, if you have a LED, can you determine simply from the characteristic curves what the most reliable (ie below Tjmax) operating point will be.......then I'm hoping to code it into a micro to do the controlling.
If the manufacturers use some sort of statistical testing methods, test to distruction or make some sort of guesstimates to figure out their maximum limits....then I'm stuffed. But do they? Could there be some sort of common physical process of all LEDs (and I presume semiconductors) that are reflected in the curves?
I wonder if the transient Vf response (ie injecting a short pulse of current) will change in a distinctive way near the maximum junction temp of any device. It should be small until the device conducts, then pretty linear along the main operating area (If) and should decrease as Tjmax is neared (at Ifmax_....that may be useful ,I remember someones comment about that in another post. Using a short pulse (on top of a static If) will enable you to measure the response, without actually heating the device any extra and affecting the measurment. So I suppose I should be asking for the transient or AC characteristics of LEDs near its maximum operating point. Here is another question, what happens to the photovoltaic effect at different temperatures.
I'm not really sure how the specify a maximum current, but I can offer some conjecture. It may have to do with maximum current density the device can handle (so directly related to size of the die). It also may have to do with meeting operating life goals. The luxeon III, for instance, has the same lifetime operating at 700mA that the Luxeon I is rated for.
Maximum junction temperature is probably a lot easier. It is likely due to temperature limits of the materials within the device, either the semiconductors themselves, or even assembly pieces like solder. I don't think that statistics have any part in determining the maximum junction temperature, I think that's something they can specify with knowledge of the materials.
I have no idea on the transient Vf response.
One thing to keep in mind about trying to make a generic, one size fits all driver - I think you will have a hard time making it work for any more than one variety of LED. Luxeons have much different characteristics than say a Nichia power LED, a Cree XL7090, an Osram golden dragon, or a lamina array. I would suspect you would even have differences in a Luxeon III vs Luxeon I or Luxeon V. It would be interesting to try, but you'll probably have to do a lot of testing with a lot of samples of different types of LEDs if you want this to be one-size-fits-all.
Re: Wanted: Vf vs Tj characteristic data of LEDs..more info found
Just come across an article that explains a bit more about how the junction (Tj) temperature is determined for a device. http://www.lumileds.com/pdfs/protected/AB20-4.PDF
Apparently it is the temperature at which the innards of the LED turn to mush, putting stress on the die/contact bond and buggering the thing up.
The efficiency, Vf and light output are affected by the temperature, but it seems to mainly to be a mechanical effect that results in the specification....that sort of spells the end of this circuit idea.
your all over my head, but the sweet spot, for the level of light it outputs would be by reading the light, not the heat :-)
to find the sweet spot, have the micro controller, keep turning up the current slowly until the output of light is no longer relative to the input of power, cause the leftovers are heat :-)
a itsy bitsy photocell hanging out on the side.
of course when the sun is out or there is other light, i guess all that goes to heck.
so reverse that, and keep turning up the current to some MAX, and when the heat is starts to go up faster than the input current rise then more of the power is going to heat.
then just like the external light, you have external heat being an issue.
i guess both of those external things could be determined with an sencor that is Not next to the led, just like an indoor outdoor thermometer.
i assume that is usually about where it is already speced for. seems the Heat goes up lots, and the light doesnt go up so much, right after you leave SPEC for the led.
whatever it is, it has to be good today, and tomorrow when the phosphors are half dead, 10 years later any calibrations , just like the differences of the leds, will all go by by again, as the heat is the same, and the phosphors just arent putting out what you want.
which sweet spot is trying to be discovered? where the led is pumping out the most light, before the heat goes balistic.
seems like that is usually closest to spec, the current/lumens charts usually show that. you can also see it, very easily with a current meter, and your eyes to see the output, and your finger , on the heat sync to feel the heat.
then if its thermal survival, how high can you go without killing the led, the problem with that is they dont pop, when you go way over spec, they just dont live as long.
if its just to stop thermal complete instant death, then simple thermal probing close to the emitter base, would be sufficient, but that wouldnt prove that it is not going to die in 30-100 hours.
that just leaves (easily) run it a bit over spec, and thermal detect if its getting extrodinarily hot, and cool it a bit. so when it runs hot for to long, or is in extreme enviromental conditions it slows a bit to save itself.
any extensive controlling beyond that will just backfire somewhere, over time, or with differences that can exist, or just trying to make it to smart.