Junction measurement of K2 and Golden Dragons

Hi,

I'm evaulating the thermal design of two arrays, one of seven equi-spaced (ie a hexagon) Osram Golden Dragons, and another of seven Luxeon K2s.

Without having to cut into the PCB to get at the heatsink slug, can anyone offer a suggestion to measure the Junction Temperature of these devices?

For the GD I've soldered a thermocouple to the anode of the center LED which should be sufficient, but am a bit unsure about the K2.

The idea is I measure the solder temp, use the datasheet Rthjs (junction->solder point thermal resistance) and input power to estimate the junction temp.

The GD datasheet gives Rthjs, but the K2 datasheet only gives me derating curves for various Rja (junction->ambient) values.

Any idea, app notes etc that can help me measure the junction temp for the K2 (and is measuring the anode of the GD reasonable)??

Phil

Well,, I never tied it, but I would try a infrared thermometer. They can measure the temperature from certain distance through the lens.

But unsertain is: does it work through the K2 dome ? And.. can the temp be measured during operation of the led ?? Or can it only be measured directly after switching off the led ??

Sorry thats all I can think of...

Ra.

mmm...I never thought about measuring using a infra-red thermometer just after switching off. Sounds like a good idea. I don't see why the lens would be a problem.

Unfortunately in this case the devices need to be measured in their operating environment...underwater..doh!

So back to the original question, where is the most appropriate spot to attach a thermocouple to calculate the junction temp for GD and K2 devices?

Phil

Measuring after the LED is switched off won't be accurate. The junction has a thermal time constant of about 20 milliseconds - after that, the junction temperature will drop to whatever the slug temperature is.

You can't directly measure the junction temperature by any means anyways. The junction in a region sanwiched between different materials of the LED, and its temperature is not directly measurable by any external means. Using an IR thermometer to measure the temperature of the phosphor is even less accurate, as it's another insulating layer between the junction and what you're actually measuring.

Your best bet is to measure the slug temperature, and use the specified thermal resistance numbers to estimate junction temperature. Lets say you're running a K2 at 1A, Vf = 3.8V. The thermal resistance is 9C/W junction-> slug. Power in is 3.8W.

If you measure 40C at the slug then the junction is at (3.8 * 9) + 40 = 74C

evan9162 said:

Your best bet is to measure the slug temperature, and use the specified thermal resistance numbers to estimate junction temperature.

Click to expand...

That's what I was thinking. I assume the manufacturers do something similar. Has anyone ever asked them how they measure junction temp?

Ok, didn't think the thermal time constant would be so small (where did you get the figure anyway btw...are there more app notes etc out there, I'd be interested in examining them)

Still, I've got the problem of measuring the slug temp, as I noted before I can't really cut up the PCB to get under the devices to access the slug. What do you guys think of this solution then:

On the bench get a freestanding K2 and GD into equilibrium at some safe current (eg 100-200mA) and measure the slug and one of the connections (eg anode) where the thermocouple will be attached.

Given the power into the device Pin=Vf * If and the temperature difference I can calculate the thermal resistance between the slug and the terminal.

Rslug->terminal = (Tslug - Tterminal) / Pin ....should be constant and a function of the device construction

So the junction temp will be:
Tj = Pin * (Rjunction->slug + Rslug->terminal + Rterminal->ambient) * Pin + Tambient

where

Rterminal->ambient = (Tterminal - Tambient) / Pin...this is a factor of the PCB+housing design etc

Rslug->terminal should be small in either case (hopefully), for K2 Rjunction->slug=9C/W and GD Rjunction->slug=15C/W.

I'm really after the max current I can run the devices at with the PCB layout housing/watercooling factors added so:
Ifmax = (Tjmax - Tambientmax) / Vf / (Rjunction->slug + Rslug->terminal + Rterminal->ambient)

Sound reasonable? Looking for a sanity check here.

Phil

The time constant has been measured by other board members in the past. The junction is quite small and has very little thermal mass, so it heats up and cools down nearly instintaneously when power is applied/removed.

Heres a problem you will run into. The only connection between the power leads and the junction is a thin gold bond wire. Once you attach the power lead to something (like a circuit board or whatnot) you will change the relationship between the temperature of the power lead and the slug temperature, as the power wire or copper pad (however you're supplying power) will sink away heat from the tab, causing it to be at a lower temperature, throwing off your measurements.

If you're looking for an absolute maximum current you can run at, you can start with the best possible case and work backwards from there, adding in thermal resistance to the body. So you know that you have 9C/W for the K2 and 15C/W for the GD. Start adding on 1C/W for the thermal epoxy, 2-3C/W for the heat sink -> ambient. Then throw 1-2C/W for a safety margin, and use the temperature of water as ambient.

So if you end up with 15C/W for the K2 in 10C water, then the maximum power dissipation is (155C-10C)/15C/W = 9.6W, which will probably be somewhere in the 2A+ range (exceeding the specifications of the K2). Running the K2 within specifications, at 1.5A will result in a Tj of 107C

Your best bet is to measure somewhere along the thermal path, even if its at the back of where the slug connects to the heat sink, I think it will be a more reliable measurement than trying to set up a relationship between slug temp and power lead temp.

It sounds like you may to do some destructive testing to get an accurate analysis. Why can't you cut into the back of the PCBs? A small hole drilled from the back of the PCB into the emitter slug to embed a thermocouple would give you great data.

If the analysis is that important, I would suggest that the sacrafice of a couple of modules would be in order. What about popping off an emitter and re-attaching it with a thin copper disk drilled for a thermocouple?

Perhaps one of our members, Newbie, who is very effective at destructive analysis, can offer some other suggestions.

Well, thanks for the input guys but I mentioned I cannot cut/modify the boards because I cannot cut/modify the boards, nor measure under the slug. The reason is that the back of the PCB will be hard fastened to the housing, and my client doesn't want to cut up the boards.

I cannot do destructive testing, sort of an important point. Perhaps newbie has some non destructive testing experience???

I want to check the thermal efficiency of the whole system as it will be in installation. As you've noted the thermal characteristics change when the LED is bonded....this is exactly the effect I'm trying to deal with, measure and determine an operating point for the boards as an installation. Cutting things up (yes I know it would be only a small hole) will affect things again.

I know the lead temp will be lower than the slug temp, hence the idea of measuring the slug->terminal resistance and factoring this into the equation to estimate the junction temp. Do you think that this will be markedly different when measured on the bench vs in situ? I might verify this by measuring a free standing LED slug/terminal temp difference then make up a little PCB, with a hole under the slug and see if there is a significant change in the slug->terminal thermal resistance when the LED is bonded. It won't be the same as the array, but it should tell me something about the effect, unless someone has already done something like this.

Perhaps if the idea of measuring slug->terminal resistance is going to be invalid do you think measuring as close as possible to the slug will be more accurate (which I can do, on top of the PCB)? I am prepared for a safety factor to take into account the inaccuracies, just looking for the most accurate, non destructive, method.

I might just go with the original plan, and add that safety factor.

Phil

Phil, if you have some spare emitters, why not create a mock-up of the planned implementation using one emitter only, a circuit board and an embedded thermocouple? You would have to mimic the manufacturer's assembly method as much as possible wrt the PCB. Best of luck with this one.

Paul

Perhaps if the idea of measuring slug->terminal resistance is going to be invalid do you think measuring as close as possible to the slug will be more accurate (which I can do, on top of the PCB)? I am prepared for a safety factor to take into account the inaccuracies, just looking for the most accurate, non destructive, method.

Click to expand...

I think this would be your best bet. If you can mock up a similar PCB and directly measure the difference, that would be even better.

Yep, on thinking about it a bit more, making the mockup of one emitter will probably do the trick. If I measure a free standing LED I'll get one resistance value, on a PCB the resistance should be smaller because there is a better thermal path from slug to terminal (or wherever the thermocouple is planted).
Thanks for the help guys.....I'll keep you posted on the results.

Another factor in the operating point determination is the thermal efficiency of the whole assembly from slug to ambient temp. There is an tool here:
http://www.futureelectronics.com/promos/lumileds/usablelight
which takes that assembly into account, guesses the effect of each LED temp in the array and optimises the current. Because the LED brightness decreases with temperature, it is not always the highest current that results in the highest flux, the assembly efficiency (which can be measured by guestimating the slug temp+ambient temp) should be determined for best results.

Should be interesting!
Phil

While you are learning, and considering, and thinking, sometimes how things are done on the manufacturers side of things, can be very useful information. I'm not sure if you have caught this one yet:
http://www.lumileds.com/pdfs/AB08.pdf


Also consider taking a look at this technique:
http://www.avagotech.com/assets/downloadDocument.do?id=50#search="agilent LED thermal measurement"

There is also the applied method of wavelength shift:
http://www.lrc.rpi.edu/programs/sol....pdf#search="agilent LED thermal measurement"

Another technique:
http://www.elecdesign.com/Articles/Print.cfm?ArticleID=11676

Thanks for the info Newbie, some good backgrounder stuff and food for thought.
I'd seen the Lighting Research Centre stuff before, that's a great site.

Not seen the other material though, the Vf non contact measurment methods sounds similiar to my automatic LED driver cct query a few months ago, in particular using a small controlled change in supplied current.

If anyone can wrap this into a robust and affordable commercial grade cct...this'd be very handy thing!

Phil