How hot is too hot for a SSC star?

[havand]

I've been doing some tests and found that a rig I built on maximum is hitting ~110-115 degrees F (it's funny how hot that actually feels) and 125 immediately behind the star. I'm using a thin sheet of aluminum to transfer heat away to another location from the star. I think the aluminum is 1/16" or 1/8" sheet, can't remember. My question is, if i'm reading 125 under the star, how hot is the led? How much hotter can it go? The problem is, I don't have any heatsink data because it is a custom built thing and not off the shelf.

 

[clg0159]

Manufacturer rating is max at 185 F, although I do not know how well it handles that temp. and for how long...maybe someone with more experience overdriving these could answer that.

 

[havand]

Well, what's interesting about that is they say that is the highest 'operating' temperature. Does that mean ambient/heatsink? As you can see from the sheet, the junction temp can be MUCH higher, but I have no way of calculating that from what I have.

Does anyone know if the 'operating temp' is the same as the heatsink/ambient temp?

 

[evan9162]

What current are you running at? Where did you get the star? Is it an official star, or a cheap knockoff?

 

[havand]

I'm running it at about 500mA. It is an official star bought from Fred.

I'm just trying to figure out how hard I can push multiple LEDs on the heatsink I have. I guess i'm just curious how the heatsink temp right around the base of the star correlates to the junction temp. Ie, at what heatsink temp will the junction temp be exceeded? Also, as I asked before, what is the 'operating temp' defined as?

 

[evan9162]

Unless it's specifically labeled maximum junction temperature, operating temperature typically means case (package) temperature.

The datasheet (http://www.seoulsemicon.co.kr/_homepage/home_kor/product/spec/X4218X.pdf) contains all the data you need, you just need to examine it closely.

The maximum operating temperature is going to be of the LED case, which is specified at 85C. The maximum junction temperature is 145C, but only if you run at less than 700mA. Above that to 1000mA, you must keep the junction temp below 90C.

Seoul specifies the Junction -> board thermal resistance for a star board at 8.5C/W, while the junction -> case (which would be no star board) at 6.9C/W. So the star introduces 1.6C/W of thermal resistance.


So let's put it all together.

You measured 125F behind the star, which is 52C

The aluminium + thermal interface to the star adds about 1C/W

So thermal resistance from your measurement point to the case is going to be about 2.6C/W.

Thermal resistance from your measurement point to the junction will be about 9.5C/W

We want to be conservative, and simplify the math, so let's call it 3C/W and 10C/W respectively

At 500mA, the Vf of the Seoul is probably around 3.4V. The LED is dissipating about 1.7W.

So the case temperature is going to be 52C + (3C/W)*1.7W = 57C
You're fine there.

The junction temperature will be 52C + (10C/W)*1.7W = 69C
Everything looks good there as well.

 

[havand]

Unless it's specifically labeled maximum junction temperature, operating temperature typically means case (package) temperature.

The datasheet (http://www.seoulsemicon.co.kr/_homepage/home_kor/product/spec/X4218X.pdf) contains all the data you need, you just need to examine it closely.

The maximum operating temperature is going to be of the LED case, which is specified at 85C. The maximum junction temperature is 145C, but only if you run at less than 700mA. Above that to 1000mA, you must keep the junction temp below 90C.

Seoul specifies the Junction -> board thermal resistance for a star board at 8.5C/W, while the junction -> case (which would be no star board) at 6.9C/W. So the star introduces 1.6C/W of thermal resistance.


So let's put it all together.

You measured 125F behind the star, which is 52C

The aluminium + thermal interface to the star adds about 1C/W

So thermal resistance from your measurement point to the case is going to be about 2.6C/W.

Thermal resistance from your measurement point to the junction will be about 9.5C/W

We want to be conservative, and simplify the math, so let's call it 3C/W and 10C/W respectively

At 500mA, the Vf of the Seoul is probably around 3.4V. The LED is dissipating about 1.7W.

So the case temperature is going to be 52C + (3C/W)*1.7W = 57C
You're fine there.

The junction temperature will be 52C + (10C/W)*1.7W = 69C
Everything looks good there as well.

Thank you so much! you've been very helpful. I'm going to have to play with those numbers to see how far I can push it. It's funny how your perception changes despite knowledge. Just placing my hand on the heatsink, I thought 'too hot' But by the calculations, it should be just fine. Thanks evan, appreciated.

 

[Oznog]

You can also measure die temp with a fair amount of accuracy by using a tightly regulated constant current and look for the change in forward voltage from when it's cold. If forward voltage drops by 160mV from when it started at 25C and dVf/dT is -4mV/degC, then you know the die is around 65C.

 

[evan9162]

You can also measure die temp with a fair amount of accuracy by using a tightly regulated constant current and look for the change in forward voltage from when it's cold. If forward voltage drops by 160mV from when it started at 25C and dVf/dT is -4mV/degC, then you know the die is around 65C.

That's an extremely difficult measurement to take. The thermal time constant of the die is around 20ms. The die essentially comes up to temperature almost instantly, then slowly increases in temperature as the rest of the package (submount and slug) start to heat up.

So unless you have a tightly controlled pulsed current source, and an oscilloscope, you won't be able to reliably determine die temperature just measuring Vf with a DMM.

In addition, the dVf/dT varies quite a bit on a per sample basis. I've attempted to characterize the temperature coefficient using a TEC to lower the temperatuer of the junction by 25C. My readings were all over the place. The coefficient varied from 2mV/C to 5mV/C over 10 samples. So I think you will have a hard time getting an accurate Tj measurement just based on Vf drop.

 

[havand]

Good suggestion, but if I had the equpiment to accurately measure that, i'd probably have an IR thermometer and extremely accurate bench supply, etc to measure the C/W of the heatsink Interesting method none-the-less.

 

[evan9162]

Really, when it comes to LEDs, you want to keep them as cool as possible given the constraints of your housing/enclosure/etc and reasonable limits on size and whatnot. Keeping the LED cooler increases the lifespan and luminous output.