Cree LED lifespan / current?

While I know that this can only be guessed at, can anyone give me a rough indication what life I can expect from LEDs are various currents?

I was thinking of running some Cree R2s at 1400mA and would like to know what life I can expect from them. Is the 50,000 hour rating given to them the rating at nominal 350mA or at their rated 3W which is about 800mA?

They don't usually die instantly, but tend to tint shift and their brightness falls off over time, all due to heat I believe. I have seen graphs before, have you checked the datasheets?

kuksul08 said:

They don't usually die instantly, but tend to tint shift and their brightness falls off over time, all due to heat I believe. I have seen graphs before, have you checked the datasheets?

Click to expand...

I haven't checked the datasheets, other than to check the basic lifespan. Given that the lifespan ratings on things like hard drives, mechanical parts and LCD panels rarely reflect the real word experience that people have with these parts I thought I'd ask in here.

So it's heat that causes LEDs to die? So, If I were to sink them very well it would contribute to a longer lifespan? How much of a difference can good sinking make versus poor sinking? Can I turn a 5000 hour life configuration into say a 10,000 hour life just with better sinking?

- Naz.

MrNaz said:

Can I turn a 5000 hour life configuration into say a 10,000 hour life just with better sinking?

- Naz.

Click to expand...

If you go from very bad to very good then I bet you could.

MrNaz said:

While I know that this can only be guessed at, can anyone give me a rough indication what life I can expect from LEDs are various currents?

I was thinking of running some Cree R2s at 1400mA and would like to know what life I can expect from them. Is the 50,000 hour rating given to them the rating at nominal 350mA or at their rated 3W which is about 800mA?

Click to expand...

The Cree life rating assumes an average of 70% lumens maintenance after 50,000 hrs for a junction temperature of 80C. If you can keep the junction cooler, the life will increase.

Three factors can affect junction temp:

1. drive current
2. ambient temp
3. thermal resistance of your thermal path to the ambient environment to cool the LED

If you can isolate the driver from the LED, so that the heat generated by the driver does not contribute to the LED's junction temperature, that really helps.

If you can keep the ambient temperature low (e.g., by always operating in the Arctic :wave:, using active cooling, or using a fan), that also can help a lot.

Of course, optimizing the thermal conduction between the LED slug or PCB and your heat sink should be a design goal.

The Cree docs state that the thermal resistance between the junction and the solder point is 8C/W.

You also have thermal resistances between the solder point and the heat sink and the heat sink and the ambient.

I don't know what the Vf for an R2 is at 1400ma. But, let's assume 4V just to plug in a number. You can easily use the actual value via simple substitution into the calculations below.

P = Power dissipated by the LED = Vf * drive current = 4V * 1.4A = 5.6W

Tjunction = Tambient + (Rj-a x P)

Rj-a = Rj-sp + Rsp-h + Rh-a

where

Rj-a = thermal resistance from junction to ambient
Rj-sp = thermal resistance from junction to solder point
Rsp-h = thermal resistance from solder point to heat sink
Rh-a = thermal resistance from heat sink to ambient

Let's use the following assumptions:

Tjunction = 80C as our target for the maximum allowable junction temp (taken from the 50,000 hr spec above)

Tambient = ambient temperature

Rj-sp = 8C/W (spec from the Cree docs)

Rsp-h = 1C/W (an assumed value for a good thermal interface)

Therefore, you can solve for the required value of Rh-a:

Rh-a = (Tj-Ta)/P - Rj-sp - Rsp-h = (80-Ta)/5.6 - 9

That means that (80-Ta)/5.6 >= 9, if Rh-a is to be a positive number. That gives a max operating ambient temp of about 30C, or 86F. And that requires a thermal resistance between the heat sink and ambient of zero C/W. This is basically as good as you are going to get with improved heat sinking unless you change some operating conditions.

If you can actively cool the ambient, then you can accept a higher Rh-a thermal resistance value. If you are willing to run at a higher max acceptable junction temperature (trading off LED life), you can run at a higher Tambient and-or accept a higher Rh-a value.

Justin, I stumbled across your nice analysis while trying to figure out how cool I need to keep my LEDs. A few comments ...

Perusing Cree documentation:

http://www.cree.com/~/media/Files/C... Application Notes/XLampThermalManagement.pdf

... they state that typically only about half the power that goes into the emitter (P in your equations) is dissipated as heat, the rest being dissipated as light (but they recommend using a conservative number of 75% of the power dissipated as heat); this helps a bit.

I also wonder if it'd be simpler to directly measure the temperature of an emitter while testing at a certain drive current. Obviously you can't directly measure the junction temperature. But if you could measure the heat sink temperature, you could then extrapolate to get junction temperature, using Rj-h pf 9C/W (per your numbers). I've tried measuring heat sink temperature using an IR gun, but it's a little tricky (maybe I don't have a very good one, but the laser targeting beam is not that accurate). I've also tried just pinching the heatsink between my fingers and seeing how much it hurts But I'd need numbers for perceived pain versus temperature for metal objects ...

cree.com gives many of their new LED's a life span of 50,000 hours.

sidecross said:

cree.com gives many of their new LED's a life span of 50,000 hours.

Click to expand...

Yes, if you keep the junction temperature within spec.

A little question that occasionally springs up in my mind.

If I remember correctly, white LEDs lose output because the phosphor degrades due to the heat, to the point it begins to impede light from the blue die underneath making it out.

But: what if one were to use a phosphorless blue LED of the same type? Exact same die and everything, just without the phosphor? Logically, one assumes without the phosphor turning into gunk the die itself should be able to keep its rated output for far longer; and yet, I don't see blue LEDs being used in long-lasting applications where lifetime would have a priority over colour. Why is this?

Fallingwater said:

But: what if one were to use a phosphorless blue LED of the same type? Exact same die and everything, just without the phosphor? Logically, one assumes without the phosphor turning into gunk the die itself should be able to keep its rated output for far longer; and yet, I don't see blue LEDs being used in long-lasting applications where lifetime would have a priority over colour. Why is this?

Click to expand...

You mean like in the original Philips Alien Head bulbs? ....or in a few other remote phosphor implementations? The reason you do not is cost. It costs more. It can also be harder to maintain color consistency and even efficiency can suffer ...then there is the limited optical control.

It's not failure so much of the phosphor, but the phosphor/silicon interface layer.

LED life if not simply a factor of junction temperature. Drive current comes into play as well. There are multiple failure mechanisms at play.

The best data is not the data sheet, but the LM80 data for the part. It's available on most suppliers websites, though with some suppliers you will need to ask. Keep in mind they don't publish this data at the extremes that flashlight people may use the part at, so you are at best guessing at those levels.