LED Lifetime vs. on/off time

[Eelcos]

As far as I know the lifetime of a LED depends on its junction temperature. The lifetime approx. doubles with each 10 degrees decrease of junction temperature.
The junction temperature, at its turn, depends at the one side of the ambient temperature and at the other side of the internal temperature rise caused by power dissipation in the LED.

At the internet I see that some people say that the lifetime of a LED doubles if it is driven with a 50% duty-cycle. Is this correct? To my opinion, a 50% duty-cycle reduces the internal power dissipation to 50% and consequently reduces the internal temperature rise to 50%.

Example:
A led has an internal temperature rise of +40 degrees if driven with 100% duty-cycle, and has a lifetime of 50k hours.
If this LED is driven with 50% duty-cycle the internal temperature rise will be +20 degrees, so the junction temperature decreases with 20 degrees. In that case the lifetime is multiplied with 2^(20/10) = 4, so the lifetime is 200k (and thus not doubled).

Is this correct?

Another question: if I have a LED backlight which is illuminated 12 hours a day and is switched off 12 hours a day, is this equal to the 50% duty-cycle case mentioned above?
Another question: and what if I leave a LED off for 5 years and then turn it on for 5 years? Must only the last 5 years be counted as 'lifetime' or is this also a 50% duty-cycle case?
Another question: and what if the LED is always off? Does it still degenerate, based on its junction temperature of 25 degrees?

 

[waross]

The life time may be influenced by the junction temperature.
When the junction is energized, the temperature rises on a curve defined mostly by the thermal time constant of the junction. When the junction is turned off the junction cools on the same curve but inverted. The time constant is the time to reach 63% of the final value.
If the on time is long in comparison to the thermal time constant then the junction will be at the final temperature for most of the on time.
A parameter is accepted to have reached the terminal value after 5 time constants.
I have no idea whet the thermal time constant of an LED may be.
The time constant may vary widely with variations in mounting and the effectiveness of any heat sinks in use.
Ambient temperature may be a factor.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[VEBill]

"...5 years..."

You've touched on the comedy of naive "duty cycle" claims. I've had to use a similar rebuttal ("50%? You mean, like one year ON, one year OFF?") to force people to provide the absolute timing assumptions behind their naive and meaningless duty-cycle claims.

There's some good info about LED failure modes on the net. Just an example:

https://www.led-professional.com/resources-1/articles/led-failure-modes-and-methods-for-analysis

Most failures modes would be driven by temperature and current. But other modes may be environmental, driven by exposure not usage (imagine an LED left exposed to the salty ocean air).

The points made about latent ESD damage and overstress during soldering are obviously important.

For most applications, it's not worth going beyond the simplistic MTBF analysis.

If you're going beyond that because of a critical application, then you'd probably need to set up a screening process for incoming batches. You wouldn't rely on OEM average MTBF data without knowing the variation from batch to batch.

 

[bacon4life]

I am unconvinced that cutting the power consumption in half precisely cuts the temperature rise in half:
* For overhead power lines, natural convection cooling is proportional to the 1.25 power of the temperature difference. A convection cooled power line would be at 23C rise rather than a 20C rise.
* Radiation losses are certainly lower than for incandescent lights, but they may need to be accounted for as they are proportional to the 4th power of the temperature difference.
* Is driving the LED at a lower voltage the same as driving with a higher voltage 100 Hz 50% duty cycle?
* Although you asked about the LED itself, the power dissipation from the driver may not be proportional to the power dissipation from the LED as the duty cycle is lowered.

 

[itsmoked]

Yes, leaving the LED off half the day will typically double its usage lifetime.

It's a non-linear device. Running with a NORMAL 50% duty cycle in the kHz doesn't reduce the temperature 50% it reduces it MORE THAN 50% because running an LED at a lower temperature causes it to have less losses.

If you leave it off for 5 years you simply haven't used 5 years of its lifetime. Barring environmental issues as mentioned prior.

Yes, they degrade while never used but slowly enough to not matter to anyone.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Eelcos]

Thank you all for your useful replies!