lifetime of LED vs Zener?

[ledtester]

Suppose you have an LED and a Zener with the same voltage drop and both are dissipating the same amount of power. What can you say about the lifetime of each?

[Zero999]

It's a silly question because there isn't enough information to even make a guess, let alone a definitive answer.

[RoGeorge]

I would expect the LED to have a shorter life.

The LED is direct polarized, Zener is reversed polarized, so the E field for the same voltage would be stronger inside the LED junction.  I would expect a stronger electric field to eventually pull atoms from their place and mess with the crystalline structure of the junction.

Another argument might be the energy level jumps of the electrons.  To emit visible light the difference in energy levels are higher, so more wiggle that, combined to the thermal wiggling, might eventually irreversibly damage the junction.

These are just cherry picked arguments based on qualitative guesses.  Didn't make any estimation, and I have no idea if these will stand true in a quantitative evaluation.

[PCB.Wiz]

LEDs certainly do have an aging effect on light output, but I’ve never seen that expressed as voltage change ?

[jpanhalt]

As implied, why not use each for its designed purpose?

Anyway, the MTBF of Zeners and LED's is available.  Here's what I found:

Zener (mil spec) >3x10^7 hours (MTBF), maybe 3.03X10^8 hours
https://product.tdk.com/system/files/dam/doc/product/power/switching-power/ac-dc-converter/reliability_data/scs_mtbf.pdf

For definition of terms: https://product.tdk.com/system/files/dam/doc/product/power/switching-power/ac-dc-converter/reliability_data/css150_mtbf.pdf

LED (commercial) 10^5 hours lifetime
https://www.electronics-notes.com/articles/electronic_components/diode/light-emitting-diode-led-lifetime-lifespan-expectancy-mtbf.php

LED (mil spec) 10^6 hours MTBF
https://alders.de/en/products/counter-indicators-leds/mil-spec-leds/

[Jon_S]

LEDs certainly do have an aging effect on light output, but I’ve never seen that expressed as voltage change ?

I previously worked for a company that also did LED lighting (back when this was a very novel thing!). I was primarily in design, but I did get involved with the testing and they did a lot of accelerated lifecycle work.
As I recall, the general trend was to see a modest increase in forward voltage as the LEDs aged, maybe 5-15% once luminous flux had decayed to 70%.

I don't recall ever seeing that sort of thing in print form though!

[Zero999]

The above is all without knowing more information.

What sort of LED? IR, UV, white, green?

Power level?

Application?

A zener is normally either used as a voltage reference which is continuosly biased by a small current, or as a voltage clamp which only conducts during spikes. An LED is normally on continuosly, execept for when it's used as a strobe.

I have used an LED as a volage reference before, but only because it was convenient and I also needed an a power indicaor. It was for an LM311 comparator, run off 5V. It was a simple circuit to close a relay, when a potentiometer's position passed a certain point, adjustable via another on-board pot, to emulate a handbrake for a car racing simulator. I connected a silicon diode and old green LED in series, across both potentiometers, to ensure the voltages at both inputs were between 0.7V and 3V, so there wasn't any deadband, due to the voltage being outside the input range of the LM311. It worked well enough because precision and tolerance weren't requirements.

[moffy]

As implied, why not use each for its designed purpose?

Anyway, the MTBF of Zeners and LED's is available.  Here's what I found:

Zener (mil spec) >3x10^7 hours (MTBF), maybe 3.03X10^8 hours
https://product.tdk.com/system/files/dam/doc/product/power/switching-power/ac-dc-converter/reliability_data/scs_mtbf.pdf

For definition of terms: https://product.tdk.com/system/files/dam/doc/product/power/switching-power/ac-dc-converter/reliability_data/css150_mtbf.pdf

LED (commercial) 10^5 hours lifetime
https://www.electronics-notes.com/articles/electronic_components/diode/light-emitting-diode-led-lifetime-lifespan-expectancy-mtbf.php

LED (mil spec) 10^6 hours MTBF
https://alders.de/en/products/counter-indicators-leds/mil-spec-leds/

Unfortunately milspec MTBFs are a calculation e.g. the opening paragraph of your first reference, not verified against real world performance. We would have loved for our milspec Inertial Navigation Systems, Radars etc. to get the predicted MTBFs, but the manufacturers know that they are a fiction requested by the customer. Real life failures were mostly stressed components and there was a definite higher failure rate amongst analogue vs digital, but that was in the 80s when digital densities were much lower.

[Phil1977]

LED lifetime is usually defined for a serious load that increases temperature and compromises lifetime.

If you keep the load low then the lifetime is nearly infinite - like the Zener.

[SteveThackery]

As implied, why not use each for its designed purpose?

Anyway, the MTBF of Zeners and LED's is available.  Here's what I found:

Zener (mil spec) >3x10^7 hours (MTBF), maybe 3.03X10^8 hours
https://product.tdk.com/system/files/dam/doc/product/power/switching-power/ac-dc-converter/reliability_data/scs_mtbf.pdf

For definition of terms: https://product.tdk.com/system/files/dam/doc/product/power/switching-power/ac-dc-converter/reliability_data/css150_mtbf.pdf

LED (commercial) 10^5 hours lifetime
https://www.electronics-notes.com/articles/electronic_components/diode/light-emitting-diode-led-lifetime-lifespan-expectancy-mtbf.php

LED (mil spec) 10^6 hours MTBF
https://alders.de/en/products/counter-indicators-leds/mil-spec-leds/

There is something very important to point out: MTBF has nothing to do with ageing and lifespan. It is a measure of the random failure rate of a component when operating within its normal lifespan (see the second reference above: MTBF = 1 / lambda, where lambda is the failure rate).  It specifically excludes failures due to ageing.

Even ignoring ageing mechanisms, only about a third of components will last as long as their MTBF.  Select a random group of 100 identical components with an MTBF of 1 million hours (say). After 1 million hours of operation 34 components will still be working (on average).

Once we introduce ageing mechanisms - such as the well-known ones for LEDs - the survival rates fall away much faster, depending on how quickly the ageing mechanisms proceed.

Last time I checked, the MTBF of a 40-year-old male human (where "failure" is taken to mean "death") is about 360* years. All that means is that there is a 1-in-360 probability of any given 40-year-old male dying in his 41st year, and out of a large population of such men, 1 in 360 will die that year. I think the life expectancy of a human male here in the West is something like 80* years? So at 40 years old, a male human has an MTBF of 360 years and a life expectancy of 80 years.

*I don't know how accurate those figures of 1-in-360 and 80 years are, but it doesn't affect the explanation.

[jpanhalt]

I used the definition given by IBM:

https://www.ibm.com/think/topics/mttr-vs-mtbf#:~:text=MTBF%20is%20a%20measure%20of%20how%20long%20a%20system%20or,downtime%20and%20reduce%20repair%20costs.

MTBF is a measure of how long a system or product is expected to operate before it fails, and it is used to plan for maintenance or replacement. MTTR is a measure of how long it takes to repair a system or product after it fails, and it is used to minimize downtime and reduce repair costs.

The numbers themselves may not be reliable, but the 10-fold or greater MTBF for zeners vs. LED's is factual compared to a guess that LED's have a shorter lifespan.

[T3sl4co1l]

Not to mention, under what transients, surge conditions, and statistics thereof? Operating temperature range, and statistics thereof?

Given perfectly smooth DC, and purely electrical means of testing, I suspect the answer to both is "longer than you can tell".

The LED has a failure mechanism due to charge trapping, causing reduced emission efficiency with use, and white LEDs the phosphor gets "bleached" so to speak.  But these don't have consequences as measured at the terminals; maybe the V(I) curve shifts some, but I'm not aware of any mechanism by which it shifts outside of nominal ratings.

We use LEDs for their emissions, obviously, so typical ratings are based on that.

I would expect the LED to have a shorter life.

The LED is direct polarized, Zener is reversed polarized, so the E field for the same voltage would be stronger inside the LED junction.  I would expect a stronger electric field to eventually pull atoms from their place and mess with the crystalline structure of the junction.

Another argument might be the energy level jumps of the electrons.  To emit visible light the difference in energy levels are higher, so more wiggle that, combined to the thermal wiggling, might eventually irreversibly damage the junction.

These are just cherry picked arguments based on qualitative guesses.  Didn't make any estimation, and I have no idea if these will stand true in a quantitative evaluation.

Strange, I would pick the zener: its junction is extremely thin (tunneling is occurring), and the bandgap, and binding energy, is lower for Si than AlGaAsP.  (Or at least, most alloys, probably.)  (Well, binding energy, or even just atomic dislocation energy, I'm not too sure of?  I *think* bandgap must be strictly less than binding energy, otherwise electrons alone could rip the stuff apart -- but dislocation, I don't know about.  Funny thought...)

Tim

[SteveThackery]

I used the definition given by IBM:

https://www.ibm.com/think/topics/mttr-vs-mtbf#:~:text=MTBF%20is%20a%20measure%20of%20how%20long%20a%20system%20or,downtime%20and%20reduce%20repair%20costs.

MTBF is a measure of how long a system or product is expected to operate before it fails, and it is used to plan for maintenance or replacement. MTTR is a measure of how long it takes to repair a system or product after it fails, and it is used to minimize downtime and reduce repair costs.

The numbers themselves may not be reliable, but the 10-fold or greater MTBF for zeners vs. LED's is factual compared to a guess that LED's have a shorter lifespan.

I can't help it if the author of that technical note got it wrong. It sounds to me like they aren't Reliability Engineers, just some freelance technical author who wrongly thought they understood the terminology and concepts. The explanation I gave is universally accepted throughout the world of reliability engineering, and is very simple: MTBF is the inverse of the failure rate, where said failure rate is measured after the early life failures and before the ageing failures (ie at the bottom of the bathtub failure rate curve)*. It is a statistical concept with little meaning to an individual component or product.

The problem with that definition you quote is that it takes no account of component ageing. It is completely normal for a component to fail well before its MTBF, even if ageing is not taken into account.

If you find my explanation unconvincing, I refer you to Wikipedia as a good starting point, and there are plenty of authoritative sources online.

Remember this simple formula: MTBF = 1 / lambda

... where lambda is the failure rate measured at the bottom of the bathtub failure rate curve.

*There is some debate about whether the bathtub curve applies to all components or only some. Those devices with no known ageing mechanisms show no increase in failure rates as time progresses, so there is no upward slope on the lambda curve at the right hand side. This doesn't invalidate anything I said: it just means that the MTBF remains constant indefinitely.

And remember: even if we completely ignore ageing, there is only a 34% chance of a randomly selected component still working after the MTBF has expired. I can give you the proof if you wish, but it's easy enough to find for yourself.

[SteveThackery]

I would just add that the lifetime of an LED is "it depends".

Unlike many electronic components, LEDs have well understood and very consistent ageing mechanisms. In fact, you can more or less choose what lifespan an LED will have depending on how hard you drive it. Given an easy life, an LED will often reach or exceed 100,000 hours before it croaks. Driven harder, as they usually are in domestic lighting, for instance, you can expect a lifespan of a quarter of that. Here in the UK at least, the expected lifespan of a domestic LED bulb is usually specified on the box; 25,000 hours is fairly common, and I've seen some quoting 15,000 hours. The lifespan is basically an indication of how tight-fisted the manufacturer is: a shorter lifespan indicates cheaper LEDs overdriven harder than those with a longer lifespan.

I've no idea what, if any, ageing mechanisms apply to a zener, so cannot comment on the comparison.

[wraper]

As implied, why not use each for its designed purpose?

Anyway, the MTBF of Zeners and LED's is available.  Here's what I found:

MTBF has nothing to do with lifetime, it's about failure rate. FWIW you can have devices that always wear out fail after say 10000 hours, but never fail before that, and you say use 8000 hours period for your failure rate measurement, you basically get infinite MTBF.

[moffy]

For any that are interested a discussion on the pros and cons of MIL-HDBK-217 on pages 46-50 with Bob Pease commenting also: https://pearl-hifi.com/06_Lit_Archive/02_PEARL_Arch/Vol_16/Sec_53/Pease_Lab_Notes/Bob_Pease_Lab_Notes_Part_1.pdf

[DavidKo]

LED lifetime is mostly dependent on the temperature of the junction (when there is no silver corrosion or issues with packaging). When you go above 125°C  at which you will get approximately 50000h depending on the producer, LED type etc. the lifetime goes drastically down. Once Philips showed me the LED replacement bulb for cars with 900h lifetime since it had operated around 150°C - I had laughed since the H7 bulb can be bought in 1500h long life version.

That is the reason why LED bulbs are failing. I have disassembled one from Lidl and it was FR4 PCB without heatsink and burns from bottom of the PCB were visible. There was no cooling of the LEDs.

[cosmicray]

LED (commercial) 10^5 hours lifetime
https://www.electronics-notes.com/articles/electronic_components/diode/light-emitting-diode-led-lifetime-lifespan-expectancy-mtbf.php

Thank you for that, as it confirms many things that the manufacturers are less explicit about. One issue tho, when multiple LEDs are operating in series (e.g. home light bulb), my understanding is that the MTBF (for the whole device) is actually closer to LED_MTBF/n, where n is the number in series. String enough of them together, and run them close to the limits, and you end up with a device MTBF much shorter than the LED MTBF.

[Doctorandus_P]

Are you comparing a quality LED with a rejected zener found on aliexpress, or are you comparing a 10W zener diode with a 0603 LED? With that "same current" one of the two can be grossly overpowered.

[tggzzz]

Suppose you have an LED and a Zener with the same voltage drop and both are dissipating the same amount of power. What can you say about the lifetime of each?

You might as well ask "if an orange and a cabbage are picked at the same time and kept at the same temperature, which one will be edible for longer?"

Please do not discuss lettuce, since in the UK that would be political

[SteveThackery]

Thank you for that, as it confirms many things that the manufacturers are less explicit about. One issue tho, when multiple LEDs are operating in series (e.g. home light bulb), my understanding is that the MTBF (for the whole device) is actually closer to LED_MTBF/n, where n is the number in series. String enough of them together, and run them close to the limits, and you end up with a device MTBF much shorter than the LED MTBF.

Correct.  Perhaps a more intuitive approach is to convert the MTBF into a failure rate by inverting it.  A component with an MTBF of 1000 hours has a failure rate of 1 / 1000 = 0.001 failures per hour.  If you have three of these components in series, simply add the failures per hour. Thus you'd get a system failure rate of 0.003 failures per hour.  Invert that again to give a system MTBF of 333.3 hours.

Why on earth bother with failure rates when you can use the formula you suggested (MTBF/n)?  Because that only works when the components are all the same.  Suppose you had a switching transistor, a current limiting resistor and an LED in series.  The MTBFs are:

Transistor: 500 hours
Resistor: 10000 hours
LED: 1000 hours

The failure rates are:

Transistor: 1/500 = 0.002 failures per hour
Resistor: 1/10000 = 0.0001 fph
LED: 1/1000 = 0.001 fph

Now add the failure rates:

System failure rate = 0.002 + 0.0001 + 0.001 = 0.0031 failures per hour
System MTBF = 1 / 0.0031 = 323 hours

Thus you can see that using failure rates makes it very easy to calculate a system failure rate and MTBF, even when the system comprises all sorts of different components.

Two points. Firstly, the MTBFs I've used are ridiculously low and not realistic, so don't quote them!  Secondly, MTBF has nothing to do with ageing or lifespan.  MTBF refers only to random failures.  Items like LEDs, with predictable ageing mechanisms, usually have an MTBF many times longer than the actual lifespan of the LED.  Take incandescent bulbs: they have a predictable lifespan of around 1000 hours (they're designed that way), but it is very rare for a bulb to suffer a failure before it has worn out, so the MTBF is probably orders of magnitude longer than the lifespan.

[uer166]

MTBF has nothing to do with ageing or lifespan.  MTBF refers only to random failures.

+100000 

You can have a device with MTBF of 10 million hours, that fails at the 1-hour mark every. single. time. due to wear-out.