Reliability of LEDs: PF Members' Views and Experience

[OmCheeto]

Just thought I'd provide some "outlier" data points:

1. None of my LED "bulbs" have failed.

Numerous LED Christmas light strings have failed. Though, from my experiments from yesterday, it appears it was just the fuses that failed. I just looked at where I think the fuses should go, and still can't figure out how to get to them, without hurting myself.​

2. The only new car I ever bought is now 8.5 years old, has 55,000 miles on it, and the only components that have failed so far have been:

1 fuse
1 tail light
2 tires
​

That is all, for now.

 

[Nik_2213]

"One thing I've done that seems to help is most totally enclosed fixtures have insulation on the top, inside surface. Ripping it off helps dissipate heat better."

Please, please be careful: down-lighters etc may use such insulation for fire resistance per 'Building Code'. Not from the light fitting, mind, but from any unrelated room fire. The insulation delays spread through the fittings to the void above and rooms beyond. If a fire investigator discovers you have compromised *any* fire-breaks, then your insurance cover may be voided or significantly diminished...
===
Cars...
I've only had the good fortune to buy a new car thrice in my life. First time, back in the 80s, I had to 'run it in' for several weeks, then get it serviced. In 2003, I was told, "Just take it easy for a few weeks to let the engine settle down." 18 months ago, when I asked about such, the delivery team shook their heads and said, "Go easy for a few days until the 'stubble' wears off the tyres. Then bring it back in two years, 50,000 miles or a system warning, whichever comes first..."

 

[sophiecentaur]

Just thought I'd provide some "outlier" data points:

1. None of my LED "bulbs" have failed.

Numerous LED Christmas light strings have failed. Though, from my experiments from yesterday, it appears it was just the fuses that failed. I just looked at where I think the fuses should go, and still can't figure out how to get to them, without hurting myself.​

2. The only new car I ever bought is now 8.5 years old, has 55,000 miles on it, and the only components that have failed so far have been:

1 fuse
1 tail light
2 tires
​

That is all, for now.

You have just been lucky I think - or else you have always bought top of the range versions. I checked my mains volts today and yesterday and the value in my shed is 232V, fairly steady. I read someone (on their thread I think it was) that the design operating voltage can be 220V. The nominal UK volts can be as high as 240V. There is just a little less headroom to deal with the odd spike of volts. Clearly the life of filament bulbs would be significantly less for nearly 10% higher volts (V^-16 law)

I think it's only a mains voltage problem, though. 12V LEDs on my boat ran and ran.

 

[OmCheeto]

You have just been lucky I think - or else you have always bought top of the range versions. I checked my mains volts today and yesterday and the value in my shed is 232V, fairly steady. I read someone (on their thread I think it was) that the design operating voltage can be 220V. The nominal UK volts can be as high as 240V. There is just a little less headroom to deal with the odd spike of volts. Clearly the life of filament bulbs would be significantly less for nearly 10% higher volts (V^-16 law)

I think it's only a mains voltage problem, though. 12V LEDs on my boat ran and ran.

I currently have 2 experiments in progress, with another completed.

edit: ps. Any idea how to describe diodes "mathematically"? Never mind. I looked it up. It was off topic anyways.

 

[Tom.G]

ps. Any idea how to describe diodes "mathematically"?

Find a graph of diode characteristics you like and make a table of x,y co-ordinates of the curve at pertinent points.

Then visit this page and enter the co-ordinates. The result is a least-squares polynomial fit of the degree you choose; probably about 6-degree for a good diode fit. I've downloaded the page and run it locally.
https://arachnoid.com/polysolve/index.html

 

[OmCheeto]

Find a graph of diode characteristics you like and make a table of x,y co-ordinates of the curve at pertinent points.

Then visit this page and enter the co-ordinates. The result is a least-squares polynomial fit of the degree you choose; probably about 6-degree for a good diode fit. I've downloaded the page and run it locally.
https://arachnoid.com/polysolve/index.html

It doesn't seem to work.
Anyways, it's a bit off topic, so I'll drop the question.

I currently have 2 experiments in progress, with another completed.

Nothing really interesting from my 3 experiments.
1. I chopped up one of my failed old Christmas light strings and found that it was either a blown fuse or a single lamp failure, as the 5 bulbs I tested were ok.
2. I hooked up a string of 4 to my car battery, and although they draw almost exactly the same power, (73 mw AC/lamp vs 79 mw DC/lamp), the DC powered lamps are much brighter. The AC lamps have been on continuously since I bought them, approximately 10 years ago, yielding a lifespan of 88,000 hours. Since the entire string uses only 2.2 watts, I use them as night time navigation lights throughout my house. (There are 3 sets: 1 each for the kitchen, living room, and bathroom) Total annual cost: $7. My guess is that this may have extended the life of my remaining CFL's, as I almost never need to turn them on.
3. I assumed the string of 4 lamps hooked to my battery, being much brighter, were drawing more current, so I was going to leave them on until they failed. Though, I now know that isn't the case, and they will probably outlive me.

You have just been lucky I think - or else you have always bought top of the range versions.

I just checked 2 that are easily accessible and they are both made by "FEIT Electric".
hmmm...
wow.
I think I did get the high end one: $8.21 at Amazon (https://www.amazon.com/dp/B00B7SB0B2/?tag=pfamazon01-20)
vs
a 10 pack for $13 at the local hardware store (https://www.homedepot.com/p/Feit-Electric-60W-Equivalent-Warm-White-A19-LED-Light-Bulb-Maintenance-Pack-10-Pack-A800-830-10KLED-10/206675105)

 

[NTL2009]

Just thought I'd provide some "outlier" data points:

1. None of my LED "bulbs" have failed.

Numerous LED Christmas light strings have failed. Though, from my experiments from yesterday, it appears it was just the fuses that failed. I just looked at where I think the fuses should go, and still can't figure out how to get to them, without hurting myself.​

... .​

​

I only have a few LED bulbs, and not for long so nothing useful to report there (no failures yet though). Although they are finicky with my CFL/LED rated dimmers. I found I need a filament bulb on the circuit to dampen the wave, or provide more load or something, or else they sort of randomly modulate the brightness (not a repetitive 'flicker'). They sell a dummy power resistor (not sure what value it is) to add to fixtures for this purpose, which of course is ironic, as that is hurting efficiency.

As far as Christmas lights, I started buying those about 6 years ago, and have now replaced almost all the old miniature filament lights, and have had ~ 8 strings of 50 in use for at least the past 4 years. Up to ~ 15 strings now. Not a single failure, not a single bad LED, no bad connections - perfect record (and these are cheap $5/50-string from Walgreens. Not quite as warm-white as I'd like ( a little greenish), but pretty good.

With the filament strings, I would replace each bad/bypassed bulb each year to avoid the extra voltage passed through to the rest of the string. I'd try to find the open bulbs, but would give up half the time before I could isolate it (I suspect there were two opens sometimes). So the LEDs have made the holiday season more relaxing - I like them, just wish I could count on buying ones that were consistently the kind of warm like the filament bulbs.

Also, if you are connecting LED strings to a battery, are you using a current limiting resistor? That's important, LEDS are current devices.

 

[OmCheeto]

...
Also, if you are connecting LED strings to a battery, are you using a current limiting resistor? That's important, LEDS are current devices.

Nope. Direct connection.
I just ramped up the voltage to 14.4 volts, and the bulbs survived!
But wow, look at the power difference:

0.072 amps @ 14.4 vdc across 4 bulbs = 1.04 watts =
260 mw DC/lamp
vs
73 mw AC/lamp​

compared to my previous experiment:

0.025 amps @ 12.7 vdc across 4 bulbs = 0.32 watts =
79 mw DC/lamp​

btw, the equation I found on the internets yesterday: I = I0(e^((qV)/(kT))-1)
yielded a current of 9E51 amps for my problem.
Seems a bit high.
As I've said before, I have no idea how diodes work.
Might be about time I figured them out.

 

[NTL2009]

Nope. Direct connection.
I just ramped up the voltage to 14.4 volts, and the bulbs survived!
But wow, look at the power difference:

0.072 amps @ 14.4 vdc across 4 bulbs = 1.04 watts =
260 mw DC/lamp
vs
73 mw AC/lamp​

compared to my previous experiment:

0.025 amps @ 12.7 vdc across 4 bulbs = 0.32 watts =
79 mw DC/lamp​

...

As I've said before, I have no idea how diodes work.
Might be about time I figured them out.

Well, your experiment demonstrates a lot about how diodes work in a circuit. To get your head around it, stop thinking about voltage across them and think about current through them. Their voltage is sort of a secondary "side effect" of the current. We call them "current devices" because that is the primary consideration in a circuit design.

So you see with 14.4 V across 4 LEDs, you were forcing ~ 3.6 V across each one, assuming they were equally matched, but they won't be exactly. At 12.7 V you seem to be close to what you got under their (normal I assume) AC condition, so ~ 3.175 V across each LED.

You observed ~ 1.13x voltage increase resulted in ~ 2.88x current increase. This is why we control current (with a current source - either active, or just a resistor and a higher voltage source) when we drive LEDs, rather than try to control Voltage. Just let the voltage fall where it will.

And this is why we run LEDs in series, not parallel. In series, they all have the same current through them. But two similar LEDs will have slightly different voltages across them at that same current. So if one LED is 3.1V @ 25 mA, and another is 3.2 V @ 25 mA, and you place them i parallel, the lower voltage one will "hog" most of the current. The 3.2 V @ 25 mA LED will draw far less current at 3.1 V (you saw that in your experiment - a slightly higher voltage equates to a much higher current.

Current limiting is the norm for driving LEDs. Some cheap/simple circuits may rely on a voltage source that has high internal resistance (like a small coin cell), so is self limiting. If you pushed that voltage across the 4 LEDS much higher than 14.4 V with a "stiff" power supply, the current would climb and climb, and it would not be much higher before the LEDs were damaged. A spec sheet will tell you the max current, if you can ID the LED.

 

[NTL2009]

Design example:

Say you have an LED like above, that you know will be approximately 3.1V @ 25 mA current, and you want to drive it at that 25 mA point.

Say you have a 5 V supply, and you will use a current limiting resistor in series with the LED to set the current. Take the 5 V minus the 3.1 V across the LED, and you will have 1.9 V across the R. Solve for R = E/I; R = 1.9/.025 = 76 Ohms. So you choose a standard value of 75 Ohms.

If some LEDs are 3.0 V @ 25 mA; and some are 3.2 V @ 25 mA, the current and power won't vary all that much. A close estimation would be 2V/75 Ohms ~ 26.7 mA and 1.8V/75 ~ 24.0 mA. It's slightly more complicated, but that's close enough for any design work I've ever seen (where LEDS are just indicators). The detail comes in as the slightly lower current will also result in a lower voltage across the LED, so you would get a higher drop on the R, so the current would actually drop a bit less - but this is minor, and works in our favor anyhow.

A higher source voltage, along with a higher resistance, makes the source appear as a more ideal current source, and those little effects are reduced even further. IOW, you will get far more LED current/power variation trying to drive an ~ 3.1 V LED with a 3.5V source and lower R, than with a 12 V source and higher R. But that also wastes more power across the R, sometimes not a big concern with indicators, but that is why active switching circuits are often used to drive LEDs with a constant current.

 

[Baluncore]

btw, the equation I found on the internets yesterday: I = I0(e^((qV)/(kT))-1)
yielded a current of 9E51 amps for my problem.
Seems a bit high.

You are wrongly assuming the LED is a forward biassed Si PN junction. The voltage across a diode depends on the doping chemistry and the mode of operation. For example a zener has a reverse breakdown voltage mode, while an LED is a forward biassed diode with a bandgap providing the voltage drop. The wavelength of the light produced by an LED is a function of the chemistry. The bandgap voltage determines the energy in eV, of the photons emitted when an electron falls over the gap. V = 1239.842 / λnm.
A grow light or a household illumination white LED will produce a primary violet light that is then “whitened” by a mixture of fluorescent chemicals. Chlorophyll has florescence sites that convert violet light into lower energy electron transport chains. Shorter wavelength LEDs, with higher energy photons have been harder to discover because the energy is often sufficient to break up the chemical bonds.
A greater LED current results in more electrons crossing the LED bandgap and so more photons being emitted.

 

[Nik_2213]

I've a covert 'spike generator' in our kitchen. I think it is the fridge-freezer, as several adjacent 'spike catcher' adaptors have died, while those opposite have been okay. Used to be you could get 'industrial strength' widgets with a toroid and three replaceable varistors, but the modern 'catchers' just seem to have one (1) non-replaceable varistor, so $ 10 a pop...

I've ordered in a job-lot of clip-on ferrites, 7mm & 9mm, will be loading the 'usual suspects' flex leads in the hope they help...

 

[Baluncore]

I've a covert 'spike generator' in our kitchen.

Old refrigerators and cool-rooms with thermostats controlling inductive loads, had a habit of generating noise when contaminated contacts began to arc. Heat annealed the spring and so mechanical hysteresis was lost from the thermostat. There was a time when the first thing to do when hunting RFI in an unfamiliar suburb, was to find the butchers shop. The community benefited when the butcher's refrigerator worked reliably.

Zinc oxide varistors catch voltage spikes, but the energy of each spike consumes part of the varistor. As an occasional spike catcher they work well, but when they are hit with a continuous train of high energy spikes they die more quickly. Heavier more expensive varistors can accumulate more spike energy and so last longer than cheaper units. Varistor lifetime is predictable. If you can measure the spike energy dissipated in a varistor, and accumulate the total, you can predict the failure of the varistor and so schedule replacement ahead of the failure.

 

[jim hardy]

More than anybody wants to know about "Spike Catchers"
http://www.electronics-tutorials.ws/resistor/varistor.html
http://www.littelfuse.com/~/media/electronics/product_catalogs/littelfuse_varistor_catalog.pdf.pdf

 

[sophiecentaur]

Old refrigerators and cool-rooms with thermostats controlling inductive loads, had a habit of generating noise when contaminated contacts began to arc. Heat annealed the spring and so mechanical hysteresis was lost from the thermostat. There was a time when the first thing to do when hunting RFI in an unfamiliar suburb, was to find the butchers shop. The community benefited when the butcher's refrigerator worked reliably.

Zinc oxide varistors catch voltage spikes, but the energy of each spike consumes part of the varistor. As an occasional spike catcher they work well, but when they are hit with a continuous train of high energy spikes they die more quickly. Heavier more expensive varistors can accumulate more spike energy and so last longer than cheaper units. Varistor lifetime is predictable. If you can measure the spike energy dissipated in a varistor, and accumulate the total, you can predict the failure of the varistor and so schedule replacement ahead of the failure.

This makes me wonder how to record those spikes using very basic equipment. It could be better to know my enemy than just to litter the house with surge suppressors and see if the frequency of casualties goes down. I could use a cheap DMM - set to Pk and give it an external supply but would it catch ms spikes, I wonder?

 

[Nik_2213]

I've now fitted three clip-on ferrites to the fridge / freezer's power lead, ahead of a spike-catcher adaptor coming 'soon'. While prowling with the bag of widgets, I've also fitted one to the clothes-iron lead close to its plug, and one on the lead to the 'mini larder-fridge' by the 'breakfast bar'...

Um, could you monitor the switching spikes non-invasively by wrapping a wire antenna around the power lead feeding a 741 amplifier and bar-graph voltage indicator such as an LM3914 ? Or all by op-amps, such as http://circuitclassics.com/bargraph-voltage-indicator.html

IIRC, powerlines may be monitored by feeding them through a toroid along with a few turns of a 'sensor' winding. Rather than an 'industrial strength' clip-on split toroid, a loose-fit, clip-on ferrite may do as a 'proof of concept'...

{ Sorry, getting hard to type as I now have three of our cats parked on desk in front of keyboard and/or screens... ;- }

 

[sophiecentaur]

@Nik_2213 I have an oscilloscope for just looking at spikes but I want to be able to capture them when they actually happen. It could be any time of the day or night. Also, I don't really want to launch out on a 'Project' to achieve this. Building electronic circuits doesn't appeal enough to make it worth while of I would find a way to do it with a Raspberry Pie or equivalent. I guess I will have to put up with LED failures or become an amateur constructor once more, after several decades.
Enjoy the Cats!

 

[Nik_2213]

You're right, of course !
{ I gave up on electronics after my eyes became 'middle-aged' and solder went 'unleaded'... }

Would this help ? : http://spikehound.sourceforge.net/

 

[jim hardy]

Hmmm. Tying the spikes to some event would be the next step.

Idle speculation here
I wonder, would a MOV "Spike Eater" in series with an old timey photography fashbulb announce a spike.? MOV would absorb the energy and flashbulbs are wrapped in plastic to discourage explosion. . I see flashbulbs occasionally in junk shops with antique cameras ..

I haven't tried it.. You seem practical enough to not make snything dangerous.

old jim

 

[Baluncore]

Turn a new light on-off-on-off-on-off quickly when you first get it, then do it slowly. If it survives that transient then the internal surge protection is probably working. If it fails, return it as faulty.

Clipping a magnetic core onto the power lead will only reduce common mode noise. Differential line noise will remain. To eliminate differential mode noise you must split and oppose the direction of the line pair where it passes through the core, you should also terminate the line on both sides with an RC snubber.

Lightning strikes and the accidental contact of 22kV power lines with one of the three 230V phases below should be eliminated by the electricity authority. They are well insured against damaging your property.

Turning on a resistive device such as a filament lamp or heater that better terminates the line will have an effect on the propagation of RFI and on the amplitude of standing waves. The network is a star of twisted pair lines that radiate outwards from the fusebox throughout the building. Adding an RC snubber to the line can also reduce standing waves and resonance without such a high power consumption.

I built and still use a passive twin-T notch filter to eliminate the power line fundamental and so show the harmonic distortion and the phase of the differential noise on an oscilloscope, synchronised to the phase of the AC line voltage.

I also built a directional coupler into a short extension cord. It was not needed to find RFI sources so much as prove which way the RF energy was travelling, and that the alleged cause was not the problem. Indeed, the device blamed was often better terminating the line at RF and so reducing the RFI.

Instead of blaming spikes you should question the temperature of the semiconductors and the on/off switching transient. After being called on many occasions to hunt the source of “spikes on the supply”, I was always able to identify the real cause of the problem. It was never the mythical magical spikes that were the problem. Blaming spikes for the death of LEDs or MOVs is a bit like witch hunting. Spikes get the blame for much of the vandalism, but they are rarely given a fair trial before being found guilty.

 

[sophiecentaur]

Instead of blaming spikes you should question the temperature of the semiconductors and the on/off switching transient. After being called on many occasions to hunt the source of “spikes on the supply”, I was always able to identify the real cause of the problem. It was never the mythical magical spikes that were the problem. Blaming spikes for the death of LEDs or MOVs is a bit like witch hunting. Spikes get the blame for much of the vandalism, but they are rarely given a fair trial before being found guilty.

The temperature factor may be (probably is) relevant to the short life. But that is no excuse and a half decent 'replacement' LED should survive in the conditions that its ancestors have tolerated for years.
You are right about the 'spike' monster but there are some badly behaved pieces of electrical equipment around in houses - despite modern manufacturing (and import) standards.

You're right, of course !
{ I gave up on electronics after my eyes became 'middle-aged' and solder went 'unleaded'... }

Would this help ? : http://spikehound.sourceforge.net/

Thanks for the idea. Yes, of course, there are dozens of digital oscilloscope modules which can be used with a laptop - I had just not thought along those lines. I might investigate the market. Up until now I have used an ageing analogue 'scope, which is very limiting in some respects.

 

[Nik_2213]

Went to fetch a 'frozen meal' out of *that* fridge-freezer, discovered top tray was seriously iced-up, whole needed defrosting the 'old fashioned' way...

While thaw was in progress, I took the opportunity to swap the 'standard' plug-top for one of the 'spike catchers' recently arrived from A*z*n. To my horror, I found the 'hot' lead's binding terminal's machine-screw was loose enough to wiggle. That can't have helped the noise-level on the power line !

I didn't swap the plug soon enough to determine if binding post had been 'running hot'. There was no obvious discolouration from arcing or scorching on post or hot-lead's 'bootlace' ferrule, but I only peered at them, did not use a hand-lens...

 

[jim hardy]

Yes, of course, there are dozens of digital oscilloscope modules which can be used with a laptop - I had just not thought along those lines.

Hmmm. I'm in the US. If somebody made one that put out NTSC video (analog television) it'd keep a lot of old TV's out of the landfills..
Did UK television go digital too ?

 

[Svein]

Hmmm. I'm in the US. If somebody made one that put out NTSC video (analog television) it'd keep a lot of old TV's out of the landfills..
Did UK television go digital too ?

I doubt you would be satisfied with the quality. NTSC resolution is 640x480 on a 4:3 TV. A cheap laptop has 1366x720...

 

[Guineafowl]

Did UK television go digital too ?

Yes, but you can buy digiboxes (digital -> PAL) and so still use your old CRT TV. We have a little old Trinitron still in service, although the colours are a little ‘off’ at times and it requires a manly thump to get them back again. When I get time, I’ll be looking for some advice on here on how to correct this properly and give the device a service, while there are still CRT people around who remember how to do such things...

 

[Baluncore]

When I get time, I’ll be looking for some advice on here on how to correct this properly and give the device a service, while there are still CRT people around who remember how to do such things...

The adice you will get is to recycle the old CRT, then use an ex-computer LED monitor with a second hand digital STB. You get a huge reduction in power consumption and a huge improvement in resolution.

 

[Guineafowl]

The adice you will get is to recycle the old CRT, then use an ex-computer LED monitor with a second hand digital STB. You get a huge reduction in power consumption and a huge improvement in resolution.

Yes, that sounds like the sensible solution. But I was thinking more in the spirit of keeping an old device alive.

 

[sophiecentaur]

Did UK television go digital too ?

Yes. Years ago. The terrestrial broadcast service was replaced over a fairly short time with FREEVIEW, which is free and replaced pretty much all conventional CTV services.
The switchover was helped a lot by a market flooded with flat screen sets. CRTs died very quickly. (No great loss except for a few lovers of old equipment. )

 

[Guineafowl]

Yes. Years ago. The terrestrial broadcast service was replaced over a fairly short time with FREEVIEW, which is free and replaced pretty much all conventional CTV services.
The switchover was helped a lot by a market flooded with flat screen sets. CRTs died very quickly. (No great loss except for a few lovers of old equipment. )

Er, yes - my love for old equipment is based largely on its repairability and build quality.

Example 1 - my mate’s newish Bosch dishwasher with a PCB fault that renders it uneconomical to repair, despite costing £300-odd. My old Candy unit is 30+ years old, is largely electromechanical, and still has parts available.

Example 2 - the Trinitron has outived a few flatscreens and then some. The chief advantage of the latter being that they take up less space in the skip/dumpster.

Example 3 - My parents have a microwave that is at least 40 years old, and has always been in daily use, in contrast to the crappy units I’ve thrown away after a few years.

 

[jim hardy]

I doubt you would be satisfied with the quality. NTSC resolution is 640x480 on a 4:3 TV. A cheap laptop has 1366x72

Oscilloscope display accuracy is a percent or three , CRT will do fine for that. Precision is in the numbers the computer writes on the screen.

Example 2 - the Trinitron has outived a few flatscreens and then some. The chief advantage of the latter being that they take up less space in the skip/dumpster.

I too have an old Trinitron, a 26 inch that's too heavy for me to move anymore. But it's convenient to the workbench..

Example 4 - the old Trinitron isn't "Smart" so it won't collude with the Russians to build a dossier on me.