What damages a LED's die in reverse, the voltage or the current?

[RoGeorge]

I would like to improvise a tweezers for testing unknown white SMD LEDs, like the ones found insight light bulbs.

Some SMD models have 2 or 3 series LEDs inside the same SMD package, so they will need about 9..12V + a current limiting resistor when direct polarized.  Other models have only one LED, and the max reverse voltage of a single white LED is usually around 5V.

I don't want to put them under microscope for finding the exact number of LEDs/SMD, or the polarity, so I was thinking of limiting the current at 1 mA, or maybe using a +/-1mA square wave current, from a +/-12V with a 2k2 or so series resistor.

The question is, would it be safe to put 12V in reverse (limited at max 1-2mA) if the datasheet specifies max 5V reverse breakdown voltage?

[SparkyFX]

They behave like a diode in reverse, so imho what kills them is the inability to dissipate enough power. Limiting the current limits the power.

[TimFox]

Note that the voltage across a reverse-biased diode conducting current is much higher than with the same current in forward bias, so the power is higher.

[exe]

My experience with pn junction breakdown is that it does measureably change characteristics of it. At least one can observe increased leakage, or lower beta of a bjt.

Is this a problem for a led? I'm not sure, but I'd do an experiment. I have a bag of leds to try. I'd limit current with something like lnd150 ( https://ww1.microchip.com/downloads/aemDocuments/documents/OTH/ApplicationNotes/ApplicationNotes/AN-D11.pdf ) , though lnd150 is a bit pricey. Any analog guru to suggest a cheaper alternative?  Possibly for lower voltage.

What I did try is repeatedly hitting leds in forward conduction with a much higher voltage (like 12V for a 2V led) at high rate through a current limiter. I didn't observe any changes in brightness. I described some of my findings here: https://www.eevblog.com/forum/projects/fast-cccv-power-supply-again/msg5559957/#msg5559957 .

I'm stil a bit hesitant to claim that it won't damage leds. And it can damage circuitry surround. Or if it's a "smart" led.

[RoGeorge]

Thank you all.

For now I've just look at the white LEDs die under the microscope, they were 3 LEDs in series inside each SMD.  So I've set the generator on 1MHz/20Vpp (+/-10V triangle wave, the AWG already has a 50ohms series output), and tested all the LEDs from the light bulb's plate.  The LEDs that were visibly burnt (under microscope the burnt LEDs have a black spot, or they have the luminofor area heavily cracked) were shining just fine with the bipolar pulses, the ones with burnt dots were not shining at all, and one burnt SMD LED was shining intermitently, like if it were a bad contact inside the SMD.

Tried a few red LEDs, too, single LED per capsule, 5mm TH, and they don't seem to get any damage from the max 10V reverse voltage during the 1-2 seconds of testing.

Not heaving to look at the LED polarity makes testing so much easier. 

[Seekonk]

For many years there was always a spec of 5V for reverse voltage.  I've tested them occasionally for 40 years and have never seen one with less than 170V reverse voltage capability.  Have some you don't care about? Put a 47K resistor in series and plug them into a 120V outlet.  A scope on the 47K will not see any reverse current.  I am willing to concede and I have heard some of the fancy smancy materials have some low breakdown problems.  I just haven't seen it in any of my testing.  Please do some testing, I wouldn't trust the data sheets.  I do high sped strobing.  Amazing the current you can pump in for a short time.

[CaptDon]

My experience with LED reverse voltage failure came about 25 to 30 years ago when we mostly had I.R. and red LED's. I did find that even the voltage of a 9 volt rectangular battery (2U6?) or 12 volts from a GelCell when applied backwards through the typical 1K to 1.8K current limiting resistor typically found in a circuit of that voltage would most definitely harm the LED in some way as to cause a much diminished lumen output when returned to the correct connection polarity. I have not experimented with the newer green, blue, white or U.V. Led's by hooking them up backwards intentionally. I can say of the recently manufactured ones that were momentarily tested backwards by mistake at 12VDC and 1K limiting resistors no damage seems to have occurred. Back in the day when you could buy a package of 30 LED's for around U.S.D. $1.79 at Radio Shack they were ALL misfits and rejects. The die was off center, lumen output was low, wrong lead was the long one, etc. etc. they were easily damaged by reverse voltage / current BUT that was 1975 LED technology and they were duds to start with!!

[Analog Kid]

I'd limit current with something like lnd150 ( https://ww1.microchip.com/downloads/aemDocuments/documents/OTH/ApplicationNotes/ApplicationNotes/AN-D11.pdf ) , though lnd150 is a bit pricey. Any analog guru to suggest a cheaper alternative?

Not a guru by any means, but how about the lowly LM334? One current-setting resistor and Bob's you uncle.

[Whales]

For many years there was always a spec of 5V for reverse voltage.  I've tested them occasionally for 40 years and have never seen one with less than 170V reverse voltage capability.

Woah 170V.  No luck for me, I had a white LED intermittently breaking down at 2V backwards, breaking a keyboard scanning matrix.  Admittedly it was 1 led in more than 100, but that's still an obscene failure rate if it were an intentional diode part.

[exe]

how about the lowly LM334? One current-setting resistor and Bob's you uncle.

Yeah, that would work too.

I just checked pricing, lnd150 and lnd250 are well below $1 a piece in single quantities on digikey. So, my pricing info is outdated, they are cheap enough for me to suggest using it. If bipolar operation not needed, it's just one lnd150 and one resistor. So, kinda lm334, less precise, but much higher operational voltage.

[jwet]

There are a bunch of half truths in all the comments above.  I like to give beginners solid advice that is not some esoteric corner of operation.  Here is some clarification to some of the comments made.

1. The fact that reversed bias diodes have a lot of voltage across them is hardly an issue- the current is near zero until breakdown so power dissipation in reverse biased diodes operated within abs maxes is near zero.

2. The traditional max reverse bias for GaAs (Red) and GaAsAl (IR) is 5v.  This is conservative but expecting anything more than 20V would be very unusual.  The doping profile that defines the PN junction in an LED is very thin.  There are a number of reasons for this but it creates very low reverse bias thresholds.  Rectifier diodes like 1N40XX (up to 1000 PIV) or even small switching diodes like the 1N4148 (200V PIV) are designed much differently.  In order to produce light, LED's are produced with thin doping creating thin depletion regions in reverse bias.  This is so photons can get out of the depletions region to make light efficiently.  This is totally different than rectifier or switching diodes.

3. Charlie Allen (of Maxim) came up with the idea of what is called "Charlieplexing".  A really clever multiplexing techniques for LED's that doubles the multiplex ratios achievable but has one major downside.  It subjects LED's in the matrix to reverse polarity to accomplish its magic.  Not all LED's can be subjected to this but most can and LED suppliers have pushed PIV ratings for LED's for this reason.  The 5V rating is conservative but is likely a worst case process corner type spec for a spec that generally is a don't care or can be worked around easily. (a reversed Si diode across the LED to limit reverse voltage).

4. Spec sheets are real, they're not a conspiracy being perpetrated on developers.  They represent what the designed products are designed to do.  Its not that suppliers can't provide specs like 170V PIV if asked but the spec sheets are real.

5. Often, when you find that you need a part or spec that doesn't exist, it usually means you're on the wrong track.  If you're doing something really unique, contact the supplier, talk to an expert and see if what you want can be done.

The responses here just kind of triggered me tonight- I'll think I'll get off the board and take a break.  Have fun.

[RoGeorge]

(a reversed Si diode across the LED to limit reverse voltage)

Can not have that, because the goal was to make a diode tester that uses AC, such that the DUT LED polarity won't matter for the tester.  This requirement is for easy testing, so I won't have to reverse the probes if a LED doesn't shine at the first try.

Main usage for such a tester would be for individually probing the LEDs found inside the domestic LED light bulbs, they are already soldered in strings, similar with these:



Even more, the Vf can be 3V..12V, because inside each SMD, depending on the model there can be 1 up to 3 series UV LED junctions (all inside the same SMD, and covered with a white phosphorus that converts UV to visible white).

Usually it's not easy to make contact with the already soldered SMDs, that's why the idea of a tweezers tester that will shine the LED no matter the polarity, and such the question about what damages a reversed LED.  Even more, I would like the tester to be a standalone small handheld thing, battery powered, so I won't have to plug anything into an AWG or PSU.

And on top of that, sometimes there are paralleled LEDs too, for example I have a LEDs plate made of 5 series sections of 3 SMDs in parallel, each SMD with 3 junctions (so I have to have enough current to light up a set of 3s3p UV junctions). 


Meanwhile, I did a few tests already with a Joule Thief (a single transistor oscillator, that produces HV spikes), but I don't like that it outputs an asymmetric waveform, which might make the same LED to shine at different intensity depending on polarity:



So I've come up with my own schematic:



This oscillator produces a symmetric AC waveform, so the DUT LED will light the same no matter how the DUT LED is oriented. 

L1 and L2 are standalone inductances, not a transformer like in the Joule Thief.  The exact value for L1 and L2 doesn't matter much.  I've tried with 100uH instead of 3.6mH and it still works.  The ones I've used in the end were 2pcs of 3.6mH, recovered from two broken LED bulbs.


Tested the circuit on a breadboard (so far without all those polarity indicators, only the DUT and the oscillator) and it seems to be working fine.  Tried to probe 2 different types of red LEDs from the 70s-90s (thinking the old red LEDs are the most sensitive to reverse damage) and they seem OK after probing them shortly, for 1-2s.



For probing those 3s3p LEDs found on the LEDs plate in the picture, I had to lower the resistor in series with the battery, even though I didn't connected any of those additional strings of LEDs seen in the schematic, at the left side of the DUT.  Those 18 additional LEDs are supposed to indicate the DUT polarity, and the DUT type (1, 2, or 3 LED junction inside the DUT), depending on which ones are turned off by connecting a DUT.

I'm not sure about the indicators.  Even if I use LEDs with 2 or 3 junctions/SMD, there will still be a lot of extra LEDs to shine.

[Seekonk]

I'd think for anyone worried about reverse voltages, an inductor is the last thing I would want in a circuit.  I don't like designing with inductors because these are items not universally available.  Just a LM358 or TL494 would do.

[RoGeorge]

That's exactly why I was asking, to know how "bad" a voltage spike from an inductance could be for the LED.

I wanted this LED tester to be small, and powered from a single Li-ion cell, so only 3.7V.  That voltage wouldn't be enough to light up a triple junction white LED.  Such a LED would require at least 10-12V to open.  That's why the oscillator is based on inductors, to also raise the voltage, not only to alternate it.

An LM358 or a TL494 can not raise the voltage by themselves either, they would need an inductor, too.


Inductors can be ordered ready made, just like would order resistors.  The ones I've used were recovered from defective light bulbs.  Often, the LED light bulbs have one or two coils inside, a typical value is somewhere between 2-10mH or so.  The exact value is not important, just that the two inductors have to be the same, so to produce a symmetrical waveform.

Same for the transistors, pretty much any small power transistor would work, PNP or NPN (for PNP reverse the battery), as long as the transistors are the same (paired by Beta).  The ones I've used were some unmarked PNP (because I happen to have a handful of these already) chosen for the same beta (mine were heaving \( \beta \) = 215, but the exact value is not important).

Same for the base resistors, any 1k..10k will work, just make them equal so to produce a symmetric signal.  The whole thing is very tolerant about the component values, and very easy to build.  I wanted the minimum number of components so to build it all on the tweezers body, without additional probing wires.

Still looking for what would make a practical enclosure for a LED tester.  I happen to still have on the desk a few vapes I was dismantling last week (those sturdy golden test pins seen in breadbord pic are from discarded vapes:  https://www.eevblog.com/forum/chat/golden-test-poles-from-vuse-go-disposable-vapes/  ).  I'm tempted to squeeze all inside the plastic body of a former vape.


Stil not sure what to do with all those LEDs that are supposed to indicate the number of junctions/SMD and the polarity of the DUT.  They are impractically too many to fit inside a vape.  Maybe I should give up to the idea of indicating the number of junctions (the numbers of LEDs inside a SMD can be seen anyway when the LED is powered at small currents), and keep only 2 triple junction LEDs as polarity indicators, or maybe I should look for a larger enclosureI don't know.

Any ideas how to keep an indicator for the number of junctions, too, but with less LEDs?

[RoGeorge]

Doh, funny news, today I've opened an OSRAM 10W/220Vac LED light bulb which is made with quadruple! SMD LEDs.  Inside each SMD there are 4 LED junctions.  Now I have to make sure the tester can light 4s white LEDs.