what is the reverse breakdown voltage of a red or green 5 mm led ?

[JacquesBBB]

I have searched the Web, but did not find any  thrustful answer.

I put a small 5mm green led as a signal light on a 27 V  AC  power, with a  1.8k resistor, but  forgot to put the diode
that I originally intended to insert in the circuit.

In fact, everything is working just fine, so the reverse voltage of this green led should be larger  than 37 V  (27 x sqrt(2)).
But  if  I look to some recent datasheets,  I see some max reverse voltage of 5 V. This does not make sense. 

So what are the real reverse voltages of the red and green ordinary through hole 5mm leds ?
I should mention that the led I used  was old (it was  already in place), probably more than 20 years old.

Thanks for sharing your experience on this topic.

[T3sl4co1l]

They only ever rate them for 5V because that's all anyone cares about (I suppose it's relevant for muxed arrays driven from +5V supplies).

AFAIK, typical breakdown is indeed in the 30V range, maybe more.  Avalanche breakdown may not cause a hard failure, but I doubt it's as robust as your average schottky or rectifier diode today.

You're welcome to test -- drive the LED with a large value resistor, from a high voltage source (30-100V or more?), see what voltage drop it develops.  Change the resistor and see what other currents do (say, 0.1mA, 1mA, 10mA..).  After a test, check that it still works in the forward direction (low leakage below threshold, same optical efficiency above threshold?).

Typical tests will vary between manufacturers and date codes, so take this with a grain of salt.  Ideally, one would test a whole lot of stuff and publish that data online...

Tim

[Zero999]

Another thing to consider is whether the breakdown is destructive or not. In my experience, old red LEDs seem to be pretty resilient. I've even connected one to the 240VAC mains before without a reverse parallel diode and it worked perfectly, again without damage. Blue LEDs, on the other hand tend to be more sensitive and die when subject to too higher reverse voltage.

What's the forward voltage of your green LED at 10mA? If it's around 2V, that means it's based on older GaP technology and stands more of a chance of surviving over voltage. If it's 2.5V or more than it's almost certainly GaN and is more likely to be destroyed by reverse breakdown.

[Seekonk]

The breakdown voltage has been about 170V for the last 40 years at least for generic ones. You will see Chinese designs that connect them to the mains without a diode for 120V operation.  I've been testing them periodically since then, placing them on the line for a couple weeks with a 47K resistor.  There are some newer ones that have a really low breakdown voltage. Never seen a spec sheet at more than 5V.  A great experiment for a couple of cents.

[wraper]

LED diameter is not relevant at all, color is not much relevant too. What is important, technology it was made with. Modern ultrabright LEDs have completely different structure compared to LEDs made with old technologies. What I can certainly say, modern LEDs are susceptible to ESD unlike old ones.

[JacquesBBB]

Thanks for these inputs.

I will do some systematic experiments, but only this week end, as I have no access to my workshop during the week.

It seems strange to see that the constructors do not put the true breakdown voltage, but an arbitrary 5V  number, well  beyond the  true characteristic of the led.

[T3sl4co1l]

FYI, turning on the 150V DC-DC converter, and using a 220k resistor, I find:

Open circuit: 149.3V (actual supply is 152.6V, so the difference is the ~10M load of the meter).

Various red, yellow and green: they seem to cluster around 30-40V, and >110V (some over 150V).  Probably some die manufacturers / designs have softer doping profiles?

One white (T1-3/4 size): 0.2V.  Hmm, I'd swear it was ~30V for a moment there.  Ah well, it's dead now (blown a microscopic hole in the die -- measures resistive now, ~65 ohms).

YMMV.  This supports the earlier assertion that newer dies may be more sensitive -- namely, GaN (blue, and probably cyan and high-efficiency green (InGaN) also) dies.

Tim

[eblc1388]

Someone would argue placing a 1N4007 diode in series with the LED serves the same purpose as a parallel reversed polarity diode.

Would it protect the LED?

[wraper]

Someone would argue placing a 1N4007 diode in series with the LED serves the same purpose as a parallel reversed polarity diode.

Would it protect the LED?

If you place, say, 47k resistor in parallel with a LED, then yes. If just a diode, and LED in series, probably yes but YMMV.

[capt bullshot]

Someone would argue placing a 1N4007 diode in series with the LED serves the same purpose as a parallel reversed polarity diode.

Would it protect the LED?

Yes, and no. It depends on the remaining leakage current of the 1N4007 and the suspectibility of the LED to reverse voltage ...
With the 1N4007 in series, a small conductance (e.g. 10k Resistor) parallel to the LED will shut the leakage and the LED should be safe (in most of the cases). A antiparellel diode is still better, but the series approaches saves half of the input power (if running off AC).

[MK14]

This 5MM Red Led datasheet (I suspect it is an older type of LED), DOES mention about extra voltages, when reverse biased.

It limits the reverse voltage to an absolute maximum of 6V.
BUT says that up to 10uA (reverse current) can take between 6 and 15 Volts (max), to be reached. (I'm NOT 100% sure on the best way of interpreting the information in this datasheet, so could easily be wrong).

I.e. It is hinting that many/some would/might take at least 15 V and appear to be ok. But still limiting it to <= 6V as an absolute maximum.

http://www.vishay.com/docs/83171/tlur640.pdf

I'm guessing. But presumably above 5V/6V there is a risk of damage (e.g. gradual reduction in brightness level for a given current), which over time would destroy the LED. If the reverse voltage is exceeded.
So it might appear to be ok with a quick bench experiment, but would NOT be a good long term (circuit design) policy.

[Ian.M]

Read: http://www.digikey.com/en/articles/techzone/2012/feb/understanding-the-cause-of-fading-in-high-brightness-leds
then Google: LED degradation reverse bias
My understanding of the situation is that under reverse bias below the avalanche breakdown threshold, the leakage current is concentrated around existing lattice defects in the junction depletion layer and if the bias voltage is high enough, the resulting hot carriers cause the defects to grow.  Its also complicated by reverse bias photocurrent . . . .

Transient excess reverse bias may not do significant damage, but the damage is culminative.  Datasheet max. reverse bias specs. seem to be targeted at multiplexing applications to ensure the design lifetime isn't significantly impacted.

[kosine]

If you check out this video, at 10mins in we start running some LEDs with about 20kV from a flyback transformer:



Surprisingly, they worked just fine, and held their own under reverse voltage - at least until they caught on fire.

Most of the voltage would be dropped by the air-gap spark, but still impressive that they worked at all.

[JacquesBBB]

I made some experiments for testing  various leds

some old ones from old stock that I got ( > 20 years) (OS) (Old Stock)
some other new from cheap chinese source  (N) (New)

I have a high voltage DC source.



I  made a small device to be able to test several leds at once.
They are all fed with the same DC source.
Then each is in series with a red led.
If  the breakdown occurs, the red led  light. I then remove the tested led.

A voltmeter and ampmeter are in the circuit.



The results for the first batch are

10 V  : 5mm green OS
23V   : 5mm red  OS
25V   : 5 mm red OS
39V   : 5mm  Yellow  OS
46V   : 5mm red N
47V   : 5mm red N
48V   : 3mm red N
53V   : 3mm red N
57V   : 5mm blue N
68V   : 3mm green N
73V   : 3mm green N
80V   : 3mm yellow N
84V   :  5mm blue N
96V   : 5mm green N
99V   : 5mm green N
113V   : 5mm yellow OS

Then I  used some other leds to make a more systematic scan fro the same batch of leds

A)  red 3mm  (N)

47V ; 51V; 53V; 53V ; 54 V

so quite consistent, and from the previous test can also be added (merging 3mm and 5 mm)

46V; 47V; 48V; 53V;

B )  green 3mm (N)

81V ; 91V; 91V; 101V; 101V ;

and from first exp ,

68V; 73V; 96V; 99V

C) blue 3mm (N)
27V, 28V, 28V, 47V, 62V

and  57V, 84V from first exp.

D) red 5mm (OS)

21V; 21V; 28V; 30V; 43V;

and from first exp :

23V; 25V;


 And in summary

 red (N)      mean  50.2 V     std    3.2V
green (N)    mean  89V         std  12.3V
blue (N)      mean   50.8V     std   21.7V
red (OS)     mean    27.3V    std     7.7V


So the green are definitely more robust than the red.  But the old red are less robust than the new ones;
Of course, this may well depend  on the kind of leds.

I have randomly check that all the leds are still in  working condition.
I may  test later on  how high they can survive on AC voltage. But this will be destructive .

[madires]

AFAIK the reverse breakdown voltage depends on the semiconductor technology used. I've found something about the impact of geometry on GaInN LEDs: https://www.ecse.rpi.edu/~schubert/Reprints/2009-Cho-J-et-al-(ElectronicsLetters)-Effect-of-chip-geometry-on-breakdown-voltage-of-GaInN-LEDs.pdf

[Seekonk]

Are you sure that supply isn't putting out some nasty spikes when you switch?   I've tested these for years and never seen results like that.

[JacquesBBB]

What are the results you would expect ?

[Seekonk]

I'm old and I get easily confused.  Here is a picture of a red LED in series with a 51.1K resistor on 120V AC power.  This is the voltage across the resistor.  Peak is 165V in one direction, can't detect any "leakage" in the other. Looks like a decent 170V diode. But then like I said, I get easily confused.

[mzacharias]

A little off-point I suppose, but I often use "soft white" 5mm LED's in audio work. I have seen suggestions to add a series 1N4000 type diode in series with the LED and limiting resistor for reverse voltage protection. However, I have tested these diodes to 60 volts reverse (the limit of my DC power supply) and cannot make them fail or exhibit measurable leakage. Since the circuit voltages are generally between 6 and 14 volts, I simply use an appropriate limiting resistor and choose not to worry about reverse breakdown in these applications.

[JacquesBBB]

In order to verify,

I did the same thing with some of the leds that had  the lowest  breakdown voltage.
I blew the first ones, so I did it with the red led (Old stock)
for which I measured a 25 V  breakdown voltage

As for you, I put the scope  over the resistor ( 11kohm)

Here are the pictures. As you can see, it is very consistent with  the 25V breakdown.
(Sorry, I did not properly  put the zero on a  main line).
It looks indeed closer to 22V, but   this method is more precise than the other one for which
I needed a significant current to light the  small red led.


To do  the experiment, I use a variac  on the main. As it is not  isolated, I put
my scope on a UPS that is thus floating.
I also do not use my main scope.  So I  took pictures  instead of   saving on USB.

[JacquesBBB]

Here is an additional test  for  a  recent green led
that  I  measured  for  99V breakdown voltage in the previous setting.

 As you can see,  the  reverse breakdown is around 98V (more easy with the cursors ! ).

I did a few other ones, and everything was consistent with the previous measures.

I made  incidentally also a few mistakes,  making some short cuts when I forgot that although I was
isolated from ground,  all the ground leds of my scope were connected together.
I knew it,  but  forgot while adding a second probe. !!

[guenthert]

I found the breakdown voltage to be around 80V for a random 5mm red LED.  Below cmds and results (A) I sent to and get from a Keithley 617 electrometer (with voltage source):
2017-01-20 10:26:32.236711 -1.83953E-11
cmd: V-78.00X
2017-01-20 10:26:35.160159 -0.29327E-10
cmd: V-78.50X
2017-01-20 10:26:37.688264 -1.66433E-11
cmd: V-79.00X
2017-01-20 10:26:40.613103 -0.25785E-10
cmd: V-79.50X
2017-01-20 10:26:44.084949 -0.30480E-10
cmd: V-80.00X
2017-01-20 10:26:46.427084 -1.19456E-10
cmd: V-80.50X
2017-01-20 10:26:48.975342 -0.45861E-05
cmd: V-81.00X
2017-01-20 10:26:50.546717 -0.53977E-05
cmd: V-81.50X
2017-01-20 10:26:52.098021 -0.61699E-05
cmd: V-82.00X
2017-01-20 10:26:53.653162 -0.69818E-05
cmd: V-82.50X
2017-01-20 10:26:55.205174 -0.77627E-05
cmd: V-83.00X
2017-01-20 10:26:56.768522 -0.85718E-05
cmd: V-83.50X
2017-01-20 10:26:58.321112 -0.93233E-05
cmd: V-84.00X
2017-01-20 10:26:59.900374 -1.00866E-05

There's some noise and probably some leak currents, but (here) between 80 and 80.5V the current jumps nearly 5 orders of magnitude (!), due to the avalanche effect as far as I understand.  The LED was in series with a 500kOhm resistor to limit the current.

For a 5mm green LED I found the breakdown around 89V and not quite as abruptly:
2017-01-20 12:16:51.194313 -0.68180E-10
cmd: V-88.00X
2017-01-20 12:16:53.885743 -0.42249E-10
cmd: V-88.50X
2017-01-20 12:16:55.985707 -1.15087E-10
cmd: V-89.00X
2017-01-20 12:16:57.742533 -1.52163E-08
cmd: V-89.50X
2017-01-20 12:16:59.557428 -0.27204E-06
cmd: V-90.00X
2017-01-20 12:17:01.130629 -1.03605E-06
cmd: V-90.50X
2017-01-20 12:17:02.681539 -0.22200E-05
cmd: V-91.00X
2017-01-20 12:17:04.234108 -0.33205E-05
cmd: V-91.50X
2017-01-20 12:17:05.797317 -0.42624E-05
cmd: V-92.00X

However I did find a red 5mm LED for which the breakdown voltage is only around 5V:
cmd: V-1.00X
2017-01-20 12:05:21.977676 -0.39669E-11
cmd: V-1.50X
2017-01-20 12:05:23.527544 -1.09342E-11
cmd: V-2.00X
2017-01-20 12:05:25.078073 -0.88612E-10
cmd: V-2.50X
2017-01-20 12:05:27.270921 -0.32288E-09
cmd: V-3.00X
2017-01-20 12:05:28.840573 -1.23969E-09
cmd: V-3.50X
2017-01-20 12:05:30.389328 -0.36609E-08
cmd: V-4.00X
2017-01-20 12:05:31.941378 -0.61346E-08
cmd: V-4.50X
2017-01-20 12:05:33.502699 -1.11485E-08
cmd: V-5.00X
2017-01-20 12:05:35.053855 -1.94957E-08
cmd: V-5.50X
2017-01-20 12:05:36.624176 -0.36391E-07
cmd: V-6.00X
2017-01-20 12:05:38.199852 -0.62728E-07
cmd: V-6.50X
2017-01-20 12:05:39.771530 -0.93276E-07
cmd: V-7.00X
2017-01-20 12:05:41.319744 -1.68114E-07
cmd: V-7.50X
2017-01-20 12:05:43.356505 -0.23189E-06
cmd: V-8.00X
2017-01-20 12:05:44.926929 -0.33065E-06
cmd: V-8.50X
2017-01-20 12:05:46.475530 -0.31177E-06
cmd: V-9.00X
2017-01-20 12:05:48.034880 -0.59507E-06

As others have already stated, the breakdown voltage will depend on the materials used and is hopefully stated in the specification.  Just grabbing a random LED out of a bin might lead to surprises.  Those two LEDs looked quite similar, one (the one with the >80V breakdown voltage) being a little bit clearer and brighter. 

But I guess the moral of the story is, if the breakdown voltage is not specified, each individual specimen must be characterized.

[JacquesBBB]

500 k  seems to be a very large value for the resistor.

Will the results be different if  you put 100k ?

Can you test on the same example ?

[guenthert]

500 k  seems to be a very large value for the resistor.

Will the results be different if  you put 100k ?

Can you test on the same example ?

Well, the 500kOhm resistor was just the nearest one.  Meanwhile I got around to retest using a 100kOhm (nominal, but it's less than 1% off) resistor.  I must have confused the test samples (or the data files ;-} as the dark red LED behaves similar to the (light) green one and the light red one is "strange" (see below).  Instead of presenting the wall of text, I spent some quality time with gnuplot and got following:

[JacquesBBB]

Nice plot.

[T3sl4co1l]

I wonder if (actually, I suppose, and suspect that!) the soft knee is due to a finite number of crystal defects, impurity sites, or whatever.

The knee is very much softer than for a proper Si avalanche diode (which is pure enough to handle large surge currents, distributing that current over the whole die area fairly evenly).

It further stands to reason that, if the V(I) curve is made up of myriad (but finite!) defects with different current density (i.e., series resistance) and breakdown voltage, then the curve should be lumpy, at least if you can zoom in close enough to observe that lumpiness, and average it over enough time (since each breakdown site will bleed erratic, noisy current).

Another way to put it: if you vary the current into the junction, the voltage will vary in a noisier way than you would expect from an ideal junction (which would simply have a smooth exponential response).  This reminds me of Barkhausen noise in ferromagnetic materials (where a "pop" is produced as magnetic domains realign across magnetically-hard "stuck" points; though that's driven by hysteresis, so not quite the same).

Tim