Suitable reverse protection diodes for white LEDs

[mmagliaro]

This subject has been discussed many times.
White LEDs typically have a maximum reverse voltage limit around 5v,
so if you connect them with the wrong polarity, they don't just "not light" like a normal LED,
but they could be destroyed if you hook, say, 12 volts across them the wrong way.

So when used in a loco headlight, assuming this is just plain DC wiring, you wire a diode in reverse across
the LED so that any reverse voltage is clamped to 0.7, and all the reverse current is ferried around the LED.

My question:
Just how fast does that reverse diode need to be as far as switching time?
I have Schottkey diodes to use for this, which are very fast indeed.  But is that necessary?

Would a simple little 1N4849/1N914 (also considered a "switching" diode because of its fast time)
or some such thing be good enough?   In other words, just how fast does it have to be?

[peteski]

I don't have an absolute answer for you, but I really think that you are over-thinkng this (and I thought that I was bad over-engineerig my projects). 

Any small signal silicone diode (or even another reverse-connected LED used as a diode) will safely clamp the reverse voltage.  If you are really worried then wire the protection diode in series with the LED.  Then you won't have to worry about the response time of the diode.  Either of these types of designs is used in factory-made models.

If you really worried (and no electronic wiz here can answer you), this is a type of a question you might want to ask the LED manufacturers.  I have accidentally applied 9 or even 12 V to blue and white LEDs (the ones very sensitive to reverse voltage) and I have not damaged them.

I really think that the rating given by the manufacturer includes a large margin of safety and that the LED destruction is based more on a reverse current than reverse voltage.  But either way, whether the diode's response time is 1 ns or 10 ns I don't think the LED will care about that.    Good question though.

[mmagliaro]

All true, Peteski.  I used white LEDs without protection and they seemed to work fine.
But I have had some mysteriously just quit working, so I suspect that back emf spikes from
the motor (during moments of intermittent wheel contact) were high enough
to kill the LED.   

Perhaps I am overthinking it.   It just seems to me that if a reverse voltage spike is placed across
the LED, it could burn out before the protection diode could switch on to protect it.  My instinct tells
me that LEDs would be slower than a  diode sold as a "switching" diode, but I have no basis for that.  It's
just instinct.

[peteski]

Lets go back to the theroy.  You are most likely talking about reverse recovery time paramter.
The this parameter is provided for switching diodes for when they are used in high-speed digital circuits.  LEDs aren't considered switching diodes, so that parameter is not usually given in their technical specs.

BUt if you look at what that parameter is http://en.wikipedia.org/wiki/Diode#Reverse-recovery_effect, you have nothing to worry about in your application.

That time is how much longer the diode conducts after is becomes reverse-biased.  So it is not how long it takes the diode to start conducting when forward-biased.  It conducts as soon as the forward-bias voltage is applied and reaches diodes forward conduction voltage.  You don't really care how many nanoseconds the diode will still be conducting after the voltage across it was flipped.

[victor miranda]

Hi Max,
like Peteski, I'd use about any diode to dump the reverse voltage.
I have a bunch of little glass ones I got for this reason.

If part count and price was not in the decision path
I'd add a small cap across the LED ( and protection diode) to
attempt to smooth any possible hi frequency signal ...

I have not seen it is needed however, as recent LEDs are pretty robust.

[mmagliaro]

Okay, boys.  You've convinced me.   The little cheap 1n4849 from my stash goes across the
SP&S E7  headlight  LED.    (yes.... remember that?   It's almost done.)

[peteski]

Okay, boys.  You've convinced me.   The little cheap 1n4849 from my stash goes across the
SP&S E7  headlight  LED.    (yes.... remember that?   It's almost done.)

[tehachapifan]

Interesting topic and one thing that I WASN'T worried about previously.

In some future DCC installs using decoders with only 2 function outputs in addition to directional headlights, I was considering running some LED's for the numberboards straight off the frame (with resistor(s) of course). This would free up the two outputs for effects lighting and the numberboards would stay constantly lit, which would be kind of cool.

Should I install a diode to protect these LED's?

[peteski]

No need. The function outputs of the DCC decoder never flip polarity so there is no need for protecting the LED from reverse voltage.

[tehachapifan]

Thanks, Peteski, but I was thinking about not coming off the decoder for these LED's, so I could save the few 2 decoder outputs for other applications.

[peteski]

I'm not sure what you are going to do.  You want to have a constant lights powered from the DCC signal on the track?  If that is it, you would be better off using a bridge rectifier and powering the LED from its output. Then you won't have to use a protection diode and the LED itself will be constantly on rather than flashing really fast (since DCC signal is a square wave.  The 4 diodes inside the bridge rectifier will do all the rectifying and protecting for your LED.

[mmagliaro]

Yes!
If you put a full-wave bridge in there, and run the LED/resistor from that, it will be nicely on all the time,
and will be protected.

In fact, har har.... that's what I ended up doing in my E7.   I didn't want the headlight to go out in reverse.
So I put a little surface mount bridge rectifier in there and hooked my LED and resistor onto that.

[tehachapifan]

OK, cool. What would be a good size/type rectifier to use for something like this?

And to confirm, going with a resistor alone won't work in the long run?

I can get an LED to light nicely off DCC track power with only a resistor.

Thanks!

[mmagliaro]

A resistor "may" work for the long run, but notice the quotes around "may".  The protection diode
really avoids the possibility that the reverse voltage won't eventually blow up a white LED that is
supposed to only be able to tolerate 5v in the reverse direction, even though many of them seem to be able
to tolerate more.

As for the rectifier...   I bought a stash of "MB6S" which are surface mount, 1/2 amp, 600 volt.
No, you don't need a 1/2 amp 600v rectifier for an LED.  25v and 50mA would be more than adequate.  But they were there on
eBay, and cheap (like 50 of them for a few bucks kind of cheap).    They shipped from China, so
they took a couple of weeks, but I have had very good success ordering lots of parts that way.

When I get my E7 project thread up, you'll be able to see them.   Just hunt around for
"bridge rectifier smd" or "bridge rectifier surface mount" and see what pops up cheap.

These things are only about 1/4" square.

The track power connects to the two "AC" terminals, and the "+" and "-" go to the LED, with a resistor in-line of course.

[tehachapifan]

Great! thanks!