How do you prove an LED is actually lit?

[Infraviolet]

Just a thought, regarding the original post's LDR idea, there is a way this could be done.

If the LED were made to flash at a very high (human imperceptible) frequency then a phototransistor (LDR wouldn't be fast enough) or photodiode circuit could be set up facing the bulb with bandpass filtering so it would only amplify signals at the expected frequency of the LED's flicker, this would be immune to other light sources so long as they didn't also match that flicker rate or an integer multiple of it. That said, with the LED "bulb" in question apparently running off AC or DC it likely contains internal circurity to smooth things out, the only way you'd be able to get a very fast flicker would be to redesign those internals, unless the bulb already has a naturally existing one which you might detect, and if it does have a naturally existing one then depending how "smart" the design is this frequency might adjust depending on things like how hot the bulb gets as it is left running.

Thermal detection on heatsinks probably isn't good enough, there is substantial time lag between bulb startup and sink heating, and between bulb failure and sink cooling.

In the end, I'd look to design for yourself as much of the unit as you can, so you can have things like fast flashing done within it controlled by your own electronics, rather than having the whole 12V unit being a commericla item with inbuild controls. With low enough duty cycles you could even do occasional ultra-fast flashes when the bulb is supposed to be off, and still detect if these fail. That is to say you could detect a failed LED when the LED is desired to be off, not only be able to detect a failure when it should be on but isn't. Knowing it has failed while it is off and knowing that before you need it to be on soudns useful to me.

[james_s]

That's not a bad idea, LEDs can be pulsed at hundreds of kHz, even MHz, and you could use a high duty cycle to minimize any side effects. Phototransistor or photodiode detector, it's getting hard to even find LDRs anyway. ROHS banned them because of the cadmium IIRC.

[Marco]

If the photodiode is right next to the LED it is pretty bright relatively speaking. Don't want to look into a lighting LED from a couple mm.

[Infraviolet]

One can always put some darkened filtering material between the bulb and photodiode/transistor to stop the sensor going in to saturation. Particularly important would be ensuring it isn't exposed to enough light to saturate it from non-LED sources (daylight and such), as if this saturates it then it won't be able to detect a signal from the LED that can then be found by bandpass filtering.

[T3sl4co1l]

It should be straightforward enough to set the sensitivity and orientation so that -- either the LED is lit, or it's inundated by such intensity that you won't be able to read it by eye anyway!

Like I said, LED chips are about the luminosity of the sun; it takes a monumental source, or more concentration than could ever be accidental (from specular reflections, refraction of glass/water/etc.), to fool that.

Tim

[Brumby]

..... unless they mean an indicator on a control panel somewhere.

That's what I was thinking.

Railway signals are traditionally relatively low wattage incandescent lamps with very directional optics. The train is on a track so you know right where it's going to be and can aim the signal accordingly.

Fair point.

Perhaps the Op could clarify. (or am I just showing my ignorance?)

[T3sl4co1l]

Point of reference, I don't have any photodiodes handy but a T1-3/4 red (clear lens, superbright) measures 1.58V (OC) and 0.26mA (SC) under about 1W of white light (high eff. white LED).  (In comparison, the forward-bias drop is 1.56V (~0.1mA?).  It's photovoltaic-ing pretty hard I would say...)  It won't take much die area to get a proper Si photodiode sinking (or sourcing, as the case may be!) much more current.

I suppose the biggest downside to a self-switched detector is, if multiple diodes in series are needed, they either need dedicated optics, or you lose specificity by trying to focus one LED onto all three.  Upside to dedicated optics is, one lens per diode means one diode per LED in the string, perhaps a redundancy feature.

Speaking of LEDs, upside with them is, a very narrow angle type means it's exactly as sensitive on-axis, and insensitive otherwise.  You still need a fixture/jig/holder to mount a THT LED or photodiode in this way (well, unless you put an SMT PD on flex circuit, which might be nice), but you need one anyway for the lens, so an integrated lens part like this might actually be a good solution.

I'd check the LED's properties in daylight, but, uh, *y'know*.  Also it's been snowy here so I might not get direct sunlight for a while anyway.

Tim

[jmh]

Railway signals are traditionally relatively low wattage incandescent lamps with very directional optics. The train is on a track so you know right where it's going to be and can aim the signal accordingly.

Fair point.

Perhaps the Op could clarify. (or am I just showing my ignorance?)

Exactly right. Traffic lights (for example) need to spread light horizontally across a junction as well as vertically to cater for cars and trucks etc. Railway signal lamps are very directional, low wattage things.

As to photodiodes this may actually be a way forward. These lamps are designed to prevent phantom lighting by having the insides behind the outer lens completely matt black. There is a hole in the centre and then the coloured discs and behind that some glass and the bulb and reflector. A photodiode placed near the bulb would get only very little external light even if the sun were shining at the lamp.

[Brumby]

Today, I learned something - and the logic of it seems so obvious.

Thank you.

[james_s]

This has a lot of information on American railroad signals. https://railroadsignals.us/  I assume there must be similar sites out there for signals used in other parts of the world.

[floobydust]

Sort of a similar application is flame detection/combustion monitoring in large burners. A UV photodiode or more typically a Hamamatsu discharge tube is used with a viewing port on the flame.
To meet functional safety requirements, there is a self-check whereby you turn a UV LED on and off, while watching the sensor's response. This confirm things are not blocked by soot, dirt etc. or that there is a hardware failure. Some implementations use a shutter you can pop up to cover the sensor.

OP would have troubles with the extra hardware and conductors in the cables to the gate arms. I can't see monitoring the lamps as being necessary because there are redundant sets at a rail crossing. Monitoring the LED current & voltage would give a lot of coverage, I've only once seen a red LED fail that almost faked the correct V_F.
Sunlight would be a problem with phototransistor-based monitoring. I'm told most crossing still use primitive relay logic for the lamps/gate arms.

OT: The UVTRON would make an awesome kitchen fire alarm, it can detect a candle flame 20m away, across a house, sunlight and smoke is ignored.

[james_s]

OP would have troubles with the extra hardware and conductors in the cables to the gate arms. I can't see monitoring the lamps as being necessary because there are redundant sets at a rail crossing. Monitoring the LED current & voltage would give a lot of coverage, I've only once seen a red LED fail that almost faked the correct V_F.
Sunlight would be a problem with phototransistor-based monitoring. I'm told most crossing still use primitive relay logic for the lamps/gate arms.

Nobody said anything about gate arms. From what the OP has said this is for railroad signals, ie traffic signals for the trains, not crossing signals for motorists. Railroad signals are quite specialized, traditionally they were kerosene lamps, then gradually they were electrified, initially using batteries as they were often installed in remote locations.

[IanB]

Nobody said anything about gate arms. From what the OP has said this is for railroad signals, ie traffic signals for the trains, not crossing signals for motorists. Railroad signals are quite specialized, traditionally they were kerosene lamps, then gradually they were electrified, initially using batteries as they were often installed in remote locations.

Indeed this, and from the context I would assume it is for a heritage railway, where funds are likely to be less abundant than at a mainline railway.

[jmh]

Indeed this, and from the context I would assume it is for a heritage railway, where funds are likely to be less abundant than at a mainline railway.

Yes it is a heritage line - mostly our own money or what we can scrounge.

Railroad signals are quite specialized, traditionally they were kerosene lamps, then gradually they were electrified, initially using batteries as they were often installed in remote locations.

And we still use kerosine lamps sometimes where there is no power, although a rechargeable 12V battery and a single LED lasts for several weeks. But those are in some of the mechanical signals.

[jmh]

Just in case it's remotely of interest this is one of the lamps taken out of it's housing.

[floobydust]

OK we got some details. I get they are railway signal lamps, not crossing lamps, and no redundancy as far as a second lamp.
I thought they would be higher power, the bulbs are apparently 200lm 2,000hrs 10V 18W 18S11/1S but not sure what the UK is using. pic related USA has lots of room in the fixtures due to the optics, and maybe a rheostat?

So far monitoring (LED) current is going to tell you a lot, I like that approach. The LED lamp failure mode depends on the # strings and # LED's in a string, as far as what happens when an LED goes open/short-circuit. It is likely a single high power LED to keep the focal point with the old fixtures, like many modern traffic lights they are no longer using arrays of LED's but a single one.

[Nusa]

I'm usually all in favor of LEDs, but in this case it seems like continuing to use a filament lamp is probably the most ideal solution. If you do decide to go with LED then add a photodiode to monitor the actual output. Since railway signals (that I'm familiar with) lack a reflector in order to prevent reflected light from making it appear lit, you should be able to mount a photodiode off to the side where it can see the LED but will not be exposed to direct light from elsewhere. I don't think anything will reach the reliability of failure detection you can get with incandescent, but the LED should be much less likely to fail.

A valid analysis. If it's the expensive bulbs that are the real concern here, perhaps explore adapting the fixture to automotive bulbs. There are many dual-filament bulbs designed for road conditions that should be cheap enough if you stock up this decade. I suspect they might last longer than rated in a stable fixture.

[IanB]

Just in case it's remotely of interest this is one of the lamps taken out of it's housing.

It is of course of enormous interest, since you have started a thread asking about this problem, but you have been strangely starving us of information, making everyone guess and speculate, as you can see above.

Is it perhaps the case the bulbs you are trying to replace are similar to this one?

https://www.proflamps.com/us/en/Products/Applications/Airfield-%26-Railway/p/13800306842296/?currency=GBP

Given that the life expectancy is only 600 hours, I can see why replacing them could become annoying. On the other hand, given that the indicated price is £1.45 per bulb, and each bulb will have to be replaced every 25 days, you will need something very inexpensive as a replacement to provide a reasonable payback on the investment, I would think?

[tom66]

I do wonder how they solve this problem in railway engineering;  LED signals are used, as these have superior visibility to halogen signals, and in theory longer lifetimes.  But it is impossible to verify with a simple voltage/current measurement that they are functioning correctly. So are they just multiply-redundant (3/4 parallel strings of LEDs independently driven) or are they actually monitoring light output?

[AndyBeez]

I suppose a Hall effect current sensor placed around the LED's power supply line can do the sensing? I assume these are watt rated not milliwatt LEDs? Perhaps an ACS712 current to voltage or an INA219 current to I2C bus, with a jellybean microcontroller doing the pony work. A LoRA module would provide status updates remotely. As for the lifespan of the LED, place the MTBF outside of the service and replace period. btw, stage 'PAR lights' provide a lot of light in a small space.

[AVGresponding]

The problem with any current sensing method is that you just can't rule out a failure mode where there's an approximately normal current flowing. The only way I can see meeting the criteria is to directly sense the light output, and even that has its problems. What if the sense circuit fails in such a way as to give a false good reading?

[jmh]

The problem with any current sensing method is that you just can't rule out a failure mode where there's an approximately normal current flowing. The only way I can see meeting the criteria is to directly sense the light output, and even that has its problems. What if the sense circuit fails in such a way as to give a false good reading?

From what everyone has been saying yes I think sensing the light output is the best way here. At the end of the day its about risk management and mitigation and what The Management agree to. All we can do is give options. Now don't shoot me for not building my own but I found a little relay module on a certain auction site which is trigged from a photosensor which will do exactly what we need. At least it's something I can experiment with and only a few quid. And not my money this time.

When I am next there we do have a LED replacement unit from a 'real railway' signal and I'll have a closer look to see how that works. Those are very different signal lamps to our old thing and replace bulbs which are a lot more watts.

[AndyBeez]

What if the sense circuit fails in such a way as to give a false good reading?

Then simply replace all of this clever technology with the Mark One eyeball.

I found a little relay module on a certain auction site which is trigged from a photosensor which will do exactly what we need. At least it's something I can experiment with and only a few quid. And not my money this time.

So how will you notify the signalman that any light source is out of operation? How do you force a failsafe or stop condition on the block?

[jmh]

So how will you notify the signalman that any light source is out of operation? How do you force a failsafe or stop condition on the block?

That's the easy part. LED fails, relay drops, drops the control relays in all following signals which then go to red (their default). Driver stops the train and calls to see what is wrong. Showing anything other than red is always a positive action i.e. a relay is energised to do so.

[PA0PBZ]

On the other hand, given that the indicated price is £1.45 per bulb...

I didn't trust the site: quantity 10, price each £ 14.50. And indeed, clicking the 'add to cart' button presents you with the total of £ 145.00