Monitoring LED Warning Lamps

What is the failure mode of Led lamps? OC ?
Dunno myself, but it's prob a good place to start?
Del

Agreed - but I dunno either. If it could be guaranteed that the only modes were OC and SC it would be easy. Is there any way it could still be drawing normal range current, but not lighting up? That's the question!

I have visions of a huge Heath Robinson device with photo cells to detect if they are illuminated or not...
Mind could just double up on LED+R in parallell with the original LED+R.
Simple is sometimes best.
Del

True, but sadly doesn't meet the requirements of IRR99.

So what happens if the warning light for the warning light fails?

Didn't Dr Suess have a story about bee watchers who needed to have bee watcher watchers behind them and they had 'watcher watcher' watchers behind him and so on and so forth?

Just asking.

The normal catastrophic failure modes of a an LED are just OC and SC. The complication comes in that a working single junction will be relatively close to a SC. So a common technique with any multiple LED lamping is to have strings of LED in series. A SC in one of these LED just makes the rest a little brighter. So if these multiple LED strings are in the same sign then it becomes difficult to see a difference both visually and from an exterior circuit sense. To be a little anal here, a SC will actually be a failure of the wiring in the package and not the junction. Since package failures still produce no light... The junction's catastrophic failure is only an OC. An OC in a series circuit is the easiest thing in the world to detect.

Depending on how long these warning lamps are intended to be in use there can be an aging problem where the efficiency of the lumen per watt decreases. This is a non-linear aging effect that includes time, junction temperature, doping concentrations (not sure about that one) and total charge carried through the diode. It usually takes years for this effect to be found. There are ways to induce this failure in an accelerated fashion for product testing. This failure though is a purely light output problem. If you must sense this failure you'll have to monitor the light output itself.

If you need citations on this I can hunt them up later.

Thanks, Fred - I think the slow degradation of light output can be catered for by visual inspection at the routine services, so it's not a monitoring problem in the same way as sudden failure. Monitoring for one shorted junction in a string would depend on the drive method - if constant current, the voltage would have to be monitored and vice versa.
Monitoring the light output would seem to be the most straightforward, but the monitoring device and circuit would itself have to be failsafe, and it rapidly becomes a case of quis custodiet ipsos custodes? - and starts to get complicated and expensive.
I've implemented monitoring on x-ray systems using filament lamps, and I've used LED warning lamps on gamma exposure systems - but fortunately there's a loophole in the regs that says you only need failsafe lamps for x-ray systems. Now various people want to use the advantages of LEDs (long life, in particular) with x-ray systems, which is where I came in.

I agree that a constant current drive will work the best. My simplest recommendation would be to use the photo-diode of an optocoupler in series with your LED array and then a comparator circuit on the voltage level going to this array. The comparator bit and the opto-coupler bit then looks for plausible failure modes. To cover all possible failures you might consider a window comparator circuit.

Check the Avago (sp?) web site. They have all of the HP optical device documentation and parts. You will probably need documentation to back this up.

good . see #5.

Typically one might find an opto coupler in series with the LED for sensing, or a direct feedback from the driver compared to a signal from the other side of the dropping resistor. but then you would have to check the checker, and check the checker that checked the checker. We have a rule that any operator who suspects that an indicator is not functional or any other part is damaged or compromised, and continues to use the equipment Assumes ALL liabilities for any incident occuring during such use. This has been made clear to the clinicians, and they readily comply. If they balk at it, they usually get a pretty clear letter from the administrative lawyers.

The warning lamps are of course only functional during an error, so they have to be checked during the power-up lamp test routine or reset.

If continuous monitoring is required then a low level below-emission bleed current must be used so there is a constant measuring capability.

The kind of lamps I'm on about are in the illuminated "X-RAYS ON" signs. If they fail, the x-ray generator should be disabled.

Thanks anyway for the contribution.

OK so they are room signs and not part of the equipment? could have said so, OP implied they were system warnings.

The regs. refer to them as "Warning Lights". Sorry I wasn't more explicit - it's jargon I've got used to over the last 30-odd years.

They are a part of the equipment to the extent that the equipment won't work without them!

Now it's much easier. Use multiple leds for your sign and two (or more) photo-diodes to sense the brightness. When the output of the photo-diodes don't agree issue an error code that blocks the operation of the equipment.

Redundancy is built-in, if a led doesn't light the comparator on the photo-diodes issues a fault, if a photo-diode fails the comparator issues a fault, now all you need is a third comparator to monitor the output of the first two.

The possible drawback here is ambient light infiltration. I'm not saying that this is a game stopper but it should be addressed.

Red,

That's what lightpipes and shields are for, to ensure that only the desired light reaches the detector. (Free design stops until the contract is signed)

Alan

The purpose of placing the lamp in-circuit with the lockout is to prevent X-Ray equipment from operating without all protection and monitoring circuits fully functional. The way that the function of the safety light is verified is with a human being. Lamps are meant to be observed, not sensed by a current meter. The loss of the "energizing" lamp or similar safety lamp prevents the actual operation of the equipment.

Simply replacing a filament style bulb with an LED is not going to provide the same amount of protection, and probably will fail the lab upon inspection.

A fully designed lockout system from the X-Ray equipment manufacturer, or your government agency in charge of inspection, should validate any changes to your safety lockout system prior to implementation.

Consider a voltage detector circuit across either the LED or its series resistor.

If across the LED, an OC would increase the detector voltage and a SC would reduce it.

If across the resistor, the opposite would apply.

Pretty easy to then feed this into a comparator and/or amplifier to process as required.

With a bit of isolation you could likely use the one amplifier circuit for all of the LEDs.

I think that, probably, the better way is to monitor the LED driver's output voltage. Because, usually, the LED driver supplies the LED with constant current you can't detect a LED which is -e.g.- short-circuited by measuring the current (as the current will be the same as in normal operation). But, as the path resistance (resistor+LED) is changed (due to LED's failure) the driver's output voltage is changed too. Make an experiment and see in which way this voltage is changed when the LED is OC or SC and then use a "window comparator" in order to detect the LED's failure.

The more I think about it, the worse it gets.
In reality the leds are possibly the least of your worries as you have to cater for any fault e.g cable to the warning light getting severed in a collision with a hospital trolley being used in a game of chasies by a cat and squirrel.

This tends to imply that the sensing needs to be at the equipment end not the lamp end.

Is the warning lamp just one big LED or a string of LEDs? How robust does the check have to be? Very little is entirely fail safe, but if the limit is say:-
To fail and not be detected must require at least 3 independant faults to occur simultaneously then you have some way of designing it.

As usual it's down to having a spec, else you get into a cycle of ... add a rely... what if that fails? Add a parallel lamp what if that fails? etc.
In reality, something along the lines of 3 parallell lamps at the indicator end, and current monitoring at the equipment end (maybe with a fairly narrow good/bad band) would do the trick.

Or just 'one lamp+ one monitor' triplicated and the results of the monitor stuffed into a logic gate at the equipment end....?
Del
(my furry brain hurts now.. can I have tea and toast please?)

Along with others I'd go for the direct approach (detect the actual light). I'd just add that I'd use a length of Plastic Optical Fiber to link the LED to the photo diode/transistor.

Thanks to all for the contributions. A bit of summarizing:

The lights in question are illuminated panels, with the legend "X-RAYS ON".

The lights must be on when the X-ray generator (i.e. the HV power supply for the x-ray tube) is on.

If the lights fail, the x-ray generator must be disabled. This is a requirement of the Ionizing Radiation Regulations 1999 (IRR99)
When filament lamps are used, this requirement is met by monitoring the filament current.

On x-ray diffaction and x-ray fluorescence systems I worked on in the '70s and '80s, a single wire was connected in series with the guard microswitches and the warning lamp filament to a relay coil. The relay enabled the x-ray generator.

The warning lamp panels I intend to use have strings of LEDs, and are sealed units. All I know at present are that they need a 12V supply; I'll need to get onto the manufacturer for more details.

I'll probably go for light monitoring using fibre optics - there's no direct access to the LEDs, only the light emerging from the panel where masking has been removed to form the lettering. There should be enough light coming out so that changes in room lighting won't affect the checking, if the trip levels are set up on-site under normal running conditions. I'll provide a lamp test button so that a visual check can be made (probably have the local rules imodified to include a check first thing every day).
I'll need to add an extra layer of monitoring such that a failure is flagged if the checking circuit indicates that the lights are on when they are, in fact, off.

Thanks again - please chip in with any other thoughts should they arise.

Some points to ponder:-

1) It may be better to use filament lamps again, if they are already fully acceptable. Maybe a slightly higher voltage version that runs relatively cool, bur still gives indication and is easy tested for operation......though I personally am a LED person.

We used this system on early computers, it increases lamp like by a factor of 3 or 4 if done properly. Use "Hardened" lamps to further increase life span. Possibly similar techniques can also be applied to LEDs, high brightness low current ones supplied from a clean supply.

2) Make sure that whatever system is used, that an accurate description of the problem is posted in some fashion (Email, SMS who knows!) so that almost any idiot can replace the problem LED/Lamp without having to search amongst many.....

3) LED/Lamp error cycles the machine down in a fully controlled manner. Not just BANG and its off!!!

4) Maybe have all the LED/Lamps in a single user replaceable "block or Strip" that almost anyone can remove and replace, when powered down and can only be inserted one way round....Trained Monkey?