LED failure detection circuit

Thread Starter

jolleym2

Joined Mar 1, 2016
7
I am currently working on a project involving a central command module which has two light stacks to display the mode of the machine.
  • The light stacks each have four modules each with a different coloured led inside wired in parallel to eachother.
  • The light stacks are powered by four(one per led) 230Vac wires coming from the command bay connected to an internal power supply, this supplies the LED's with 24Vdc.
  • There is one returning neutral.
  • The lights are 15 meters from the command module
I have a circuit which uses an LDR to detect the LED failure however the project has requested that the circuit that is designed will sit inside of the command module and it will not be convenient to run the 8 wires per light stack from the command module to each light stack.

The initial problem i am struggling with ideas for is the fact that an LED can fail open or short, when it fails open a comparator circuit could be used to detect the voltage drop across the LED, if an LED fails short how can that failure be detected?

The second problem is: can the failure be detected by using the control signals from the command bay to the light stack and the neutral returning wire, or does the circuit require modifications to the light stack such as mounting a circut with inputs from either branch of each LED?
I will upload a schematic later. For reference the light stack is AEUR modul70

Thank you in advance
 

AnalogKid

Joined Aug 1, 2013
10,987
The initial problem i am struggling with ideas for is the fact that an LED can fail open or short, when it fails open a comparator circuit could be used to detect the voltage drop across the LED, if an LED fails short how can that failure be detected?
With a 2nd comparator. Two comparators combined like this are called a window comparator. Since many comparators have open collector outputs, you can assign the signal polarities such that when the voltage across the LED is in the good range, both comparator outputs are open. If either comparator trips, the wire-OR output goes low signalling a fault. Is it important to know which kind of fault it is, or is one good/bad signal enough?

The second problem is: can the failure be detected by using the control signals from the command bay to the light stack and the neutral returning wire,
Maybe, but unlikely. More after schematic.

Web search returned this: We did not find results for: +AEUR modul70. Can you post data or a link?

The AC powered stack lights I found run on 115 Vac (what country are you in) and draw 25 mA. Sensing the neutral might not work because 2 lights on looks the same as 3 lights on with one shorted. If the lights are mutually exclusive, then sensing the neutral would find an open light IFIFIF whatever is doing the AC/DC conversion plays along. Tantalizingly close, but still too many unknowns.

ak
 
Last edited:

Thread Starter

jolleym2

Joined Mar 1, 2016
7
With a 2nd comparator. Two comparators combined like this are called a window comparator. Since many comparators have open collector outputs, you can assign the signal polarities such that when the voltage across the LED is in the good range, both comparator outputs are open. If either comparator trips, the wire-OR output goes low signalling a fault. Is it important to know which kind of fault it is, or is one good/bad signal enough?
The good/bad signal is enough, thank you for that suggestion i will research it further!

Maybe, but unlikely. More after schematic.

Web search returned this: We did not find results for: +AEUR modul70.

Can you post data or a link?

ak
This is the light stack which is used http://www.auer-signal.com/files/productfiles/Modul70 (Leuchtmodul-Eintonschallgeber)_1 2|1.pdf
 

AnalogKid

Joined Aug 1, 2013
10,987
Those things are very tightly sealed. Unless you can drill holes in the plastic to run your own wires, it looks like the only sensing points are the 240 Vac lines and the neutral. You could stick optical sensors on the outside of each light, combine them with a local microcontroller onto a 2-wire power and signal system like a 4-20 mA loop or something custom, and take it apart back at the command bay.

Or, run the four 240 Vac power lines through individual current transformers for a local, isolated signal you can window detect. Again, the window works only for some of the possible failure modes.

ak
 

Thread Starter

jolleym2

Joined Mar 1, 2016
7
Those things are very tightly sealed. Unless you can drill holes in the plastic to run your own wires, it looks like the only sensing points are the 240 Vac lines and the neutral. You could stick optical sensors on the outside of each light, combine them with a local microcontroller onto a 2-wire power and signal system like a 4-20 mA loop or something custom, and take it apart back at the command bay.
Yes unfortunately I believe they're IP65 so I might have to drill holes, the optical sensors may be the best option.
Or, run the four 240 Vac power lines through individual current transformers for a local, isolated signal you can window detect. Again, the window works only for some of the possible failure modes.

ak
This won't work if the LED 's short circuit will it? :(

Thanks for the advice
 

AnalogKid

Joined Aug 1, 2013
10,987
Depends on the power supply in each lamp. Something is turning 240 Vac into a few volts DC. Simple or complex, it is what is actually being measured.

ak
 

WBahn

Joined Mar 31, 2012
29,979
If you can do current sensing, then look for the current to be between some minimum and maximum values. It is very unlikely that an LED would fail yet still draw the same current.

Another thing that might be worth considering is locating LDR (or other optical) sensors on a remote module and then send the data back serially. Depending on the specifics, you could get by with just two additional wires (or even less) relative to what you already have.
 

AnalogKid

Joined Aug 1, 2013
10,987
Yeah, I mentioned that. But my guess is that this is not a production product, and that the project can't support the development cost of a one-off serial data datacomm gadget. Thinking about it, since everything is one-way for only 50 feet, an old 40-pin UART could handle the data with no uC or code development.

ak
 

Thread Starter

jolleym2

Joined Mar 1, 2016
7
If you can do current sensing, then look for the current to be between some minimum and maximum values. It is very unlikely that an LED would fail yet still draw the same current.
This could be a good possibility and wouldnt be as intrusive as other designs, I could place a shunt resistor in series with the load and then use a differential amp to sense current changes.

Another thing that might be worth considering is locating LDR (or other optical) sensors on a remote module and then send the data back serially. Depending on the specifics, you could get by with just two additional wires (or even less) relative to what you already have.
Unfortunately this is my favorite idea but the system is safety critical and there are too many variables that will be raised by h&s about the influence of ambient light.
 

Thread Starter

jolleym2

Joined Mar 1, 2016
7
Yeah, I mentioned that. But my guess is that this is not a production product, and that the project can't support the development cost of a one-off serial data datacomm gadget. Thinking about it, since everything is one-way for only 50 feet, an old 40-pin UART could handle the data with no uC or code development.

ak
The project can cost however much, ideally the cheaper the better however it is time and safety critical as the system has to be implemented within 6 weeks and no PLC or the like can be used due to the "safety critical" nature.

I think my two options are to dismantle the lights and lose the IP rating but then have the ability to attatch window comparators into each light module, or to use the method WBahn suggested with current sensing
 

AnalogKid

Joined Aug 1, 2013
10,987
Again, current sensing where? If you can conduct an experiment, place a DVM in AC Current mode in series with one of the 240Vac lines (or put in a 1 ohm resistor and measure the voltage across it) to get a normal reading. Then short out the LED. Then remove the LED. Those three readings will tell us everything we need to know about the feasibility of reading the current on the AC line without opening and modifying the lights. I like that method because it can be done back at the controller without running any extra wires.

Just winging it here, I think the two failure conditions have to be at least 20% off of the normal reading to be able to differentiate them in an industrial environment accounting for part-to-part variations. If the AC/DC converter in each light is just a series capacitor and a diode, then an open will be obvious but a short will be a current change of only a few percent, probably not enough to see reliably. Is a shorted LED really enough of a possibility to matter?

Good news about the budget...

ak
 

Thread Starter

jolleym2

Joined Mar 1, 2016
7
Again, current sensing where? If you can conduct an experiment, place a DVM in AC Current mode in series with one of the 240Vac lines (or put in a 1 ohm resistor and measure the voltage across it) to get a normal reading. Then short out the LED. Then remove the LED. Those three readings will tell us everything we need to know about the feasibility of reading the current on the AC line without opening and modifying the lights. I like that method because it can be done back at the controller without running any extra wires.ak
I will conduct this experiment and post the outcome, I believe a shunt in series with a window comparator will allow me to detect the failure

Just winging it here, I think the two failure conditions have to be at least 20% off of the normal reading to be able to differentiate them in an industrial environment accounting for part-to-part variations. If the AC/DC converter in each light is just a series capacitor and a diode, then an open will be obvious but a short will be a current change of only a few percent, probably not enough to see reliably. Is a shorted LED really enough of a possibility to matter?

Good news about the budget...

ak
The short is really really unlikely but in the case of having to prove the circuit to the health and safety authority, they would need to know it works under both conditions, its a very unlikely event anyway because the average lifespan of an LED under normal conditions is 57 years!
 

AnalogKid

Joined Aug 1, 2013
10,987
the average lifespan of an LED under normal conditions is 57 years!
All the LED vendors say things like that, but reality is a bit different. As it was explained to me by someone who knows this stuff, that is the basic reliability of the diode chip, one bonding wire, and two leads, all hermetically sealed in resin. When you add just one external connection to a current limiting resistor, the statistical reliability of the circuit plummets. It still is a very large number, but not all those decades. MIL-HDBK-217 is the military's guide to electronic circuit reliability prediction, and it is a cruel master.

ak
 

Thread Starter

jolleym2

Joined Mar 1, 2016
7
Had a small change in the lights which are being used, they now contain two parallel LED arrays, each array is wired in series with 5 LED's. The window comparator circuit will need to be repeated per LED, any ideas of how to reduce the circuit size?
 

AnalogKid

Joined Aug 1, 2013
10,987
Had a small change in the lights which are being used, they now contain two parallel LED arrays, each array is wired in series with 5 LED's. The window comparator circuit will need to be repeated per LED,
Why? The forward voltage of 5 LED's in series will vary a little from one light to the next, but not enough to mask a single-LED short. A +/-0.5 x Vf window still looks good on paper. An LM339 quad comparator will get you two window circuits in one package, one for each array.

ak
 
Top