LED Fault Monitoring

[Kisen]

Hello,

Im new to the forum.

I am working on a project and have become a bit stumped, I could do with some help.

I am working on an LED lighting system. I will try to include as much detail as i can to give you an idea of what i need. Although at this point no specific components have been settled upon, this is all "theoretical" at the moment.

So the LED light will be composed of several individual constant current power supplies, powering a chain of 4 LEDs each.
All 4 LEDs will be the same colour and specification.
The LED System is an essential part of the overall system (a bio reactor) and needs to be kept running optimally.
The constant current power supply will have the ability to PWM the LEDs to control brightness, likely via a dedicated PWM IC.

Now the whole LED unit when completed will be housed inside a enclosure that stops the user from simply seeing a failed LED. So I would like to add in some kind of monitoring system.
I have been looking into LED shunts that should one of these 4 LEDs fail open circuit it will route the power to keep the circuit complete and the remaining LEDs will function.

I need the ability to detect the state of each one of the 4 LEDs (open or closed circuit fail) and feed this back to say a uController for analysis. The system can then decide whether to activate the shunt and keep the system running or shut down that particular PSU for the string and flag for maintenance.

I have looked at window comparators, but right now i feel that component count will be pushed too high for my PCB real estate.
Also though about measuring the voltage at each LED cathode and use a PIC to look for a window voltage (same as the comparator is guess) but im not sure if the PIC would then start feeding each cathode to GND through the PIC.
Ideally the Monitor circuitry should not interfere with the LEDs function at all, almost like it wasnt there to start(if thats even possible)

So my question to you all is...How would you guys go about checking each LED individually for a open/closed circuit fault?

I hope i have included enough to give you a decent picture of what i am trying to do.

 

[ronsimpson]

checking each LED individually

Why?

You have 4 LEDs in series: I think it does not matter which LED is open/shorted, you just need to know that one is bad.
The voltage across the LEDs is constant (more or less). If one of the LEDs is shorted the voltage will drop to 3/4. That is detectable. If one of the LEDs is open the constant current supply will output a much higher voltage. Also detectable.

You could add a Light Detector and measure total output.

 

[Kisen]

Why?

You have 4 LEDs in series: I think it does not matter which LED is open/shorted, you just need to know that one is bad.
The voltage across the LEDs is constant (more or less). If one of the LEDs is shorted the voltage will drop to 3/4. That is detectable. If one of the LEDs is open the constant current supply will output a much higher voltage. Also detectable.

You could add a Light Detector and measure total output.

I did think about monitoring the string of all 4 LEDs. This would work, however it doesnt give me the ability to know which LED has failed in order to trigger the shunt.
If all LEDs have an automatic shunt then i would never see a problem with the voltage, would i?

 

[ronsimpson]

I almost never see a open LED. I do see bad solder joints that cause no current flow.
Do you have data on LED failure? (resent data) Before you spend time and money on this problem I think you should see what percentage of the LEDs will not work. If you are getting LEDs from ebay.com then you have problems, but LEDs from a good source will work.

 

[Kisen]

I almost never see a open LED. I do see bad solder joints that cause no current flow.
Do you have data on LED failure? (resent data) Before you spend time and money on this problem I think you should see what percentage of the LEDs will not work. If you are getting LEDs from ebay.com then you have problems, but LEDs from a good source will work.

I dont have any data at all on failure rates. The LEDs will be cree. Looking to potentially use these...

https://uk.farnell.com/cree/mlered-a1-0000-000u01/led-xlamp-red-mle-13-9lm/dp/2097587

I can imagine that LED failures may well be quite rare, but they can happen as they age. As the device is being built with the intention of being largely un-monitored by a human being, this is why i would like there to be some level of automated fault detection.
Basically, If the LED system fails and is left undetected, the whole system crashes.

 

[Externet]

Explore using light conduit to monitor the operation of each led, bringing the bundle ends outside the led housing to phototransistors.

----> **broken link removed**

Most important; never drive leds at its full rated capabilities shown on data sheets. Want more light ? ---> use more leds. Specially important in 'essential'/critical applications. Want even more reliability ? Use redundant circuitry.

 

[ronsimpson]

After the factory; LED failure is related to temperature. Spend some time thinking on how to pull the heat out of the LEDs.

What are you going to do when a power supply fails? (more likely)
How are you going to send failure data to a center point?

 

[Kisen]

After the factory; LED failure is related to temperature. Spend some time thinking on how to pull the heat out of the LEDs.

What are you going to do when a power supply fails? (more likely)
How are you going to send failure data to a center point?

You raise a good question on PSU failure. One i had thought of but only because i wanted to monitor each individual LED. In short if all 4 LEDs appeared as open circuit i.e no voltage, then i could assume that the PSU for that string had failed.
I expect the same could be said if i had 1 sense resistor for the 4 LEDs, No voltage means PSU is dead.
Each string of 4 LEDs has its own PSU, the Microcontroller board has its own PSU and a Supercap backup system. So if there is a powercut etc or an all out failure wireless comms can still be maintained for some time.

The heat of the LEDs is already taken care of. The LEDs are mounted onto a water cooled system that takes the heat from the LEDs and passes it into the bioreactor tank, heating the water in there to a nice 28c. If there is excess heat its sent to a circulation tank where heat goes to atmosphere.

 

[Nigel Goodwin]

You raise a good question on PSU failure.

As long as you use a decent quality supply you should be fine, many PSU's run for decades without problems - the vast majority of problems come from fitting low quality electrolytics, which have short live spans. A classic example was the original Sky HD satellite box - the PSU (made by Samsung) used the same useless caps as Samsung TV's, and the datasheet for the caps at the average running temperature gave a live of about 8-9 months only.

 

[Kisen]

So moving back on to the topic at hand, Is it looking like the best way here is to monitor the string of LEDs as a whole and report a failure in that module do you think.
I can see that being done with 1 sense resistor, but means that should 1 LED go open, then the other 3 will go dark too.

 

[Nigel Goodwin]

So moving back on to the topic at hand, Is it looking like the best way here is to monitor the string of LEDs as a whole and report a failure in that module do you think.
I can see that being done with 1 sense resistor, but means that should 1 LED go open, then the other 3 will go dark too.

If it's a string of LED's then one going O/C will kill them all - but the single sense resistor will detect that and report it.

LED's though are really pretty reliable, and what kills them is heat - if you run them at a slightly lower current then their reliable will increase accordingly.

 

[JonSea]

Big Clive in a recent teardown video looked at an LED light that had a zener diode across each LED of a series string. He surmized they were there so an open LED wouldn't kill the whole string.

If you're using a constant-current supply, you don't need a sense resistor. Voltage across the LED string will be equal to the sum of the forward voltages. If an LED shorts, the voltage will drop by the forward voltage. Voltage across the string would tell you how many LEDs are shorted. If any LED is open (assuming no zener shunt), voltage across the string will be equal to the maximum voltage output of the constant-current supply.

 

[Kisen]

Big Clive in a recent teardown video looked at an LED light that had a zener diode across each LED of a series string. He surmized they were there so an open LED wouldn't kill the whole string.

If you're using a constant-current supply, you don't need a sense resistor. Voltage across the LED string will be equal to the sum of the forward voltages. If an LED shorts, the voltage will drop by the forward voltage. Voltage across the string would tell you how many LEDs are shorted. If any LED is open (assuming no zener shunt), voltage across the string will be equal to the maximum voltage output of the constant-current supply.

That is most helpful, thanks.

Just a though, if i were to use these ...
https://uk.farnell.com/littelfuse/p...X0EBoCVxcQAvD_BwE&CAWELAID=120173390000082512

I assume that if an LED goes open circuit, i would never know it happened? The voltages would all appear correct?

 

[Nigel Goodwin]

Big Clive in a recent teardown video looked at an LED light that had a zener diode across each LED of a series string. He surmized they were there so an open LED wouldn't kill the whole string.

They are, and may be special diodes designed for this purpose, rather than normal zeners.

 

[JonSea]

I think a zener would be better. At any rate, a TVS or zener would have to have a breakdown voltage getter than the LED's forward voltage. Say the zener has a breakdown voltage 50% higher than the LED's forward voltage. If an LED goes open, voltage across the string would increase so you could tell.

 

[dknguyen]

I did think about monitoring the string of all 4 LEDs. This would work, however it doesnt give me the ability to know which LED has failed in order to trigger the shunt.
If all LEDs have an automatic shunt then i would never see a problem with the voltage, would i?

Assuming you use the automatic shunt, you already seem to have considered placing comparators across the shunt/LED to measure the voltage to tell if it's functioning, open-circuited or short-circuited but are worried about component acount.

You said you have a PIC though. Could you just straight-up run the voltage at each LED node to the PIC so you could just use the PIC ADC to read the voltages relative to ground and calculate the voltage across an individual LED by using the voltages of the LED's below it closer to ground.

If these are not just bare LEDs so they have some other stuff right? That would require you to step down the voltage before running it to the PIC ADC? That would require more components for the resistive divider to step down the voltage at each LED...and maybe op-amp buffers.

 

[KeepItSimpleStupid]

Couple of points.

LED will have a strong temperature function.
Not all LEDS have the same Vf
Select LEDS by Vf for each string (possibly normalize with a resistor).
LEDs get dimmer of time.
Heat is the killer.

That said, why not measure power of the entire set-up.

Initially I'd say TRMS meters and multiply. Might not be the exact value.
A shunt, hall effect or transimpeadance amp to measure current.
Could possibly use math functions on scope.
Do the math in the MCU and display.

 

[JonSea]

If LEDs are in series and powered by constant current, I don't believe there is any advantage in matching Vf.

If you simply measure the voltage across the string, you'll get pretty good diagnostic information. In normal operation, the voltage will vary somewhat with temperature, but a drop of 1 Vf will be distinctly noticeable.

 

[Kisen]

Thanks guys for your input, it has been most helpful.
From this thread i think i will be using a Zener for each LED as a shunt and i will measure the voltage of the string and monitor this for open/closed circuits.
I now understand how using a Zener to help keep the system running, can also be detectable (previously i didnt understand how this was possible.)
Time to start drawing up the schematics

 

[KeepItSimpleStupid]

If LEDs are in series and powered by constant current, I don't believe there is any advantage in matching Vf.

If he has 4 in series, matching Vf will generally match intensity for the string. You can "tweek" the strings to each other with an added resistor, so that Vf of 4 LEDS + 1 resistor is the same. Now you have uniform light from all strings. Matching Vf is more important if these are RGB LEDS.

Monitoring the LEDS for burnout, I'm suggesting measuring power. PWM and LED's likely complicate things a little. Average(V(t)) * Average(i(t)) will give you something to monitor, but it may not be actual power. Integrating v(t) and i(t) using an RC filter can give you simple numbers and little complication.

Spectral shifts happen with the change of operating current, but not looking at that.

I think there are different ways of dimming and I don't think you PWM I.