Calculating LED Requirements for Reuse: A Guide for Scientists

[Flyboy]

I recently got hold of an LED aircraft landing light after it started taking on water (bad seal on the junction between the front plastic window and the heatsink core structure was letting water in during takeoff and landing), but it is otherwise totally functional. Or it was, until I broke off a surface mounted component on the driver board during the dissection of the light. The LEDs themselves are totally functional still, and I would like to find a way to reuse them on another project. However, with the driver board damaged, I don't know what the specs on the LEDs are, and I can't gather the data by hooking the driver up to a suitable power supply.

Here's the data I do know...

• Total power draw of the light is 40W, per the installation manual included in the replacement bulb (which is a like-for-like replacement)
• Driver board accepts either 12V or 24V DC (the two most common bus voltages in aviation), with the current limits being adjusted accordingly to maintain 40W total power.
• There are four LED chips in the circuit, wired in series.

Is there a way to (carefully) gather data for voltage and current limits for the LED chips via a power supply and/or multimeter? Or is there a formula I can use to determine this without any other data?

EDIT: Forgot to ask this in the initial posting, but... I noticed thermal paste dots on the backs of the chips and on the interface between the heatsink core and the aluminum back shell of the light, so I know I need to provide cooling for the LED chips. I just don't know how much. Any insights on how to tackle that issue?

 

[berkeman]

Or it was, until I broke off a surface mounted component on the driver board during the dissection of the light.

Can you repair the board? Do you have a hot air rework station? (You probably need that anyway if you want to desolder the LEDs and reuse them.)

Is there a way to (carefully) gather data for voltage and current limits for the LED chips via a power supply and/or multimeter?

Sure, just use a variable power supply with a reasonable current limit adjustment, and turn up the voltage until the LED starts to light. Then adjust the voltage and current limit of the power supply until it looks like you are getting reasonable light out of the LED.

I noticed thermal paste dots on the backs of the chips and on the interface between the heatsink core and the aluminum back shell of the light, so I know I need to provide cooling for the LED chips.

40W for 4 LEDs is around 10W per (minus some power consumption for the LED drive circuit). So you need to have a heat sink that limits the junction temperature of the LEDs to around 100C or so (conservative), so (100C-25C)/10W is around 7.5C/W junction-to-air thermal resistance for the heat sink setup...

 

[Flyboy]

Can you repair the board? Do you have a hot air rework station? (You probably need that anyway if you want to desolder the LEDs and reuse them.)

The LEDs are actually on independent chips, as they were arrayed around the heatsink core, facing out into the reflector buckets built into the backshell. I'm eyeing using them as-is, as they're on metal-backed chips, which is better for heat dissipation and transfer, if I'm understanding things correctly.
As for repairing the board, one of the tiny, rice grain sized capacitors got knocked loose during the dissection and promptly made itself scarce, as small parts do in a shop environment. I have no clue what it was rated at, so replacing it isn't really an option unless another one breaks down sometime soon and I can split the driver board housing open with more care. (Unlikely, as the parts guy informed me, after I had already split this one open, that there's actually a warranty on the lights.)

Sure, just use a variable power supply with a reasonable current limit adjustment, and turn up the voltage until the LED starts to light. Then adjust the voltage and current limit of the power supply until it looks like you are getting reasonable light out of the LED.

Okay, sounds simple enough. Limit it to the rated current at the target voltage range? I.E., if I were to be driving it at the 14v spec for the assembled unit, which is 2.9A, I would limit the current to, say, 2A?

40W for 4 LEDs is around 10W per (minus some power consumption for the LED drive circuit). So you need to have a heat sink that limits the junction temperature of the LEDs to around 100C or so (conservative), so (100C-25C)/10W is around 7.5C/W junction-to-air thermal resistance for the heat sink setup...

... Okay, I don't fully understand what you're getting at here. My understanding of air cooling stuff is more towards piston engines in light aircraft and heat exchangers on jets, so dealing with electronics cooling is definitely a little out of my experience zone. I understand the limiting the temperature bit (I want to keep the magical smoke inside the LEDs, thank you very much), but I'm not tracking on the "junction-to-air thermal resistance" bit. I'm assuming (and I use that term with the caution it deserves) that it means the effectiveness of the heat transfer from the chip to the heat sink? Basically, I'm not sure how to translate that into surface area to radiate the heat away. And using the surface area of the backshell isn't exactly the best reference, as it was inside of a housing to protect wiring and such, and only a ~1in^2 plate was exposed to airflow at ~100-180 knots.

 

[berkeman]

Okay, sounds simple enough. Limit it to the rated current at the target voltage range? I.E., if I were to be driving it at the 14v spec for the assembled unit, which is 2.9A, I would limit the current to, say, 2A?

No. A simple LED has a forward voltage (Vf) of around 2V. I don't know if each of the 4 LED modules that you have consists of a single LED and junction, or perhaps several matched LED junctions in series. So I would start with your power supply set at V=1.5V and I_limit+100mA and increase the voltage to see when the LED starts to turn on. Then adjust the V and I accordingly to figure out where the operating point is approximately.

The LEDs are actually on independent chips, as they were arrayed around the heatsink core, facing out into the reflector buckets built into the backshell. I'm eyeing using them as-is, as they're on metal-backed chips, which is better for heat dissipation and transfer, if I'm understanding things correctly.

So you can use them mounted to their original heat sink(s)? Sorry, I'm not understanding. Can you upload a couple pictures? Use the "Attach files" link to the lower left of the Edit window to attach JPEG or PDF copies of the images. Thanks.

 

[Flyboy]

Okay, first off is an image of the light in question from the manufacturer:View attachment parmetheus-pro-par-46-featured_2048x2048.webp
You can kinda see one of the diodes there, pointed out into the reflector bucket. The black core structure is both support structure for the forward lens and heat sink. It's screwed to the back shell by a pair of screws that also help hold the driver board in place.

And these are the LED chips:

(Please forgive the messy toolbox top in the background.)
Chips are wired in series, joined by splices. Chips are ~ 1/2" on a side. Backing plate appears to be aluminum, and has small spots of residue from thermal paste.

For good measure, I'm throwing in a shot of the driver board, too:

C1 is what got snapped off. For scale, holes are for #8 screws on the positive and negative pads, and #6 for the mounting points.

Hopefully this helps you understand what's going on.

 

[berkeman]

Does your DVM have a "diode" setting? If so, you could verify if each module has a Vf of about 1.5-2V and hence is a single LED junction and not several in series...

 

[Flyboy]

I know the one in my toolbox does, will have to see if my home one does or not.

 

[Tom.G]

Mostly speculation, but here goes.

It looks like one end of C1 is connected to Minus Supply (Gnd), with the other end connected to the junction of two 150Ω resistors in series. It is difficult to tell for certain from the photo, the labelling for C1 is in the way. Anyhow, if that is the case, it indicates that C1 is likely part of a Pi filter for whatever the resistors are supplying power to.

Try connecting a 1μf to 10μf capacitor to the C1 pads. Capacitors in that value range are usually electrolytics, so get the polarity correct... otherwise the go 'Bang.' (sometimes with shrapnel)

Please let us know what happens!

Cheers,
Tom

 

[Flyboy]

Mostly speculation, but here goes.

It looks like one end of C1 is connected to Minus Supply (Gnd), with the other end connected to the junction of two 150Ω resistors in series. It is difficult to tell for certain from the photo, the labelling for C1 is in the way. Anyhow, if that is the case, it indicates that C1 is likely part of a Pi filter for whatever the resistors are supplying power to.

Try connecting a 1μf to 10μf capacitor to the C1 pads. Capacitors in that value range are usually electrolytics, so get the polarity correct... otherwise the go 'Bang.' (sometimes with shrapnel)

I can do that this weekend pretty easily. Is there a way to hook it up to a power source and probe the pads and see which orientation the capacitor is supposed to go?

And safety glasses are absolutely a must. "Safety First, Not Safety Third" and all that.

Please let us know what happens!

Cheers,
Tom

Absolutely!

 

[Tom.G]

Is there a way to hook it up to a power source and probe the pads and see which orientation the capacitor is supposed to go?

Didn't you already answer your question?

C1 is what got snapped off. For scale, holes are for #8 screws on the positive and negative pads

 

[Flyboy]

Didn't you already answer your question?

I’ll take a look tomorrow at the board under magnification. I think I understand what you’re getting at... looking at the traces on the board.

 

[Rive]

I broke off a surface mounted component on the driver board during the dissection of the light.

That looks like a clean break. Any intent on a repair instead?
I would start with piling up 10u 50V ceramics there till it starts working.

 

[hutchphd]

Absolutely. For instance

https://www.digikey.com/en/products/detail/murata-electronics/GCM21BR71H105KA03K/4903983
This is a 0806 surface mount part (meaning ~2mmx1.5mm) which I think will fit those pads. Caps are hard to kill by soldering......I bet it will work fine

 

[Dullard]

It's been mentioned, but to be clear:
Heat dissipation is the dominant factor in the design of LED lamps. Operating them with an insufficient heat sink will likely kill them.

 

[Flyboy]

It's been mentioned, but to be clear:
Heat dissipation is the dominant factor in the design of LED lamps. Operating them with an insufficient heat sink will likely kill them.

Yeah, that's another thing I'm completely over my head on. No clue how much heatsink and radiator fin surface to use.

 

[berkeman]

Yeah, that's another thing I'm completely over my head on. No clue how much heatsink and radiator fin surface to use.

I don't know if this will help you at this point, but the way you calculate this is to look at the datasheet for the maximum working junction temperature of the LED (not the absolute max), and look at the $\theta_{JA}$ for the heatsinks (and airflows) that may fit the application.