Question about mounting high-power bead LEDs

[Chris C]

I want to mount a bunch bead LEDs in sets of two, with the two LEDs in each set as close together as possible.  Example LED:



Common single-LED stars are too big, and result in the LEDs being spaced too far apart.  If another PCB exists more suited for my goal, I'd appreciate a tip.  But I bet I'm going to have to mount the LEDs directly to my aluminum U-channel.  The question is, how?

I can easily insulate the anode/cathode leads from the aluminum with Kapton tape.  The thermal pad is trickier, as I've read that some LEDs have an electrically insulated pad, and some don't.  Unfortunately that detail is not listed for the LEDs I'll be ordering.  If they happen to be insulated, then I'm set, I can just use thermal glue.

However if they don't, I'm not sure what to do.  I have some cheap thermal tape from Ebay which will ensure electrical insulation, but given the included warning "don't use this for a CPU or GPU, you will fry it", I'm not sure this would be sufficient for a bead LED either.  On the plus side I'll be running these LEDs at no more than 0.5W each, on the minus side that thermal pad sure is a small contact area.

Suggestions?

---

Here's some more information on the project in case you're interested.  I'm trying to build some LED lighting with both relatively high CRI (color rendering index) and PAR (photosynthetically active radiation).  It will be placed where it can be viewed directly, so I don't want any blinding points of light, or for it to appear excessively multicolored.

As structure and heatsinking, I'll be using 1" square aluminum U-channel.

Into the open face of the channel, I'll be mounting ChromaLit remote phosphor.  This converts royal blue light to white, with a 4,000°K color temperature, and CRI of 80.

Inside the channel, I'll be mounting royal blue LEDs, at 2" center-to-center spacing.  Since these will indirectly produce most of the light, I'm choosing 5W Cree XT-E for high efficiency, which will be pre-mounted on stars for good heat transfer and long life.  They'll be run at no more than 3W.

Between each Cree will be two other LEDs, one cyan and one deep red.  These will fill in the two biggest gaps in the spectrum produced by the ChromaLit, increasing both CRI and PAR.  Not much power is needed in relation to royal blue, so 1W LEDs will more than suffice.  Efficiency isn't so important either, so cheap no-name LEDs will be used.  I've already verified the ChromaLit will pass these two colors without conversion.  But it still acts as a diffuser, blending together the colors, with uniformity dependent on how close I can get the LEDs together.  That's why I want to avoid using stars on these LEDs.

The final detail is an internal reflector.  It may not be optimal, but I'm going to try some decent quality aluminized mylar.  I'll cut strips wider than the channel, and punch out holes where the LED lenses are.  With its edges glued to line up with the channel's, it should naturally hold a roughly parabolic shape.

I'm still debating the drivers.  I might go with Meanwells for convenience.  Or I might use cheap Ebay boost converter modules, adding my own current limiting as I've seen others do.

[tszaboo]

You should solder them on Aluminium (the correct way to write it) PCBs. The stars are basically those.
Some reading material:
http://www.cree.com/~/media/Files/Cree/LED%20Components%20and%20Modules/XLamp/XLamp%20Application%20Notes/XLamp_PCB_Thermal.pdf
The package might be different, the principles are the same.

[Chris C]

You should solder them on Aluminium (the correct way to write it) PCBs. The stars are basically those.

That would require an aluminum PCB to exist which meets my stated spacing requirements, so unless you know of one, that's no help.

Single-LED stars won't work, as they are more than twice the width of each bead LED.  I've seen some three-LED stars, but only for Luxeon LEDs which much more expensive and powerful than what I require, and use of those would nearly double the total cost of my project.  And there's stars for RGB LEDs, but I need very specific colors, and none of the colors in RGBs are appropriate.

[tszaboo]

You an design the PCB, where you place the LEDs exactly to your requirements. Or you can outsource it, someone else can design it. Metal core PCBs are very common today, some manufacturers offer even pooling options, meaning you can order them without setup costs.

[Chris C]

Sounds expensive, but maybe I'll be pleasantly surprised.  Is there one you can recommend?  Keep in mind this is a one-off project, with a budget of $200.

Another possibility, I just saw a lighting project where someone glued 1W bead LEDs directly to aluminum stock using JB Weld!  Now approaching four years of use, with no issues.  Mine will be running at only about 0.5W, so perhaps at this power level a hackish approach is quite adequate.  If the LED's thermal pads turn out not to be electrically insulated, I could use proper thermal glue to attach the LED first to a bit of aluminum stock, several times the area of the original thermal pad, as a heat pre-spreader much like a star PCB.  That would then be attached to the main aluminum channel using thermal tape, which provides electrical insulation.  Even though the tape isn't such a good thermal conductor, the larger transfer area should help.

[tszaboo]

Oh, it is a one-off. Still I;ve checked some prices for you, 50mmx300mm (2 inch x 12 inch) panel which can be broken down to small strips, cost 96 EUR+VAT ~ 105 USD, in single quantity, from EU manufacturer. So I guess it is doable. Soldering might be a bit trick though.

[Ian.M]

At 2W/linear inch (3W + 0.5W + 0.5W every 2"), I suspect the U section extrusion will run rather hot as the inside will be effectively insulated by the Mylar film and the remote phosphor screen.  If I were you, I'd make up a 4" test section, with expanded polystyrene foam end plates and a plastic cover the same thickness as the phosphor screen, with a U of Mylar film under the cover and test it with 8W applied to its base inside it using 2 power resistors on 2" centres.   It needs to be mounted in the same orientation and to the same type of surface as the actual light.  The base interior surface temperature in the center should be closely comparable to the heatsink temperature the LEDs will experience. 

I strongly suspect that you will end up needing heat spreaders for the bead LEDs directly soldered to their thermal pads to keep their junction temperatures low enough for reliability.

[Chris C]

[Ian.M], that's a good idea.  I will mock up a test as you describe.

If it runs only a bit hotter than generally considered acceptable, that's fine.  My understanding is that with regular white LEDs, because the phosphor is applied right over the hottest element (the LED die), the phosphor tends to be the point of failure if heat is not drawn away from the die quickly enough.  Plus heat caused by phosphor conversion losses also adds to die temperature, due to their proximity.  Remote phosphors eliminate both issues, and it's claimed a light using remote phosphors can be built with less heatsinking.  Exactly how much less I don't know.  But I do notice commercial lights using remote phosphors are quite small for their wattage.  The maximum power my fixtures will run is 23W, with 36 square inches of exposed outside aluminum surface area.  I have a commercial remote phosphor-based 23W screw-in LED, with a fraction of that surface area, and mostly plastic!

Guess I'll find out.  It will be a nice experiment.

[Seekonk]

I have a 30W Philips DLM lamp.  That has a massive 6 pound finned heat sink and it doesn't even have driver electronics!  A 9W E27 LED fixture has only plastic and the package touts 9,000 hours life.  Remember when they said 30 years.    Have a number of LED now that have taken to twinkeling after not much use.  I have a feeling that many LED are more efficient the less you drive them.  They are so cheap that you can use twicw as many of them and use mounting methods that are less critical.  More just give better light distribution anyway.

[MikeW]

I have a feeling that many LED are more efficient the less you drive them. 

This is very true.

[Chris C]

I have a 30W Philips DLM lamp.  That has a massive 6 pound finned heat sink and it doesn't even have driver electronics!

I have a 13W Cree DLM I picked up at a thrift store for $5.  It too has a shockingly large heat sink.  I guess because it's intended to be installed somewhere with little ventilation?

I agree it's best to treat LEDs gently.  On the last LED fixture I built, I did so, but probably to unnecessary extremes.  At some point increases in efficiency and longevity become so small that they'll never pay back the extra money spent on the build.  So on this one, I want to test the limits so I can find out where the happy balance is.  I have a LUX meter and can test for power vs. efficiency, and lumen loss over time.  These will be built in 12" modules, and if I see signs of degradation, I can add more modules and turn down the power.

[MikeW]

I have a 30W Philips DLM lamp.  That has a massive 6 pound finned heat sink and it doesn't even have driver electronics!

Some of those Philips DLM series contain the driver electronics in the led module itself.

 

[Chris C]

Did some rough estimation today.  Each 12" long, 1" square profile U-channel will run at 23W.  I'm not sure why, but folks generally neglect that some of that power is converted to light, not heat.  Assuming 25% of the power goes to light production, heat production is actually only 17.25W.

Using some reference graphs, I believe the heat sink temperature will average somewhere around 125°F.  That's the maximum I'd be comfortable with, so I'll still do a mock-up test.

[Chris C]

Ok, that makes sense.  I guess I just prefer using all knowns, then derating to allow for unknowns separately and at the end.

Placed orders last night, for just enough parts to build and test one 12" fixture with the real royal blue LEDs and remote phosphor.

Plus some Cree XP-E2 Blues to play with.  There's a spectral gap between the royal blue and supplemental cyan in my previously described design.  Blue is partially converted by the remote phosphor even though it's not the optimal wavelength, so I'd need more power relative to the other supplemental colors for enough to bleed through to cover the gap.  And blue LEDs are apparently less efficient than royal blues, so this would result in a fixture with less overall efficiency.  I'm just curious whether it will make any noticeable difference in the light quality.

[Zero999]

And blue LEDs are apparently less efficient than royal blues, so this would result in a fixture with less overall efficiency.

It depends on what you mean by efficiency.

Blue LEDs have a higher luminous efficacy due to the response of the human eye but royal blue LEDs gives more power out vs power in.

For some reason LEDs are more efficient at the extreme ends of the visible spectrum and less efficient in the middle which is why phosphors are used.

[Chris C]

It depends on what you mean by efficiency.

Blue LEDs have a higher luminous efficacy due to the response of the human eye but royal blue LEDs gives more power out vs power in.

True.  As the blue light will be mostly converted by the remote phosphor, I figure "efficiency" is best represented in this case by the luminous efficacy of the LED.  I expect that by using a mix of blue and royal blue instead of just royal blue, perceived brightness of the fixture will drop overall, due to loss of green which the eye is most sensitive to.  But if it noticeably improves color rendering, it might be worth a reasonable drop in other stats.

[BurningTantalum]

The CREE datasheet was an interesting read- I was unaware of the use of thermal vias on glass fibre PCBs.
Recently I repaired 3 LED stage lights with 36x 8mm bead LEDs in each- red green & blue.
All 3 had failed due to moisture ingress killing the SMPSs but one actually had water swilling about in the case so had caused the failure of 4 of the LEDs on the aluminium-backed PCB due to the leads corroding. As the lights are for personal use now, I cleaned up the board and replaced the LEDs. The originals had pads on the PCB under the LEDs but as far as I can see the LEDs were not glued down, or soldered. I dobbed some thermal paste under the ones that I replaced.
Am I to understand that the correct procedure for mounting would be to solder, with paste, or glue the LEDs to the PCB ?
As the units seem to be running OK I don't think that I will bother to remove 108 LEDs to re-seat them, but the life expectancy may be reduced, as I understand it.
BT

[Zero999]

It depends on what you mean by efficiency.

Blue LEDs have a higher luminous efficacy due to the response of the human eye but royal blue LEDs gives more power out vs power in.

True.  As the blue light will be mostly converted by the remote phosphor, I figure "efficiency" is best represented in this case by the luminous efficacy of the LED.  I expect that by using a mix of blue and royal blue instead of just royal blue, perceived brightness of the fixture will drop overall, due to loss of green which the eye is most sensitive to.  But if it noticeably improves color rendering, it might be worth a reasonable drop in other stats.

LEDs with a high CRI and efficiency use violet and red LEDs with a range of phosphors to produce the full visible spectrum.