best LED for DIY fixed lighting?

[idleprocess]

Going on ten years ago I homebrew-ed an outdoor lighting installation using 3-up stars populated with Lumileds Rebel LEDs. Heatsinking was ... mininal with ⅛" x 2" barstock sections cut all of about 3" long and whatever additional heatsinking that the minuscule shell the Al sample containers offered that enclosed the LEDs:

The installation is still going strong at a >80% duty cycle but it's time to upgrade. Suspect I'll use a similar style of fabrication for the replacement. I'd like to run 3 of the "pucks" at ~500mA and another ~9 at ~150mA.

What's the best LED for the job that's available mounted to the semistandard triple MCPCB stars that's compatible with the usual suspect 105XX Carclo optics? Ideally this will be 3000K and will tolerate running 24/7 for a decade without complaint. High CRI would be nice but isn't critical since this is safety lighting more than anything and it's bracketed between a MHI streetlamp a few houses over and a HPS streelamp a few houses over the other direction. If I need to beef up the higher-power pucks with thicker - or more - barstock that's fine. If 500mA isn't advisable I can change up and run up to 5 of them at something closer to 350mA and should slightly exceed the illuminance that the original 4 achieve.

 

[brickbat]

I don't know - depends on what you mean by 'best LED', and the reason you're replacing the old Rebels...

What beam angle do you need?

 

[Microa]

The Cree XPG3 would be the best LED for the Job ( low price, high light output at 1 watt power ).

 

[idleprocess]

I don't know - depends on what you mean by 'best LED', [...] What beam angle do you need?

My requirements:

What's the best LED for the job that's available mounted to the semistandard triple MCPCB stars that's compatible with the usual suspect 105XX Carclo optics? Ideally this will be 3000K and will tolerate running 24/7 for a decade without complaint. High CRI would be nice but isn't critical

TL;DR I need a LED that can handle continuous operation for a decade with minimal heatsinking in 3000K with enough output that three triples driven to ~500mA can match the output of four triples ~2010 Rebels driven to 350mA - or alternatively five triples driven to ~350mA will neatly exceed my original four.

I can improve the heatsinking nominally on the new fixtures but not by much - I'd like them to be as concealed as possible under the eaves of the garage. Longer 2x⅛ barstock, using some 2x1x⅛ U-section, spacing the heatsink off the soffit with a washer stack by ~⅛" to make the heatsinking more effective are my options.

and the reason you're replacing the old Rebels...

Expansion (present arrangement covers a driveway; new arrangement should cover 3 sides of the garage), conversion to 3000K (present LEDs are 4000K), and age (365.25 * 10 * 24 * 80% = ~70 thousand hours runtime).

The Cree XPG3 would be the best LED for the Job ( low price, high light output at 1 watt power ).

Will they tolerate up to ~4.5 watts per triple with perhaps 10-12 square inches of heatsink - and not all of it in close proximity to the MCPCB - for a decade? The rebels were tolerant of ~3.15 watts per triple with ~6 inches of effective heatsinking for a decade.

 

[brickbat]

I'm not inclined to do all the hard work, but curious to see where you land. The only way that works for me when solving the 'best component' problem is to throw together a spreadsheet -l label a handful of columns with parameters I care about, and start sifting through data sheets and filling rows for each candidate. Since your efficacy spec seems to be something relative to a 2010 Rebel, you'd obviously need to look that up first.

I think your toughest constraint (i.e. the one that will rule out most of the candidates) is compatibility with a triple optic, which is why I asked about your desired beamwidth.

 

[idleprocess]

I'm not inclined to do all the hard work, but curious to see where you land. The only way that works for me when solving the 'best component' problem is to throw together a spreadsheet -l label a handful of columns with parameters I care about, and start sifting through data sheets and filling rows for each candidate. Since your efficacy spec seems to be something relative to a 2010 Rebel, you'd obviously need to look that up first.

I mostly need information that's not in spec sheets - namely an idea of long-term thermal ruggedness.

My performance needs in terms of peak output or efficiency are not demanding given that LEDs from decade ago driven at 350mA more than sufficed - and circumstances forced me to only install four fixtures rather than the orginally-envisioned six. Nichia and Luminous LEDs run hot. Cree and Samsung seem to be more efficient. I'm not as familiar with Lumileds' offerings but I could certainly grab some triple Rebels again easily enough. Osrams are probably tolerant of the sort of prolonged abuse I intend to throw at them but they do not seem to be as readily available as other offerings.

I think your toughest constraint (i.e. the one that will rule out most of the candidates) is compatibility with a triple optic, which is why I asked about your desired beamwidth.

I was going to use the Carclo epliptical optics (10510) for all but two corner fixtures, which will likely feature wide spot optics (10508). The present fixtures have no optics whatsoever beyond scuffing up the glass lids for the sample containers that serve as the housing.

 

[brickbat]

I mostly need information that's not in spec sheets - namely an idea of long-term thermal ruggedness....

Reputable suppliers might have easily-located LM-80 results. See here, for an example from Cree...

https://www.cree.com/led-components/media/documents/LM80_Results.pdf

If you can't find similar reports from a candidate supplier, I can't imagine what other method you'd use to assess 'thermal ruggedness'.

 

[idleprocess]

Reputable suppliers might have easily-located LM-80 results. See here, for an example from Cree...

https://www.cree.com/led-components/media/documents/LM80_Results.pdf

If you can't find similar reports from a candidate supplier, I can't imagine what other method you'd use to assess 'thermal ruggedness'.

Perhaps I should have been more explicit: if someone that's done something similar to what I'm attempting could relay their experiences, it would be helpful. I have little in the way of instrumentation save for a rather unreliable non-contact thermometer and a multimeter.

 

[brickbat]

I see. In the event you don't get that, we could look at the LM-80 results. I think the only missing puzzle piece is your heatsink performance. (i.e. measure the HS and ambient temps, and tell us the total input power to the puck.)

If I was going to the trouble to design and build my own lights, as you're doing, I'd avoid the components for which there's no published LM-80 data. If an manufacturer did the testing, and it showed favorable results, there's no good reason they'd hide it. It's to their advantage to publish it, so potential customers can see how great their LEDs are.

Gross generalization - If you can't find an LM-80 report for a candidate LED, don't pick it because the chances of it being crap are higher.

 

[Ken_McE]

I don't have a good crisp answer to your question. :sigh:
Heat and longevity seem to be your big concerns. Under driving and over cooling are of course your friends here.

I have an LED driveway light, but I went with mounting a cool white rope light on the wall. It is about 4' off the ground and aimed down at about a 45 degree angle. It gives a mild, very evenly distributed light similar to full moonlight. It will never be a work light, but you can see and identify all objects, see some color, and move quickly and easily.

If you'd care to post a build log of the next generation lights I'd be interested to see it.

 

[idleprocess]

Heat and longevity seem to be your big concerns. Under driving and over cooling are of course your friends here.

Yeah, this is the opposite of the approach to flashlights where overdriving in a small package is entirely possible for a device that sees single-digit overall duty cycles and sub-single digit duty cycles at peak output. Since I've realized a decade out of a setup with such modest thermal design and the heatsinks have never been anything beyond warm to the touch I'd like to imagine that ~40% more drive current won't cause problems.

I have an LED driveway light, but I went with mounting a cool white rope light on the wall. It is about 4' off the ground and aimed down at about a 45 degree angle. It gives a mild, very evenly distributed light similar to full moonlight. It will never be a work light, but you can see and identify all objects, see some color, and move quickly and easily.

I's difficult to photograph the current setup in such a fashion as to demonstrate its effect, but the four wholly unfocused fixtures (save for really medicore diffusion on the lenses) manage to illuminate the driveway in an acceptably even fashion. Reflection off the white garage door certainly helps. The intensity is lower than the original dual floodlamp fixture but I recall that the spec sheet lumens were greater and the distribution is far better.

Actual optics will alter this equation a tad over the driveway but I doubt the impact will be negative.

If you'd care to post a build log of the next generation lights I'd be interested to see it.

I expect to. It's occured to me that the local makerspace likely has some thermal instrumentation thus I could fab up a prototype or two and see how they perform at various drive currents.

 

[Bimmerboy]

I've been reading along for a few days now, and am also interested in how this project turns out.

To get a fuller picture of the setup, are the individual 3-up stars running in series or parallel? I assume just one star per puck? Are the pucks in series or parallel to each other? What driver, and how are you able to change the current?

Random thoughts and questions:

Those old Rebels are tough. They were a leap forward in reducing lumen sag while hot. But, they came out over 10 years ago. Are most of todays LED's not just as tough, or tougher? Plus, modern die packaging seems to have a lot better thermal transfer, and many LEDs can be soldered direct to one layer copper stars.

I would have to think the latest emitters are putting out more light and less heat per watt than the Rebels, so even at the same drive current there should be a slight bump in output, and run slightly cooler. I think you can get away with a small bit of drive increase while just re-using the old heatsinks. Maybe 400 - 450 mA just to be on the safer/more conservative side? Without doing any math whatsoever, I'm guessing you'll get a noticeable improvement in output at roughly the same amount of heat.

Speaking of thermal stuff, how are the old pucks bonded to the bar stock? Could some improvements be made there? For example, if the old ones were, say, just stuck on using thermal adhesive or something, could you screw them down this time for better contact pressure, and use thermal grease?

Using wide angle optics this time should extract a lot more lumens out of the pucks over scuffed glass or plastic. Big gains right there, and I think it's a great decision.

I was gonna' say something else, but forgot while typing the other stuff... lol. Perhaps it'll come back to me. Anyway, looking forward to further updates!

Edit: Just thought of this. Why use canisters with bottoms? They're just going to impede heat transfer. Bond the stars direct to the bar stock, and surround them with a small cylinder. Or, if you like the size and function of the ones you're already using, perhaps just cut the bottoms out?

 

[idleprocess]

To get a fuller picture of the setup, are the individual 3-up stars running in series or parallel? I assume just one star per puck? Are the pucks in series or parallel to each other? What driver, and how are you able to change the current?

Present arrangement is two strings of six since I was using current-limiting AC-DC drivers.

Future arrangement is likely to be parallel with a 12V DC Bus to each puck. I am seemingly misremembering Meanwell making DIP style CC drivers for ~$1 each (LDD-L series closest to a DIP package and more like $3 each) so I may well opt for the simple power resistor instead for current limiting.

Those old Rebels are tough. They were a leap forward in reducing lumen sag while hot. But, they came out over 10 years ago. Are most of todays LED's not just as tough, or tougher? Plus, modern die packaging seems to have a lot better thermal transfer, and many LEDs can be soldered direct to one layer copper stars.

I would have to think the latest emitters are putting out more light and less heat per watt than the Rebels, so even at the same drive current there should be a slight bump in output, and run slightly cooler. I think you can get away with a small bit of drive increase while just re-using the old heatsinks. Maybe 400 - 450 mA just to be on the safer/more conservative side? Without doing any math whatsoever, I'm guessing you'll get a noticeable improvement in output at roughly the same amount of heat.

I would assume newer is similarly rugged to the Rebel after some quick glances as LM-80 data as well as more efficient.

Speaking of thermal stuff, how are the old pucks bonded to the bar stock? Could some improvements be made there? For example, if the old ones were, say, just stuck on using thermal adhesive or something, could you screw them down this time for better contact pressure, and use thermal grease?

LED >> Arctic Alumina adhesive >> sample container >> Arctic Alumina adhesive >> barstock. Could reuse the heatsinks for some of the fixtures, but I'm also ~tripling the size of the project as well.

Using wide angle optics this time should extract a lot more lumens out of the pucks over scuffed glass or plastic. Big gains right there, and I think it's a great decision.

The housings were really shallow (going to guess that I saw >90° of exposure) so the lumen loss would primarily be the hilariously bad diffusion. But they're also shallow enough for triple 10mm optics, said optics are cheap, and the main function of the lower-power pucks is going to be wall-washing with the elliptical optics so they're an obvious choice.

Edit: Just thought of this. Why use canisters with bottoms? They're just going to impede heat transfer. Bond the stars direct to the bar stock, and surround them with a small cylinder. Or, if you like the size and function of the ones you're already using, perhaps just cut the bottoms out?

Simplest method I could think of for enclosing the LED so as to keep detritus and bugs off the emitters. They're made from aluminum mercifully rather than steel like most so the main loss is the double-layer of AA adhesive.

EDIT
This is the best photo I've been able to take lately of the existing arrangement:

Exposure is kind of bad since the pavement is wet but this approximates how it works today.