An LED lighting project

[jtr1962]

Dumb question...why copper pipe and not plastic?

To dissipate the heat from the LEDs. The copper tube provides enough surface area to dissipate ~2.5 watts of heat without becoming overly warm. I'm even thinking of putting vertical fins on the copper tubes for the chandelier in my dining room because of the required light output. The plain copper tubes are adequate if you only need about 300 lumens per candle (this is approximately equal to a 30-35 watt small base bulb). If I need to replace a 60 watt small base bulb ( ~700 lumens), then I need about twice the surface area for heat dissipation.

As for why go through all this instead of just buying LED bulbs, for starters there don't exist any viable candelabra-base LED bulbs. This isn't surprising given the very small area available to dissipate heat. Second, it's much more reliable having 5 or 6 LEDs connected to one well-engineered ballast, as opposed to each LED having its own, individual ballast crammed into a tiny space. Note that most LED screw-base lamps only have lifetimes in the 25000 to 35000 hour range. This is because most run the LEDs/electronics very hot. By running the LEDs cooler, 100,000+ hour life should easily be possible. I also obtain greater efficiency than commercial products. 3000 lumens/34 watts = 88 lm/W. I could obtain over 100 lm/W if I ran the LEDs at 700 mA.

 

[time]

The spectra of LEDs are continuous, not spiky like fluorescent spectra.

Fluorescent lights use a UV light source to stimulate 3 or 4 different types of phosphor. The phosphors emit 3 or 4 different color spectrums that combine to produce white light with a reasonable spectrum width.

Power LEDs use a blue light source to stimulate different types of phosphor. The phosphors emit different color spectrums that combine to produce white light with a reasonable spectrum width.

Spot the difference?

White LEDs can also be made by coating near ultraviolet (NUV) emitting LEDs with a mixture of high efficiency europium-based red and blue emitting phosphors plus green emitting copper and aluminium doped zinc sulfide (ZnS:Cu, Al). This is a method analogous to the way fluorescent lamps work. This method is less efficient than the blue LED with YAG:Ce phosphor, as the Stokes shift is larger, so more energy is converted to heat, but yields light with better spectral characteristics, which render color better.

 

[e_dawg]

The problem i have with both LEDs and CFLs is that they while they are highly efficient, the quality of light i find is lacking in 2 primary areas:

1. Both suffer from significant gaps in the emission spectra compared to a blackbody radiator incandescent. LEDs are generally more continuous than CFLs, but they still usually lack some output in the reds as well as the cyans.

2. Additionally, LEDs usually are very point-source and highly intense (and therefore appear very bright), but in practice, they don't put out much total light yet require lots of diffusion to prevent them from being harsh on the eyes. Perhaps the new "switch" LED bulbs will address this issue.

To be clear, i'm not a huge fan of incandescents due to their inefficiency and amount of heat production, but i do prefer the quality of their light output in general (at least the daylight / neodymium coated bulbs) and feel their continuous emission spectrum is valuable, especially when combined with CFLs or LEDs to offset their primary output deficiency: the red wavelengths.

My current solution is to pair a 100w incandescent with a 23w GE Reveal CFL and a 23w Daylight CFL in a 3 socket fixture suspended from the ceiling with a frosted diffuser dome. I find the 3 emission spectra work together relatively well to form a relatively continuous spectrum for most wavelengths while reducing the impact on efficiency by having only 1 incandescent.

 

[ddrueding]

I agree e_dawg. At work I even combined 5000K and 6500K 4' fluorescent tubes. They both claim 90-95CRI, but it seems that each fills in the gaps of the other.

I know JTR combines assorted color LEDs with white ones to level out the spectrum when he can. This fixture is backed by a pair of 75w halogen cans, so it shouldn't be an issue.

The LED candles arrived from JTR on Thursday, and I did a test-run yesterday. Chandelier assembly begins today, hopefully finished early next week.

 

[jtr1962]

I completed the chandelier in my mom's bedroom. This chandelier formerly used 25 watt candelabra bulbs and was a consistent source of complaint about both light quality and quantity. Also, despite not being used often, light bulbs frequently blew out.

There was room to put the ballast under the medallion as shown in the following 2 pictures:

As a result, there is only low-voltage wiring going to the chandelier. Here are three more pictures of the end result. I left the candles natural copper color, but polished them out, and coated them with clear to prevent tarnishing.

Here's how it looks with the yellow C7 plastic covers. Obviously we're going to use clear, but they're still not available. I just wanted to get some idea of how everything would look.

Light levels as measured with a light meter are generally 50% to 100% more than before, while power consumption is 16 watts instead of 150. Light quality is much better, close to daylight actually (I used 3A Cree XPGs, which are around 5000K).

 

[ddrueding]

That looks awesome, JTR. I've assembled the chandelier, but I'm still waiting for the metal works to return the shortened chain and redesigned ceiling mount. Originally it was for a 4m ceiling and a hook coming out of the plaster.

 

[ddrueding]

After 8 hours of 100% load in 72F of mostly still air, the base of the copper "candle" was 100F. That is really amazing because the LED itself was only 107F; fantastic heat transfer. The transformer itself was 111F on the outside of the plastic housing. This is now the bit that I am most worried about.

I plan to mount the transformers on a 2x4 post rising out of the insulation in the attic, possibly with heat sinks strapped to the sides. The ambient up there can be much higher, particularly were the attic fan to fail. What is the functional temp. range on those transformers before I start significantly shortening their lifespan? I'd rather avoid active cooling if possible.

 

[jtr1962]

It's marked 90°C maximum case temperature on the ballast label. You're getting about 21-22°C rise above ambient, so you should probably be OK up to attic temperatures of around 65°C. The hottest I've ever measured our attic was around 60°C. Our attic is uninsulated, we have a dark colored roof, and this was on a day when the outside temperature was pushing 110°F. The summers by you seem pretty mild by comparison. I honestly think you'll be OK mounting the ballasts the way you plan to. I think (but I'm not sure) that they're thermally protected. If so, then they'll likely just shut down if they overheat, and work normally once they cool down. The ballast in the chandelier I did is essentially in an enclosed, insulated space. We'll see how it does, but I'll bet even mounting like this there won't be any problems.

By the way, the temps you measured are pretty much in line with my readings. I was getting an LED temperature rise of about 20°C above ambient at 1 amp. This should be conducive to very long life according to this document.

 

[jtr1962]

According to the ballast data sheet, there is protection with auto recovery against over voltage, over current, over temperature, and short-circuit. It looks like failure due to overheating shouldn't be an issue at all.

 

[ddrueding]

Thanks JTR, that is good to hear.

 

[Will Rickards]

Yeah the yellow covers don't look very good. Maybe the picture is just making it seem worse.

You have that on a dimmer, correct? Because that seemed like too much light for a bedroom. Looks like matching the mirror color might have been a better choice for color than the copper. But it seems to go with the rest of the room.

 

[time]

Two great projects Jtr.

I'm curious as to why you chose copper over aluminium though. I can understand copper for the pill, but isn't copper pipe like that super-expensive? And I seem to recall that aluminium is actually slightly better at dissipating heat, as opposed to conducting it?

 

[ddrueding]

I like the (exposed) copper, something about being honest with the materials and the craftsmanship.

 

[Howell]

And I seem to recall that aluminium is actually slightly better at dissipating heat, as opposed to conducting it?

Like a processor you want very good conduction to get the heat to travel as far away from the heat source as possible. An ideal conductor would show the temperature at the base of the copper at the same temperature as the heat source.

Here is a list of thermal conductivity values. The lower values show good insulation and the higher values show good heat sinks. You will see that copper is a doubly better conductor of thermal energy than aluminum alloy. Interestingly gold is not as good as copper but silver is better. Were the rod aluminum jtr would have measured a higher temperature at the LED and lower at the base.

Once the heat is in the rod, it is convection (primarily) that can remove that. The convection and conduction equations are both dependent on temperature differential and surface area. And the convection counterpart to thermal conductivity is the heat transfer coefficient.

 

[Howell]

The ballast in the chandelier I did is essentially in an enclosed, insulated space. We'll see how it does, but I'll bet even mounting like this there won't be any problems.

Is that a metal or plastic housing for the ballast?

 

[jtr1962]

Yeah the yellow covers don't look very good. Maybe the picture is just making it seem worse.

You have that on a dimmer, correct? Because that seemed like too much light for a bedroom. Looks like matching the mirror color might have been a better choice for color than the copper. But it seems to go with the rest of the room.

Yes, it's on a dimmer. In fact, the reason I chose that particular ballast is so it could work with a standard triac dimmer. And yes, the yellow covers really do look that bad in person. The original incandescent lights were nearly that bad-dim and very yellow.

I could have matched the mirror color with gold paint but honestly I hated the thought of painting over my nicely polished finish.

Two great projects Jtr.

I'm curious as to why you chose copper over aluminium though. I can understand copper for the pill, but isn't copper pipe like that super-expensive? And I seem to recall that aluminium is actually slightly better at dissipating heat, as opposed to conducting it?

Howell explained it pretty well. In the end, an aluminum tube with a thicker wall thickness for similar overall thermal resistance might have worked as well as the copper. The primary reason I chose copper was the need to be able to solder various pieces together since it was unlikely I would find a piece exactly the size and shape needed.

I am considering these as a starting point for the dining room chandelier. I need at least 700 lumens per candle, which in turn means greater surface area to dissipate the heat (either that, or 200 lm/W LEDs which as of yet don't exist outside the lab).

I like the (exposed) copper, something about being honest with the materials and the craftsmanship.

My thoughts as well. Another possibility was to gold electroplate the candles to match the gold leaf on the chandelier but I don't have an electroplating kit. It probably would have been too pricey anyway.

Is that a metal or plastic housing for the ballast?

Plastic housing. The same ballast is also available with a metal housing for a few dollars more but I didn't realize this until after I bought it.

 

[ddrueding]

I had this weird thought last night about having a single fan in the center and forcing the air through the "legs" and then through candles that are finned on the inside and solid on the outside.

That or running heatpipes from inside the copper tubes throughout the main mass of the chandelier to increase the surface area.

 

[jtr1962]

Yeah, that would work. In fact, you probably wouldn't even need fins on the inside. Just forcing air at a few m/sec would increase heat dissipation by a factor of 4 to 5, meaning the candle would be good for ~10 watts of LED power, ~7 watts of waste heat, and 900-1000 lumens (those are roughly the figures for an Cree XML driven at 3 amps). In a few years the problem will likely solve itself anyway once we have LEDs which put out 75% light, 25% heat.

The heat pipe idea is even better in that it's passive. A chandelier has plenty of surface area if the heat is channeled properly.

 

[ddrueding]

The heat pipe idea is even better in that it's passive. A chandelier has plenty of surface area if the heat is channeled properly.

The only problem with that is finding heat pipes that are both flexible and allow for any orientation. Most of the ones that can be arranged arbitrarily have complex internal wicking structures that make them fragile.

 

[ddrueding]

Chandelier is up and running beautifully. We've chickened out with the colored diffusers, and I've programmed it to run at 15% brightness with bare LEDs. I'll get some pictures tomorrow when the construction stuff/scaffolding around it is cleaned up.

 

[mubs]

Looks very good, jtr. You should go into business making custom lights like this. An ex-boss of mine got into pipe-making (the smoking kind) and racks for storing the pipes. It was a specialized niche and he was well-known in the relevant circles.

 

[ddrueding]

Another LED lighting project. This one isn't nearly as complex, as I plan on building them from scratch myself. The only trick is that I still want the source to be dimmable via a regular 120v dimmer.

The fixture is going to be very compact and only have a single bulb (something like this maybe). Due to the compact nature, I need to find a compact ballast (ideally, a cubic inch or so). Anyone have experience with this sort of thing? I don't even know where to look.

 

[jtr1962]

I don't think anybody makes a 120 VAC LED ballast that small which is dimmable with a regular triac dimmer. I'm not even sure if it's possible.

If I find out differently, I'll let you know. The best way to have a small, dimmable light source off 120 VAC would be to use a wall wart (12V, 1 amp is fairly small these days) as the primary power source, then have a switching converter powering the LEDs. Dimming could be done via a potentiometer. I already have boards which can do this sort of thing.

 

[ddrueding]

The space inside is basically a cylinder 1.25" tall and 4" in diameter. I can lose the dimming, and the enclosure is machined aluminum, so no heat sink necessary if it is laid out right. Does this get me closer to possible?

 

[jtr1962]

Maybe. I think Deal Extreme has 120VAC drivers small enough to fit in that. Like everything else from them though the quality is a crapshoot.

 

[ddrueding]

Chandelier is up and running beautifully. We've chickened out with the colored diffusers, and I've programmed it to run at 15% brightness with bare LEDs. I'll get some pictures tomorrow when the construction stuff/scaffolding around it is cleaned up.

Picture as promised (only a year late!).

We haven't found clear diffusers yet, so we are currently running without any. This means that the majority of the light hits the ceiling and reflects downward. This actually works remarkably well. The light falls off nicely as you move away from the mounting point.

Disclaimer: Taking pictures of intense light sources sucks. And I'm not a great photographer. I was trying to simulate what it looks like in person, and failed pretty hard. Cameras (and monitors) just don't have the dynamic range to display these things correctly. I ended up taking 14 different exposures and trying to do a composite.

 

[jtr1962]

That lights the room up pretty nicely. It also looks like it renders colors reasonably well.

 

[CityK]

Picture as promised

I see that DrunkenBastard is visiting

 

[ddrueding]

I see that DrunkenBastard is visiting

Nah, I've just been learning my scotch, and whiskey. At the moment my favorites are the Macallan 15 and Jameson 18.

 

[CougTek]

Is that jeans clothing on the chairs? First time in my life I see that.

 

[ddrueding]

Is that jeans clothing on the chairs? First time in my life I see that.

Indeed. We have white ones for formal stuff (never actually used them), but the denim is durable and works well.

 

[mubs]

This Philips LED system seems awesome; check it out.

 

[ddrueding]

At the end of that review they mention a different product that sounds even more exciting. LIFX

 

[mubs]

Things are getting interesting! In about 5 years, we can have it anyway we want.

 

[jtr1962]

Yep, variable color lamps are one of the big reasons to switch to LED. I think in the future we'll probably see low-cost general lighting LED lamps which produce varying shades of white. This is actually fairly easy to do compared to products which produce every shade of the rainbow. You just use maybe 2700K and 6500K LEDs. By varying the proportions, you can get any shade of white from 2700K up to 6500K. I'll guess that these lamps will have a simple rotary switch to change the CCT instead of a wireless interface.

 

[ddrueding]

What I still want is higher CRI. The LEDs that jtr wired up for my latest project were supposed to be some of the best, but next to 98CRI florescents of the same color temp there is clearly some missing/exaggerated spectrum going on. It just isn't as nice to look at. I suspect that these color changing bulbs don't fully address that issue.

 

[jtr1962]

What I still want is higher CRI. The LEDs that jtr wired up for my latest project were supposed to be some of the best, but next to 98CRI florescents of the same color temp there is clearly some missing/exaggerated spectrum going on. It just isn't as nice to look at. I suspect that these color changing bulbs don't fully address that issue.

The color changing bulbs use red, green, and blue LEDs as far as I know in order to get all the colors of the rainbow. RGB isn't particularly good at CRI. Even the best carefully chosen wavelengths will only give a CRI in the mid 80s. You need 4 colors to get into the low 90s, and 5 colors to get into the high 90s.

Here are some integrating sphere tests of the Nichia 219. CRI goes from 89 at lower currents up to nearly 95 at 2000 mA. There are some LEDs which do better, such as the Xicato Artist Series with a minimum CRI of 95 and a typical CRI of 97. Unfortunately they're only available in CCTs of 2700K, 3000K, 3500K, and 4000K. My holy grail is a ~5000K LED with a CRI in the high 90s. That would essentially be artificial sunlight. By the nature of LED's continuous spectrum, as opposed to the discontinuous spectrum of flurescents (even the high-CRI ones), a 95+ CRI LED should give better light than a 98 CRI fluorescent.

 

[ddrueding]

My holy grail is a ~5000K LED with a CRI in the high 90s. That would essentially be artificial sunlight. By the nature of LED's continuous spectrum, as opposed to the discontinuous spectrum of flurescents (even the high-CRI ones), a 95+ CRI LED should give better light than a 98 CRI fluorescent.

Exactly. When you see those, we'll have some work to do