Friday, October 09, 2015

LED Torchiere Conversion

Twenty-three years ago, my wife and I moved into an apartment with very little built-in lighting, and purchased a couple of torchiere floor lamps to put in the living room and office. Their 300 watt quartz-halogen lamps provided a powerful, white light when turned up, and could be dimmed to a warm glow for quieter evening settings.

Twenty years ago, we moved into the house we now own, and placed the two torchieres in the living room, making for nice, soft indirect lighting on either side of our two chairs.

Perhaps 15 years ago, we picked up a couple of wall-mounted Quartz-halogen projector lamps in the "as-is" section of the Burbank IKEA. I mounted these as reading lights on the shaft of the torchieres (one for each of us), removing the torchieres' failing internal dimmers and routing the reading lights' wires through the now-available dimmer control opening and down through the core of the torchieres' poles to their DC power "bricks." I added external, remote-controlled AC dimmers for the torchieres.

Some years after that, I mounted rear satellite speakers for a surround-sound audio system high on the poles of each of the torchieres flanking our seats, firing rearward and slightly inward to reflect off the wall behind. The speaker wires joined the AC wires feeding the torchiere lamps, and DC wires powering the reading lights.

I've been slowly converting our home lighting from incandescent to LED (and very thankful never to have been faced with the ugly colors of fluorescent lighting). I'd noodled solutions for replacing the two 300 watt torchiere bulbs with lower-power LEDs for more than a year, researching available LEDs and power supplies (I still wanted dimmable lighting, which is a bit more exotic for LED driving circuits). Of late, there have been an increasing number of BIG, postage stamp-sized LEDs which produce light output adequate to replace any household application. I'd almost committed to buying separate components and fabricating a heat-sink/cooling system, when I stopped to look more closely at a LED ceiling down-light conversion kit at Costco. After some thought, I bought the $27 kit, containing two fixtures promising "same output as 120 watt" incandescent reflector flood lights, and 1,250 lumens of light output from only 21.5 watts. Manufacturer quotes of output of the 300 watt Quartz lamps in our torchieres quote nearly 5,000 lumen output, however, that's omnidirectional - light radiates in all directions from the glowing filament. LEDs output most of their light perpendicular to a plane, which is exactly where I want the light going - up into the ceiling. So I thought I might get away with a lower rating to achieve similar light levels.

Once home, I removed the diffusing lens (just breaking it off, then discovering that I missed two screws to remove it reversibly, which turned out to be moot) screwed the medium-base bulb adapter (the familiar threaded light bulb mount for the U. S.) into a work light socket and fired one of the fixtures into the ceiling at the same height as the torchiere lamps. It wasn't bad - about the same output as one of our torchieres which has an old, darkened lamp which was probably putting out half as much light as it could. But it was disappointing next to a freshly-relamped instrument.

I considered just going with the reduced light level, but then decided that I might be able to Siamese the two LED fixtures by sawing about 1/4 of each of their housings off. They would both then fit (sort of) within the upturned shade of the torchiere after I surgically removed all the original socket and reflector parts.

Further exploration and experimentation revealed that I could free the array of 18 LEDs and the steel disk to which they are bonded - I realized it wasn't glued as I first suspected, but merely stuck by silicone heat sink grease. The hockey puck-sized dimmable power supply was easily separated, though I had only about 1/3" of wire with which to solder on four connections between the power supplies and LEDs.

A couple of hours wandering around Home Depot yielded some galvanized steel plates (made for joining construction lumber), pop rivets, some threaded rod and a Plan.

I riveted the LED arrays from both of the fixtures on my fabricated mounting plates/heat sinks (with yet more heat sink grease), and tucked the power supplies underneath. The whole assembly floats above the torchieres' shallow, upturned bowl-shaped shades on two threaded rods.

Before (dimmed very low for photo)

After (also dimmed low)

When I fired it up, I was thrilled to find that the combined output of the 43 watts of LEDs equals or exceeds that of the brand new 300 watt quartz-halogen bulb, and the pattern on the ceiling and walls is perfect. The lights dim as hoped, though there is little to no color change (having warm colored lighting at night may be a better idea to prevent unwanted wakefulness), and the light levels abruptly change at some points - which matters not. Power consumption will drop to 1/7th of the original, and in the summer, the air conditioner will have to contend with over 500 fewer watts of heat in the living room at night.

The original halogen light on the left, and the LED on the right - the dimmers are set at only about 30 per cent, but the LED exhibits none of the reddish color change of a filament bulb.

Success! Our 20th-Century lamps are now 21st, and will continue to serve as reading lights and Surround Sound satellites for years to come.


PZ said...

Great post! It's been a while since you've posted so I'm not sure whether you are still seeing comments...
Just 1 question about the four 1/3" long wires between the power supply and the LED array. In the photo, it looks like you added more wire to make the resoldering easier. Is this correct? Also, it looks like these wires can be desoldered from the LED array, rather than cutting the wires. Is this correct?
Thanks for all your help.
This info saved me either a lot of trial and error or a lot of money!

Ellsworth said...


In the original product, they bring the wires from the power supplies up through a hole in the disk bearing the LEDs, then trim it to about 3/4" and solder it to the contact points on the LED disk. When I cut that in half to free up the components, it left very short (which I described as 1/3") stubs. I didn't want to have to crack open the plastic cases for the power supplies. I also wasn't confident that I could solder to whatever was exposed on the LED disk, so I stripped those 1/3" long pieces of wire remaining on the disks, and soldered my 22 gauge zip cord extensions. I clamped a hemostat on the wire stubs close to the existing solder joint to keep it from melting when I soldered on the extension.

You might be able get away with desoldering the wires from the disk - I just wasn't sure whether that might be something strange like a multi-layer circuit board, with little to no exposed conductor, or some solder-resistant surface.

Good luck with the project. Let me know how it goes.

PZ said...

Thank you Ellsworth. That sounds like a plan.
Now have 2 final questions.
You mention tucking the power supply under the tie plate. If it is secured to the plate, I can't see how or where. Or does it just sit on the base of the torchiere shade?
Also, I can't really see the 2 threaded support rods in the photos. I would have used 3 rods so the tie plates wouldn't wobble. Also I'm thinking of putting dishwasher rack tine repair caps on the end of the rods where they contact the shade base. My thinking is that if the lamp is moved the caps would keep the rods from sliding around as much. Does this make sense to you?

Thanks again for all you help. This is a great project!

Ellsworth said...


I think I mounted the power supply pucks to the underside of the LED heat sink plates with some 3M VHB foam tape. I remember thinking I would just let them rattle around, but I think this just made it easier to finish the mounting, not having to juggle all the parts. In retrospect, there's some potential for hum from either the LEDs or the power supplies, so this might have helped minimize sound conduction between the pucks and plates.

You can see the ends of the rods in the photos. In the very first image of the working LED arrays and the next-to-last "after" photo, you can see two galvanized nuts atop the lower rectangular galvanized plates, which you can see in the gap between the two LED support plates. I threaded the sections of rod into some existing steel bracketry from the original lamp, and then drilled holes in those lower galvanized plates to match the positions. There are nuts under the plates serving as rests, and then nuts on the top to clamp the the plates down. I just adjusted the support nuts until the plate was in a suitable location, then chopped off the excess threaded rod, reassembled and ran the jam nuts down on the top, clamping the original lamp and the LED assembly together rigidly. It's all bendable, so any off-level aspect after assembly could be fixed by bending the rods or plates.

It's almost a year later, and we burn these things every day, and no problems so far. I will mention that the X10 remote dimmers to which the lamps are connected (and controlled the 300 watt quartz-halogen lamps for decades) dim the lights in a very non-linear fashion, with a very steep curve from dim to full bright at a very low setting for the dimmer. So it's tricky to pick some intermediate dimmer levels, especially with some of our remotes, which appear to send command pulses at different rates. Different dimmers may produce better results - I've noticed descriptions of some household dimmers being "LED/CFL compatible" - perhaps this reflects some non-linear curves in their operation.


P.S. Thanks for the thanks. Glad I could help.

PZ said...

So I finally got around to this project and after a little trial and error, I think it is an unqualified success. The major deviation from your plan was that my wife didn't think it was bright enough. It looked like there would be room for as many as four LEDs in the torchiere, so my plan was to try three and if that still wasn't bright enough, then try four (which would have comparable lumens to the 300W halogen). Well, three was substantially brighter that the 300W halogen, so we settled on that.

I did want to ask your advice on heat dissipation. You could have fit the two LEDs on a single 3x7 tie plate, rather than joining two 3x7s together with 1x5s. Did you do this for added surface area for heat dissipation? I riveted a single 3x7 to a single 3x5 in a "T" configuration with a LED at each of the three ends of the "T", using heat-sink compound (I would post a photo but I don't see how to do this in my comment). Do you think I should worry about overheating?

And finally, I did a couple things just a little differently:

I cut the wires from the power supply to the LED right at the LED and soldered the extension wire to the LED. This gave me a little more wire to work with at the power-supply end.

My local hardware store doesn't stock rivets smaller than 1/8" so instead of using rivets I used the original screws to reconnect the LEDs to the power supplies on the under side of the tie plate.

Ellsworth said...

My thinking about thermal load/dissipation was that although the original lamp fixture provides considerably more connected conductive surface area (the entire fixture) than the tie plates provide, the original instrument was obviously intended to be mounted in a down-firing orientation, and thus had the expectation of retaining quite a bit of heat in the downturned concavity of the reflector. The original design also traps heat between the diffuser lens and the fixture. Further, there is no provision for allowing air to flow across either the LED array plate itself or any of the fixture - if anything, some air might be exchanged above the fixture, but even then, that assumes that there is available air in the ceiling above the fixture.

In my design (and probably yours), the LED mounting plate and the supporting heat sinks are free-floating, and air can pretty readily be drafted in from the sides of the fixture to chimney up from the hot LED array.

These are my intuitions, and I'm prepared to suffer the consequences of premature LED failure if I've miss-guessed. So far, so good. My strategy in the event of thermal failure was to escalate to active cooling, using a personal computer CPU cooling block - which incorporates a fan and a heat sink. This unfortunately may introduce fan noise into the environment, but there are fairly quiet CPU fans, and also manually- and thermostatically-adjusted fan speed controllers. A kooky possibility I also considered was a liquid-cooling CPU cooler, which allows one to move the thermal load away from the heat source via tubing. This would allow the possibility of building a remote sound-baffled box which contains the liquid-cooling radiator and noisy fan at some position from the light fixture.