LEDBrick Project - DIY pendant w/ pucks

theatrus

100-mile-commuter
Buying an LED lighting fixture is easy, but its not fun. As such, I've decided to do a start to finish project to design a replacement for my current T5 lighting setup using no "kits" or other pre-fabricated bits (excepting AC/DC power supplies, because those are hard to design).

My intent is to produce a monolithic "puck" LED emitter board containing a series of emitters arranged into channels on one metal-core single layer PCB, in addition to a driver and controller arrangement.

The board is designed to fit in the footprint of an 80mm fan, which will be connected via threaded rods through the heatsink to the emitter board. The drivers and controller will then be stacked above the fan on the same footprint, forming a wide pendant light (way less sexy than Kessil).

fan-on-heatsink-sm.jpg


The above image shows the heatsink stock (Heatsinkusa.com 4" serated) cut to size, along with a very dirty old fan as reference of the desired shape. Any heatsink could be used here - this is a bit oversized for good measure.

The design packs a large number of LED positions on the single layer metal core emitter board (not all have to be stuffed of course).

The pads and suggested LEDs are:

  • 4 Rebel UV emitters (these are very expensive, but more efficient than the popular Chinese SemiLEDs emitters. They however do not have a primary optic - more on this later)
  • 6 Cree XP-G(2) White emitters (or any Cree XP package)
  • 6 Cree XP-E Blue emitters (or any Cree XP package)
  • 6 Osram Oslon Squre Deep Blue emitters
  • 4 Osram SSL Hyper-Red
  • 4 Phillips Rebel PC-Amber
  • 4 Phillips Rebel Cyan
  • 4 Cree XP-E Green (or any XP)

for a total of 38 emitters on the PCB. Emitters can be skipped by simply jumpering over the pads.

sch.png


ledbrick-board-top.jpg


(~ 3x3in board)

Why so many LEDs? Lower currents, higher efficiency, and future flexibility. Remember, this is not a build-cost efficient fixture :)

A downside to the large emitter count is the inability to provide secondary-optics, focusing the light tighter than 120 degrees from the primary optic. The emitters are simply too close together to use most commercially available lenses, which feature large footprints. I intend to run these pendents very close to the water, which would not need the use of an optic.

Also integral to the board is a Microchip MCP9808 I2C temperature sensor. The sensors ground pad is brought out to one of the mounting screws, in an effort to get a good thermal path.

All connectors to the board are a series of Molex PicoLock blade-style high current flat wire-to-board connectors. I've successfully crimped the contacts, however annoying it may be, with the Engineer PA-09 which qualifies it for hobby use (as the Molex official crimping tool is $500). I've used the 10-pin and 6-pin connectors, as those are the only ones available in stock at any distributor (a downside to more unique connectors).

I have ordered prototypes (metal core prototype prices are nowhere as bad as they used to be) and stencils of the emitter board, and am continuing the design work on the driver and controller boards to sit in the stack.

Stay tuned to this thread!
 
Sneak peak at an early revision of the driver assembly (4 corners hole spacing is identical):

driver-r0.jpg
 
Which driver are you planning on using? I'm looking to design something with the LM3414, so I'm just curious.
 
The above board was designed with the ONSemi NCP3066.

Upsides:
  • 1.5A switch current
  • Cheap

Downsides:
  • Its inefficient, due to a bi-polar output element (not a FET)
  • Thermal sinking is bad (compounded problem from the above)

I had actually considered the LM3414 as the second runner - I'll end up with a test board of each most likely.

Currently, my main MCU is slated to be the K24 (_120 family). At the price points, there is little reason to stick to an 8-bitter if you also want an RTC, USB, and just lots of resources to be totally wasteful with. I'm planning on adding a home for a stack-on-board Bluetooth LE/Smart module (such as the RN4020), even if I don't end up with firmware for it at first.
 
Look into the A6211. O2Surplus happens to have a large cash of them at the moment. about the only way you'll get them. He ordered a few hundred a while back for assorted projects. Shoot him a PM, I'm sure he can spare you a few.
 
Look into the A6211. O2Surplus happens to have a large cash of them at the moment. about the only way you'll get them. He ordered a few hundred a while back for assorted projects. Shoot him a PM, I'm sure he can spare you a few.

I am actually on the train for a whole bunch of them when they show up at a US distributor again.

Looking again, the LM3414 has an interesting feature as the current control is set internally and not a comparator from a sense resistor, which means you can use other means to set peak current and not depend on resistor swapping. A6211 uses the classic sense resistor.

I don't want to drive LEDs any harder than ~900mA anyway, so I may be making an early Rev1 to explore this with the National/TI driver.
 
Looks very interesting :)

As you say tho, can't see it being the cheapest solution, but where's the fun in that? :lmao:

Tim
 
if you want to tighten the beam a bit you could use a reflector like i did for a freshwater led build. using light tube aluminum sheet is an easy way to tighten the beam and in my opinion helps with mixing and avoiding hard shadows. just a thought, looking forward to the build.
 
I am actually on the train for a whole bunch of them when they show up at a US distributor again.

Looking again, the LM3414 has an interesting feature as the current control is set internally and not a comparator from a sense resistor, which means you can use other means to set peak current and not depend on resistor swapping. A6211 uses the classic sense resistor.

I don't want to drive LEDs any harder than ~900mA anyway, so I may be making an early Rev1 to explore this with the National/TI driver.

The current setting feature on the LM3414 is what drew me to them. Originally I was planning on using the LM3404 but then I saw the LM3414 and thought "why not use this since it's much easier to implement and because I'm probably not going to go over 1000mA". I'm kind of hesitant of getting a PCB made because I don't have a o-scope to fully check the circuit once I build it and I don't want to blow up LED's, lol. Also, I keep thinking maybe someday I will need to go above the 1000mA limit. But I do have all the parts to build 5 of these, except the PCB.
 
Very interesting. This definitely solves the multichip led problem, where they often offer only one or two colors, where here you can adjust everything.
 
The current setting feature on the LM3414 is what drew me to them. Originally I was planning on using the LM3404 but then I saw the LM3414 and thought "why not use this since it's much easier to implement and because I'm probably not going to go over 1000mA". I'm kind of hesitant of getting a PCB made because I don't have a o-scope to fully check the circuit once I build it and I don't want to blow up LED's, lol. Also, I keep thinking maybe someday I will need to go above the 1000mA limit. But I do have all the parts to build 5 of these, except the PCB.

I have a full lab setup - I'll happily be a guinea pig on a design :)
 
LED emitter driver boards arrived:

ledbrick-board-1.jpg

ledbrick-board-back.jpg


I also got my captain stencil - will hot-plate on an expensive set of LEDs later this week. Also missing my set of M4 screws, nor have I drilled and tapped a heatsink yet.

(Also, naming dilemna, I guess this is the LEDBrick Emitter board, even though its not named that :) )
 
I'm really interested into how hot this thing gets. (did you mention a current? I scanned through the post and didn't see how many mA you plan on running everything at). But that's not much heatsink, even with a fan (which FYI looks like 90% of the old fans I have too :D)
 
At ~700mA for all channels (with dimming it would be less) and assuming 30% efficient emitters we are at about a conservative 65W that this needs to sink away. Most of the emitters are better than that (low 40%). The thermal resistance from package to heat sink is an unknown at this point, so I don't have a good estimate on die temperatures.

Forced air is required, but probably not much.
 
Don't suppose you will end up having spare LED PC boards to share with the rest of us at some point?
 
Just a curious observation as I've not researched them but don't the luxeon violets need larger thermal pads (in their case also the wire traces) to sink away the heat? They are just as powerful or more than the larger Semi size chips, where does all the heat go?
 
Just a curious observation as I've not researched them but don't the luxeon violets need larger thermal pads (in their case also the wire traces) to sink away the heat? They are just as powerful or more than the larger Semi size chips, where does all the heat go?

Yeah, this is a concern - they don't have a dedicated thermal pad, and I laid them out with just traces, which is a concern. I'll make a note to check the pad temperature when I power it up at first.

Edit: According to http://www.philipslumileds.com/uploads/446/AB114-pdf I am doing it wrong. I'll correct that for a new revision :)
 
Back
Top