I suck and that's sad

I messed up my PCB design. The prototype board I haven’t even received yet will not work. It has two parallel groups of three 1 watt Luxeon I emitters in series and the heat sink slugs for each LED are electrically connected to two big copper planes on the PCB. If I had read the application brief describing assembly in addition to the datasheet and design guide I would have known that you aren’t supposed to do this. Why you ask? The reason is explained in the reliability datasheet which isn’t even listed on the product page where the datasheet, thermal design guide and other application briefs are on their website. I had to use google to discover it.

Inside the LED package are two parallel circuits. The first is the LED connected between the anode and cathode: anode -> LED -> cathode. The second goes like this: anode -> zener diode preventing current from anode -> heat sink slug -> zener diode preventing current from cathode -> cathode. The reverse breakdown voltage of the diodes is 7V so if you have three LEDs in series with +10V on top after the current regulator and GND on bottom the voltage at that first zener diode at the top of the chain is actually 10V and it will break down and nothing will work.

At this point I have two options. Option the first: there exists thermally conductive, electrically non-conductive epoxy. You can even buy small batches of it for purposes like attaching heat sinks to ram chips and so on. Arctic Silver makes some that is apparently good and which can be had for $13 or so from Fry’s. Applying this to the bond pad for the LED would theoretically electrically isolate the slug from the board but still conduct heat to it. This is a bit ugly and may not work depending on your interpretation of the Arctic Silver website and documentation. Option the second is to order the updated board in which I have fixed the problem. This is appealing because I have also fixed a few other minor, not-show-stopping things and have included some extra labeling on the silk screen. It would mean another two weeks or so wait and double the cost of the thermal adhesive.

If nothing else this has been very educational. Thanks to Brian Neltner whose website (ahem) illuminated my design error during his discussion of LED choice. His art collaboration project is very cool. I will probably go with the new board.