High Power SMD LED Array - Heatsinking

[Chisatholm]

Hi, I'm new here so please forgive me if I don't know the dos and don'ts just yet.

I'd like to put 12 high power (1000mA) SMD LEDs on a board in a 3x4 array, to occupy as little space as possible. My first experiments with SMD soldering have gone well, so I think the SMD aspect per se will be OK. What I'm concerned about is that these devices kick out a lot of heat in a package roughly 4x4mm. Although my duty cycle will be low (1 second on, >10 seconds off), I still believe I'm going to have to heatsink the LEDs. Manufacturers typically offer a version with a sink, but they are bulky - if I put 12 together, the overall device will be too large.

The way I plan to try is to i) make a PCB on 35 micron prototyping board to accommodate the LEDs, ii) drill through the board where the centres of the LEDs will be, iii) solder the LEDs in place, iv) back fill the holes with thermally conductive / heatsink compound, so I have a 35micron layer behind/under each LED and then finally v) glue that assembly onto a sheet of aluminium to act as the heat sink. It feels like it's going to be hard backfilling the holes in such a way that I don't get greasy heatsink compound everywhere and ruin my chances of a good mechanical bond to the aluminium as well as a good thermal connection.

Here's the data sheet of a typical LED I plan to use, and Farnell's listing of the 35micron board:

http://www.farnell.com/datasheets/1796056.pdf
http://uk.farnell.com/cif/abc16/board-fr4-2-sides-4-10-35µ-100x160mm/dp/1643090

If anyone has any experience of such a project or thoughts on where my plan might fall down or be improved, I'd be very glad to hear!

Thanks,

Andrew.

 

[Harald Kapp]

I don't think the hole-plus-thermal compound is very effective. Thermally conductive compound is usually applied in very thin layers to just fill the gaps between a component and the heat sink.

I recommend you create an etched PCB using the recommendes solder pad layout from the LED's datasheet. Route the cathode connnections (1,4) on top (where the LEDs are placed). Route the anode and the thermal sluf on bottom (opposite the LEDs). Use lots of through holes under the thermal slug to connect to bottom. This will be rather effective in diverting the thermal energy from top to bottom. You can then glue the PCB onto an aluminum heatsink using thermally conductive glue.
Note that in order to efectively remove heat from the LED you definitely have to solder the thermal slug. As this is on the underside of the LED, only some kind of reflow soldering will be suitable.

 

[Chisatholm]

Hi Harald

Thanks for your reply. I need to be able to control the current through all 12 LEDs individually. (I plan to use this to expose light sensitive colour photographic paper and control the colour that the paper 'sees'. The LEDs colours will be 10 red - the paper is not very sensitive to red light - 1 green and 1 blue, and I need to be able to control the brightnesses of the three colours separately) The control circuit I designed has all the anodes and cathodes at floating voltages ie the anodes are not common to all LEDs. So following your suggestion would, I think, mean I'd have to isolate each anode from the others on the heatsink somehow or redesign the control circuit with common anodes. Is that right? Isolating each anode seems very fiddly. Redesigning wouldn't be too hard.

However, my reflow soldering is beginner level - I only started experimenting with it in order to do this project. Could you describe how I would actually fill the hole under the slug? Would I first reflow solder the LEDs in place on the PCB using connections 1 & 4 but without any heatsink, then when the board has cooled, put fresh solder paste into the hole (you say drill plenty of holes, but I only have a hand drill and good eyesight!) under the LED, and then reflow again with the heatsink in place, soldering the LEDs to the heatsink via a slug of solder?

I had not heard of heat conducting glue, so I looked it up. Here's an epoxy-based glue that seems potentially useful: http://www.farnell.com/datasheets/651746.pdf It says it has a thermal conductivity of 1.1 W/mK. So I could reflow solder the LEDs in place and then glue the PCB to the heatsink using this epoxy. If I used the 35micron board (but no heatsink compound) and had a 2.5mm dia hole behind the LED which I filled with epoxy, then I would have a solid cylinder of epoxy 2.5mm dia x 35 micron thick. If there were a 1K temperature difference across the ends of this cylinder, then the thermal power going through the slug would be P = 1K x pi x 0.00125^2 x 1.1 W/mK / 35E-6 = 0.15 W. So a temperature increase of 1K at the device shifts 0.15 W of thermal power, ie I have a thermal resistance of 6.5 K/W. Let's say it's actually 15 K/W. I can tolerate a temperature rise of 100K (the operating temp is 125C) and the maximum thermal power output is 2.4W. So provided my final heatsink has a resistance of no more than 25K/W, my total temperature rise is 2.4 x 15 + 2.4 x 25 = 96K. Given that I have a low duty cycle and that I built some margin for error into the calculation of the epoxy resistance, perhaps this is workable?

Any comments/ideas gratefully received!

 

[Colin Mitchell]

3.6 x 1 x 12 = 43.2 x0.1 = 4.3 watts Some of the wattage will be released as light but 4 watts on a small board will require nuts and bolts from the LED heatsinks to fins on the back of the board.

 

[Chisatholm]

Hi Colin

Thanks for your reply. I couldn't see what numbers you were combining there to get 4.3 watts, in particular where 3.6 was from. Is that the forward voltage x current x no. of LEDs? The LED datasheet says the thermal power output is 2.4W (per LED). So I think that's 2.4 x 12 = 29W total, with a duty cycle of < 0.1.

It would be great to use nuts and bolts - well inside my comfort zone of construction techniques!! But the LEDs I want to use are tiny (4.4mm square), I want to pack them into the smallest space possible and the only thermal connection offered is the metal back.

I realised that the thermal resistance I calculated in my last post was per LED, by the way, so that's really 25 / 12 = 2K/W for the overall heatsink, which is much more ambitious.

Thanks!

 

[Harald Kapp]

How about not using a PCB at all?

You could use small screws with a flat head (hexagonal, may need some grinding to become really flat), srew them into the heatsink in a spacing that's appropriate for your purpose. This way you build 12 islands elevated above the heatsink. Use thermal glue to fill the gap between screw and hole and to ensure the screws dont get loos.
Alternatively you could route grooves into the heatsink to form the islands if you have access to a good router. This is quite more ambitious than using screws.

To each island you glue one LED with its thermal slug using thermal glue. This will drain the heat from the LEDs.

Use thin wire to connect the LEDs to your control circuit. It's a bit trickier than soldering to a PCB, but not impossible.

 

[Colin Mitchell]

I took it as white LEDs but red LEDs are much lower at 1.8v to 2.2v or it seems like 2.4v so the dissipation will be less. Gluing them to screws will dissipate the heat better than a PCB. 3 watts on 20mm x 20mm will need a LOT of effective heatsinking. Screws are the only answer and the heatsink will have to be 100 x 100 with lots of 30mm fins. Even a clip-on fan would be a good idea because LEDs get destroyed with heat.

 

[Chisatholm]

Thank you both for your suggestions & thoughts! I'm not quite sure how I'm going to proceed, but I'll let you know how I get on.