I'm looking at DIY projects where a COB LED is cooled by an all-in-one CPU liquid cooler, however the contact area in a CPU heatsink will only fit one high-power LED. I'm considering how I could use several, perhaps less powerful LEDs with a single heatsink.

Is it feasible to mount LEDs on a larger copper plate, which is then tightly fixed to the heatsink (screws + paste, or thermal adhesive)? What can be expected from and what should be considered for this solution?

UPDATE - already got a great answer with general principles, but here are the specific numbers for my current project:

  • 4 COB LEDs of 25W, 19x19mm large. Will be at least 70mm apart because of lenses.
  • Cooler is a Liquid Freezer 120 with a surface diameter of 60mm and recommended max TDP of 250W.
  • Based on the answer below I'm considering a 6mm thick aluminum disc with a diameter of 144mm, for mounting the LEDs on and then attaching to the cooler.
  • 1
    \$\begingroup\$ It's feasible but you sound like you are more in need of opinions relating to possible LEDs you might consider and this doesn't make it a good fit for the site. Try asking a much more specific question like "I have these LEDs and they generate this much heat and will they overheat on this plate arrangement". \$\endgroup\$ – Andy aka Feb 6 '18 at 13:54
  • \$\begingroup\$ I'm literally wondering how much of the capacity of a 100-200W TDP heatsink might remain if I expand its area this way (say 4x the original contact area), considering extra distance and inefficiencies in any thermal compound used for the extra layer. I apologize if it's the wrong type of question. \$\endgroup\$ – Toerndev Feb 6 '18 at 15:40
  • \$\begingroup\$ A possible benefit is that if using a cross-shaped piece of copper, the "wings" outside of the heatsink area could be slightly bent to point each LED at a different angle. But I can't find examples of this kind of mod so I'd like to know if it's advisable. \$\endgroup\$ – Toerndev Feb 6 '18 at 15:45
  • \$\begingroup\$ Indeed it is possible to mount LEDs on a copper plate connected to a heatsink. What can be expected will depend on the LEDs, the copper plate and the heatsink, of which you don't provide any information. \$\endgroup\$ – Dmitry Grigoryev Feb 8 '18 at 8:15

This sort of thing is reasonably standard in the RF world where the usual problem is small geometry power devices mounted onto very large typically aluminium heatsinks, power levels can exceed 1,000W. We typically machine a copper block then solder or (vastly inferior) screw the power device to it in order to increase the contact area between the heatsink and the spreader by many times over the contact area of the raw device.

The phrase to search for is "Heat Spreader", and I would note that surface finish is reasonably critical, face milling of both components is commonplace.

Your case is slightly the reverse as your loads are bigger then the heatsink area, I would go for a bit of 10mm MIC-6 or similar ally tooling plate (Copper is not really that useful to you as you do not really have the power density to make it worthwhile and it is a pig to machine), tooling plate is nice as it has already been face milled so will be reasonably flat, something that cannot be said of extruded bar stock.

Another option would be a single sided metal core PCB made with a very thick substrate as a heat spreader.

TDP is not really a useful number, being as it combines a mess of thermal resistances and thermal limits (which may or may not apply to an LED), far better to use the thermal resistance (measured in kelvin per watt) which makes doing sums to see how hot the die will actually get fairly straightforward.

| improve this answer | |
  • \$\begingroup\$ Thanks! That's several new keywords for me! Regarding the 10mm MIC-6, is there a term or search phrase to use to find a ballpark suitable thickness for a given effect? E.g. whether 1-2-3mm copper/alu plate would suffice for removing heat from 5-10-20-30W LEDs. Assuming that the plate is then attached to a decent heatsink, and temporarily disregarding the part about surface finish. \$\endgroup\$ – Toerndev Feb 7 '18 at 11:16
  • \$\begingroup\$ Ignoring surface effects (Which cannot really be ignored), the thermal conductivity of pretty much any metal is easily looked up. Then it is just a matter of some modelling in a spreadsheet to see how you want to trade off the heatsink size, transfer plate temperature drop and junction temperature. \$\endgroup\$ – Dan Mills Feb 7 '18 at 12:39

Not the answer you're looking for? Browse other questions tagged or ask your own question.