0
\$\begingroup\$

I have a PCB design with a long row of 1206 power resistors. Although only a few of them will be active at a time, they can dissipate up to 1 watt each, so I would like to heat sink them. If they get too hot they may overheat the components they're connected to, or just fail sooner.

Picture for reference:

pcb image

For electrical and space reasons I can't attach either pad of the resistors to a large plane. I've placed thermal pads under the resistors, but they cannot be electrically connected to it, only thermally.

My original plan was to "pour" thermal epoxy over the resistors after reflow assembly, this would spread the heat out from the active resistors, and also couple it into the nearby thermal pads. I tested that, and it works, but the issue is that the epoxy is much thicker than I anticipated (like frosting). This makes smearing it onto the board without covering adjacent LEDs quite difficult and time consuming, not really suitable for production.

I also considered placing a dab of thermal compound (not epoxy) under each resistor before reflow, but the issue with that is that the thermal compound must be able to withstand reflow temperatures. I found one (boron nitride), but it's a very watery consistency and it dries to a crust in seconds, so that also proved to be impractical.

Does anyone know of an elegant way to heatsink these resistors?

\$\endgroup\$
13
  • \$\begingroup\$ The usual way to heatsink SMD parts that don't have a dedicated thermal pad is through the contacts, with large copper areas on the PCB adjacent to them. \$\endgroup\$
    – Hearth
    Commented Mar 9, 2021 at 4:59
  • \$\begingroup\$ 1. split the resistors into two (calls for new layout) or use higher wattage resistor (digikey.in/product-detail/en/rohm-semiconductor/LTR18EZPFU10L0/…) \$\endgroup\$
    – User323693
    Commented Mar 9, 2021 at 5:00
  • \$\begingroup\$ Epoxy is a thermal insulator, how do you know chip temp? \$\endgroup\$ Commented Mar 9, 2021 at 5:06
  • 2
    \$\begingroup\$ Have you considered through hole resistors mounted in the air? \$\endgroup\$
    – Passerby
    Commented Mar 9, 2021 at 5:08
  • 3
    \$\begingroup\$ You must rethink your overall design for efficiency and not have to lose 1W per R or 45W total. that would need a CPU cooler. \$\endgroup\$ Commented Mar 9, 2021 at 5:18

2 Answers 2

2
\$\begingroup\$

One cleaner way to go is to use a thermal pad. Those come into sheets that you can cut to the appropriate size, you can then screw an aluminum heatsink to it. There is a long thin heatsink that would fit there.

enter image description here

The advantage of this method is that you don't have to do anything special for the SMT process and can be mounted afterward. It also avoids the mess of using epoxy / thermal compound, can be disassembled easily.

These pads come in different thicknesses from less than a millimeter to 5mm, so it "squeezes" when compressed, allowing for small height differences between components.

If you cannot afford screws, 3M manufactures a thermal pad that is sticky on both sides. I've been using that for years and yet one has come apart. It also comes in sheets that you can cut.

enter image description here

\$\endgroup\$
8
  • \$\begingroup\$ Are these pads you recommend electrical insulators? \$\endgroup\$
    – K H
    Commented Mar 9, 2021 at 5:22
  • \$\begingroup\$ That’s what I was thinking at first but if there is 45W load, not a chance to get low Rth on non-coplanar SMD’s with 1W/sqin extrusion 1-sided using tape \$\endgroup\$ Commented Mar 9, 2021 at 5:23
  • \$\begingroup\$ Yes they are insulators @KH \$\endgroup\$
    – Damien
    Commented Mar 9, 2021 at 5:23
  • \$\begingroup\$ @TonyStewartSunnyskyguyEE75 OP said " Although only a few of them will be active at a time". Also these pads comes in different thickness so it can handle some variance in height, but screws would be a better solution. \$\endgroup\$
    – Damien
    Commented Mar 9, 2021 at 5:24
  • 2
    \$\begingroup\$ What is on the other side of the PCB? | Is this a small or high volume product. An aluminium core PCB may be in order. \$\endgroup\$
    – Russell McMahon
    Commented Mar 9, 2021 at 5:55
1
\$\begingroup\$

Think outside the box.

If you're shedding that much power in the dropping resistors that you're exceeding the rating, consider a separate, lower VDD rail for the LEDs, made with a DC-DC step-down converter. Then you can optimize the voltage so that the resistor values can be lower, and therefore shed less heat for the given current.

Efficiency will be better and heat will be less. Win-win overall.

Also, do your LEDs need cooling? I'm guessing they do. Consider a thermal solution on the back of your board using thermal vias and a heatsink. Maybe mount the whole board to the sink with a thermal pad in between. If you have no backside components that would be easy to do.

\$\endgroup\$
4
  • \$\begingroup\$ There is a thermal solution on the back of the board, the LEDs are connected to it. The voltage drop is unfortunately necessary to even out and differences in LED forward voltage. The high currents can also cause voltage drop in the planes, and I don't want the LEDs to be noticeably effected if the bus voltage goes down by 100mV. \$\endgroup\$
    – Drew
    Commented Mar 9, 2021 at 6:05
  • \$\begingroup\$ Are your drivers current sources? If so, then this will do the compensating for you. Something to consider. \$\endgroup\$ Commented Mar 9, 2021 at 6:09
  • \$\begingroup\$ The daisy chainable drivers can be used that way (WS2811s), but I need to drive much higher current than the WS2811 is capable of, so I'm buffering the output with mosfets. \$\endgroup\$
    – Drew
    Commented Mar 9, 2021 at 6:16
  • \$\begingroup\$ You may need to go as far as having separate supplies tuned for each color's Vf. Then you can use minimum series dropping. At the power levels you're contemplating it might be worth it. \$\endgroup\$ Commented Mar 9, 2021 at 6:20

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