10
\$\begingroup\$

I have prototyped some DC-DC converter modules, and they work well. However due to size, I don't have a huge amount of copper pour for them to dissipate their heat. I have used 2oz copper but the bottom layer copper pours get really hot at higher current. Even through the solder mask!

I experimented with adding a heatsink to this bottom side of the PCB with some thermal grease and the results were surprisingly good. The heatsinks get hot under heavy load. The buck-boost-inverter went from maxing out at about 1.5A to being stable at 2.5A! This is my current setup: First try at heatsinking

However I can't help but think that I can improve this. I am thinking of removing the bottom solder mask around the heatsink area for better heat transfer. Also I want to use a Sil-pad instead of thermal grease for easier assembly, and because I don't want to risk shorting different copper pours when the solder mask is gone. Like this: Heatsink new idea

So my questions are:

  • Is this a good way to do this? (given my limitations)
  • is there anything that could affect the long term life of my PCBs with this setup?
  • Are there any other suggestions people have?

Thanks!

\$\endgroup\$
3
  • \$\begingroup\$ use a fan, any size fan \$\endgroup\$
    – analogsystemsrf
    Commented Jul 20, 2019 at 17:32
  • \$\begingroup\$ I suggest that you experiment before you finalize on a solution, that you run the numbers, and that you do all of this in consultation with your manufacturing engineering department. I lean toward suggesting that you retain the solder mask and use a thin thermally conductive adhesive -- but you need to experiment to find an interface that meets your thermal needs and is manufacturable. \$\endgroup\$
    – TimWescott
    Commented Jul 20, 2019 at 17:33
  • \$\begingroup\$ I would make the final step, and solder the heatsinks onto the back of the PCB. No silicone pad, no soldermask. Best heat transfer to the heat sink you can get. \$\endgroup\$
    – Nick Alexeev
    Commented Jul 20, 2019 at 17:47

3 Answers 3

Reset to default
6
\$\begingroup\$

This link contains useful information, also this video.

Typical solder mask has 20-25µm thickness and 0.2 W/m.K thermal conductivity. This means a 1cm2 area of solder mask will have a thermal resistance of 1°C/W. This can be a problem... or not, that depends on your application and how much power is dissipated. For a few watts, an extra 1°C/W doesn't matter, just do the calculation. For a larger contact area, thermal resistance drops accordingly.

However, soldermask has another very important role. If you use immersion gold, large copper areas without soldermask may result in a thick gold layer, and your PCB fab will be asking who's gonna pay for the extra gold. If you use HASL, solder thickness may not be even, which will require a thicker interface material to even out the bumps, and increase thermal resistance too. There could even be a little drop of solder left over on the edge of a via, and then your heat sink won't be flush, and if you try to remove the bump by hand, it'll make a mess. And of course, wave soldering would result in a mess too. So... soldermask is nice to have.

Anodized aluminium is insulated by the oxide layer, but it can get scratched off. So, a bare heat sink on top of vias with just conductive grease between them would work... in theory... still a bad idea. It's better to tent the vias and protect them with soldermask.

Thermal grease is better than silpads because it is thinner. However, silpads are insulating and thermal grease is not. Why not simply check the datasheet of your silpad and calculate the thermal resistance versus contact area, and check if it works?

Another option is a SMD heat sink. Pros: thermal conduction path is 100% metal. Cons: thermal path has to go horizontally through the copper layer, which isn't that efficient.

Anyway. If your IC only dissipates a few watts, keep the silkscreen or use a SMD heat sink.

\$\endgroup\$
2
  • \$\begingroup\$ Thanks, this is the most helpful comment. The surface finish will be HASL, and I hadn’t thought about the advantage of having a more flat surface to work with. The vias are fully tented already on the bottom solder mask. The only heatsinks that I can find which fit perfectly, and are reasonably priced, are un-anodized aluminium. do you think that the thin layer of solder mask and thermal grease will be enough to keep separation over a long time period? The highest voltage potential between any two pads is 27V but potentially I could design it so that it’s only 12V on the heatsink area. \$\endgroup\$
    – Benjamin
    Commented Jul 21, 2019 at 10:54
  • \$\begingroup\$ How much power do you want to dissipate? \$\endgroup\$
    – bobflux
    Commented Jul 21, 2019 at 11:56
1
\$\begingroup\$

So, if you want to avoid the risk of shorting copper pours, stick with the solder mask. It's probably lower in thermal resistance than the difference between a pad and a thin layer of heat conducting paste.

Also, how hot does the top side of your IC package get? Maybe a small, second heatsink stuck to the top of the package is helpful, too.

Also, if you mount your board upside down, convection will greatly improve the cooling efficiency of your fins.

\$\endgroup\$
5
  • \$\begingroup\$ I will still need to have a thermal interface material between the bottom of the PCB and the heatsink anyways, so what is the benefit of keeping the solder mask? I can't just use the solder mask to thermally couple the copper to the heatsink. The top of the IC packages are remarkable cool. When the heatsink get's warm, I can't really feel any heat from the top side components. When the heatsink is HOT then they get warm but, but nothing compared to the heatsink. I don't think the top side of the components radiate heat very well. \$\endgroup\$
    – Benjamin
    Commented Jul 20, 2019 at 16:46
  • 1
    \$\begingroup\$ soldermask is a very thin layer. heatsink paste is a very thin layer. usually, the thermal transport through these two layers is better than through no soldermask + pad. So, removing the soldermask and replacing the paste by a pad is a downgrade to your system, likely. \$\endgroup\$
    – Marcus Müller
    Commented Jul 20, 2019 at 16:49
  • 2
    \$\begingroup\$ If your production people don't kill you for it, use thermal epoxy and keep the solder mask. The stuff I've seen is basically a two-part epoxy with thermally conductive filler, and a formulation that can stand up to high temperatures. \$\endgroup\$
    – TimWescott
    Commented Jul 20, 2019 at 17:31
  • \$\begingroup\$ @MarcusMüller, thanks that makes sense. I think I will stick with the solder mask to also gain the benefit of it being flatter surface area. Tim Wescott, I’d rather not use epoxy. I want to be able to rework these if necessary, and that is all but impossible with a heatsink permanently attached to each plane. \$\endgroup\$
    – Benjamin
    Commented Jul 21, 2019 at 10:59
  • 1
    \$\begingroup\$ @Benjamin sounds like you'd want to have some screwholes around the area then to spring-loadedly fasten your cooler to the board then (and thus press it firmly against the PCB) \$\endgroup\$
    – Marcus Müller
    Commented Jul 21, 2019 at 11:19
1
\$\begingroup\$

I think it’s a fine idea. About 2/3 of an IC’s power gets shed in the PCB.

The reason you don’t see this approach more often is that many designers want to use the underside of the board for bypass capacitors, and they in general have a mechanical bias for tall components on the top side.

The challenge you will have with fully exposed copper flood is in assembly: you won’t be able to use simple wave soldering without a blocking plate if you need that, and its also an issue for components that are applied by SMT.

Maybe split the difference and use a stipple pattern of exposed copper in the mask, and ensure any traces that go from the exposed area to component pads have soldemask dams.

More stuff:

Phase change material will give better performance than the silicon pad, at increased cost. Thermal compound also.

Using epoxy - that’s the lazy way, assembly people hate it. Work out how to mount the heatsink using push-pins or a spring clip.

\$\endgroup\$
2
  • \$\begingroup\$ Do screws work? \$\endgroup\$
    – DKNguyen
    Commented Jul 20, 2019 at 18:33
  • 1
    \$\begingroup\$ Not really as well. It’s hard to ensure consistent pressure on the heatsink and thermal layer. Ideally you want even, repeatable force that won’t damage the thermal material. This is especially important for phase change which liquifies when it gets warm. \$\endgroup\$
    – hacktastical
    Commented Jul 21, 2019 at 0:05

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