I have lain out the board with this approach in mind. It's pretty close to the data sheet layout, and if I later want to have boards made professionally I can easily switch to a thermal via pattern without touching the rest of the circuit.
I'm looking around for an adhesive plastic film to try. Any specific recommendations?
--- In Homebrew_PCBs@yahoogroups.com, "Phil@..." <yahoo@...> wrote:
>
> I have no direct experience but I know what I would do. I would put the
> rivets in the thermal pad area first, cover the rest of the board with
> adhesive plastic film, and sand the rivets with maybe 320 grit wet until the
> heads were flush with the plastic, then bend the SMD leads slightly to
> compensate for the extra thickness.
> --
> Phil M.
>
> ----- Original Message -----
> From: "Kaelin Colclasure" <kaelin@...>
> To: <Homebrew_PCBs@yahoogroups.com>
> Sent: Wednesday, July 10, 2013 7:55 AM
> Subject: [Homebrew_PCBs] How to tackle thermal pads on SMT packages?
>
>
> > I'm looking to stretch my luck by trying a design that uses all SMT
> > packages. At the moment I am pondering how to tackle the thermal pads
> > commonly found on SMT power converter packages (for an onboard buck
> > converter).
> >
> > The data sheets commonly specify that a pattern of plated vias should be
> > used to conduct heat from the thermal pad of the package on the top copper
> > layer to the bottom layer. The idea, obviously, is to provide more copper
> > surface area for radiating away waste heat. My problem with this is that I
> > don't have a way to plate through holes. I am using old-fashioned copper
> > rivets to connect between layers. And while the rivets are smashed pretty
> > flat by the tool, you can definitely feel where they are, and the tool
> > also tends to leave a slight deformation in the surrounding copper layer.
> > I am concerned that this could cause problems with reflow soldering.
> >
> > I thought about simply making a larger copper area on the component side
> > of the board, but the relevant data sheets are also adamant that circuit
> > paths to supporting components must be as short as possible.
> >
> > I've got a couple of ideas to try:
> >
> > 1) Use riveted vias in a copper area that's as close as possible to the
> > package without actually being embedded in the thermal pad landing. But
> > will they still conduct away enough heat?
> > 2) Forget the through-holes and stick some kind of heat sink on top of the
> > package. But again, since the package is quite small, and not designed for
> > a heat sink on top, will it work?
> >
> > Any other suggestions? Has anyone here tackled this sort of thing before?
> >
> >
>