Well, we've decided to give SMT a try again, for making our guitar PCBs. So the first thing to do is make an SMT version of our earlier (working) guitar board:
We tweaked it a bit - making some pads bigger and spacing a few out more. We also had problems with the first version of the board, connecting the guitar strings to the two rows of six pads at the top and bottom of the board. If we didn't take care soldering, the round pads would lift quite easily, so we've changed them slightly for this board.
This board was designed in ExpressPCB, printed to PDF using CutePDF then opened in Inkscape to flip it horizontally (mirrored) so the final transfer was the right way around - when you're working with through-hole components, you actually want the PCB to be mirrored on the underside so print it as you see it works fine. For SMT work, the image needs to be mirrored so that when it's transferred onto the copper board, it ends up the right-way around.
With our fancy new laser cutter (how longer before it stops being fancy and new, and just becomes "our laser cutter" I wonder?) we cut a stencil design into an acetate sheet. We won't do it again - acetate stinks! And it burns and browns and covers everything in a sticky goo. But here's the solder paste stencil we made anyway:
Maybe the stencil holes aren't fine enough, or maybe we didn't use it correctly, but after applying solder paste through the stencil, we just ended up with big blobs of paste on the board. Manually applying the paste with a small paintbrush may not give a better result, but is about ten times quicker to do!
So we cleaned the board off and applied the paste again using our tried-and-tested method with a small brush
Even with quite a large tip on our soldering iron we got quite satisfactory results. Simply touch the tip onto the pad to start the paste flowing. Any paste in the gaps between traces simply burns off (or flows onto a tinned pin). To be sure of a good contact, we touched the tip of the iron onto the pin to be soldered to make sure the solder doesn't just flow under it.
You may see solder on the bits of board between traces. On a professionally manufactured board, these bits would be covered with solder resist. This is where we just applied solder paste all over the pins then heated it to remove the bridges between traces. Obviously, where the solder paste settled on these "in-between bits" it's resulted in excess solder on the face of the board. We *could* have etched all this excess away and not bothered putting a "solid plane" between traces - but then the Ferric Chloride would take an age to etch the board fully.
Here's the finished board, with multi-core cables attached on one side.
It's not a complete working board, but we've proved that we can solder SMT components just as quickly (we're not going to say as easily!) as their through-hole counter-parts.
On a professionally manufactured board, with solder resist and only the pads exposed, soldering would be even easier - but we found hand-soldering SMT components onto a home-made board relatively straight forward (and pretty quick to do!)
In the example above, we tried two different types of wire - what we're calling "half-pitch" (0.025") IDE cable and "regular" 0.5" IDE cable. We thought that the bigger cable would be easier to solder to the board but thought we'd try the two side-by-side to compare them. Although the regular cable is easier to solder (the traces can be bigger and slightly wider spaced) it's also more likely to fail when working on a large number of boards - because each core is multi-stranded, you have to take great care when soldering as the strands can fan out and easily create bridging with the other contacts. The smaller cable is a bit more fiddly to work with, but because each core is solid, once it's in place, there should be no worries about bridging with other wires.
We've yet to settle on a final design but whichever we choose, will probably be a compromise between easy of soldering and likelihood of failure - we want something that's easy to put together, but if we're likely to get a high failure rate, we'd be better off spending a bit more time on something a bit more fiddly, but be confident that it'll work once it's done!