here's the board with most of the pcbs pins soldered in place
on the top side the PCB pins stand about 0.9mm proud of the PCB surface
The through-hole vias are big clumsy PCB pins from Maplin and some of the pads didn't line up perfectly on both sides (though never more than 1mm off, so all the pins did actually hit a trace on the opposite side).
The heads of the pins are proud of the top surface which is a bit of a pain - we're going to have to create a much fatter separator layer, to avoid getting lots of dimples appearing on the top (printed) layer, once assembled. But it should still prove that the PCB at least works (or doesn't, as the case may be) before we send the design off to be manufactured.
the final board, all soldered up and cut to size
adding magnets to the edges of the board
To be honest, Steve did most of the assembly work. He has an eye for detail that means everything is cut and aligned "just so" - far better than I might have left it (copper boards cut with a hammer and joined with string and sellotape).
Soldering the magnets onto the edges was more fiddly than we'd expected.
Firstly, of course, they stick to the tip of the soldering iron! And if you let a magnet get too hot, it starts to break down and stops being so magnetic. And these magnets were mounted into acrylic edges on the board, so as well as stopping the magnets from breaking down, we also had to make sure the plastic didn't melt while the solder was applied!
The final board looks very good and all the pads have been tested for continuity.
Assembly of lots of boards might be problematic with magnets as edge connectors.
Personally, I love the idea of putting two boards edge to edge and they just go "clink!" and join together. But it's fraught with difficulties. If any one of the four magnets is slightly more proud than the others, there's a potential for a break in the continuity of the entire circuit. The magnets will pull the boards together as closely as possible - but if one it further forward than the others, not all magnets might actually touch.
To solve this, future versions will use pin headers and sockets to connect the boards together. This should also make assembly of each board section easier too, as well as make the connections between boards a little more robust.
But this board "works" in that the pads on the top can be bridged with a piece of conductive material, and the continuity traces out to all the correct pins on the microcontroller pads.
Next step is to program up a PIC (we've got some 28 pin 16F722 chips here to try out) and actually send the correct serial data back to a host controller as each pad on the surface is pressed/released.
Exciting times indeed!