Not in this picture are the other components (SSD1963, ESP32, Clk etc.). I have messed up circuit boards before in the early stages of prototyping, hence I am designing this board to be able to be broken up and be used as individual breakout boards and breadboarded in the event that does happen.
There is an alternative strategy. If you're prone to messing up pinouts, you can route a bus from e.g. display controller to connector, with jumper resistors in series. That might look like so:
(Own work.) This was a display controller (FT800 I think?) sending 24-bit parallel video to an FFC connector.
This design was driven by two instructions:
- Series resistors where possible, either to jumper for patching of prototypes, or for source termination of logic signals
- Test points on every node so that component assembly can be tested automatically (flying probe)
With top-side placement and routing, test points mostly look like vias here. I think I didn't use bottom-side SMT test pads at all back then, actually; I would've added those to the library not long after this board was made.
Speaking of test points, if you want to get actual wire into the pad, I'd recommend 0.5-1.0mm ID; the low end is enough for fine wire, enameled, wirewrap, etc.; the upper end fits most component leads, header strip, etc. If in an array (like this), placing them on a regular grid 2.54mm allows header installation. (Or something else nearby and convenient; 2mm would also work, and 2mm header is readily available. Finer pitch, I'd recommend against; it's less available, and harder to use.) Do make considerations, like every-other grounded, particularly on relatively fast signals like this.
These test points might even be 0.3mm ID, which is still enough to fit wirewrap wire (30 AWG), if a bit fiddly; but the resistor pads (0805s at that) are easy enough to tack a wire onto, which we did from time to time.
Note the design: the board is 2-layer, ground plane poured on both sides, stitched with vias (you can just see the dark, tented vias around the bus here, around the top-left bus, connectors, etc.), which makes traces into microstrip transmission line, which has predictable impedance and propagation velocity. Where buses cross, overlap area is kept to a minimum, and stitching vias are placed all around the crossing to minimize ground-return path length. (There's nothing routed under this video bus, of course; the controller's SPI interface is routed underneath the top-left bus, around R87-R85.)
Even if long jumper wires need to be routed over such a board, they will also manifest as microstrip (in this case, some hybrid-dielectric, round wire over plane, geometry; basically the same characteristics as planar microstrip). It's typical to glue such wires down to the board, enforcing a low height above plane -- you even see boards (moreso back in the day, 70s to 90s) with such wires, in production -- it's not the end of the world, to have a few jumpers, anyway. (Less so today, what with validation tools, and PCB fab costs being so low, not to mention consumer product density so high you can't get a patch wire in anywhere anyway. But there's still niche stuff like this where it can happen, at least in prototype quantities like this exact board was.)
You can't apply this practice fully everywhere; high-pin-count devices, it gets very tedious (as you can see here!). Decide if a compromise is necessary, or check the pinout and signal type/level obsessively to direct wire it. Don't jumper out power pins: verify and triple-check them against the datasheet instead; you need the short path length to nearby bypass capacitor or ground plane. (As consolation, you could still leave a short stub track, which can be cut with a craft knife, and the soldermask scraped away to expose solderable copper. Patching becomes more tedious, but you've made sure it's very unlikely to need.)
Alternately, boards are cheap. Just wing it. It costs a couple weeks time, and not much budget these days, to spin another board. If it's not a professional project, just something you're whacking together, time is easy to write off.
