Aside from all the procurement and design issues, making a board that's widely useable depends quite a bit on the board quality, particularly the precision of registration of solder mask and pads.

If you can get the solder mask to run precisely between the pads on a TQFP, it's probably not as difficult to solder cleanly as one might assume. The key, however, is that the solder mask has to allow the IC pads to contact the metal only where they should.

My recent experience is limited pretty much to dry-film solder mask, rather than the more traditional silkscreened mask, so I may be off base with this, but I've found that properly registered and produced solder mask, held to reasonable (~1 mil) tolerances will yield boards that are inherently hand-solderable. With a good mask, even one as half-blind as I am can solder the leads and subsequently remove the excess with solder-wick. It's all in the mask layout.

That opens up a range of possibilities, including the provision of a site for a TQFP-packaged JTAG programmable CPLD, or even FPGA of considerable size. That way, if one wants to use external memory, packaged in a TSOP or SO package, it's in the cards, just as it wold be if there were a DIP that were in use.

One thing I'd consider, in your position, and I'm not in your position, would be a 3-volt MCU and a XILINX CoolRunner-II CPLD, say, the XC2C256, in TQFP144. There's quite a lot one can do with that part, and if one's using 3 volt memory and a 3-volt MCU, there's no need to worry about the supply voltage. XILINX, Altera and Lattice make some pretty interesting FPGA's that don't require a boot ROM, so that might prove interesting as well. I suspect a guy could amuse himself for quite a while with any of those options.

If you want, you can provide the programming file for the external latches with which to interface to external memory and I/O, and you can probably depend on some of the guys who use programmable logic to kick in some elementary building blocks by means of which one could implement hardware features that otherwise would be pretty cumbersome to accomodate on a not-too-complex board. To make the board widely useful and flexible in its application, it would be important to have a prototype are to which one could attach various common connectors. If you take a look at the SAMTEC repertoire, you might get some ideas as to how one might bring out signals.

You can consider this or not, as the one doing the work is the one who gets to make the decisions.

Good luck to you!

RE