SMD design - ouch! Personally I'd advise through-hole design where possible. Many TI'ers are now of a certain age where hands and eyes are not as good as they once were, and it's likely to put some potential builders off.
Stuart, thanks for the comments. I have to say I am not that young myself either... Even if I often like to see myself that way
I debated with myself about going entirely SMD. I eventually settled to go that way, because many components were going to be SMD anyway: the CPLD, the micro controller, USB connector and SD card socket. But to be honest, I wanted to go this way because I wanted to see if I could do it

so this is my small mount everest to conquer. We will see how it pans out. The good news is that the boards turned out to be very affordable - the order went in yesterday. If it turns out to be difficult to construct, I can modify the design with through hole components for the parts where that option is available. I tried to align parts in a way where there would be some maneuvering space around them - towards the end of the layout I realized I had placed some parts too close to each other to facilitate hand assembly.
To continue on the SMD theme, I also want to say that I believe the board layout was easier to do this way. Currently the only through hole parts are the pin headers and the crystal. Going through hole with the other components would probably have required a larger board. I attached a printout of the top and bottom of the board (I eventually settled on Eagle, dare I say due to the fact it seemed to have more complete SMD part part libraries...). With the XC9572XL and its 100 pin package there is quite a bit of routing pressure to bring the signals out. Even though the board does not look like much, it took me probably 50 hours or so to manually place and route after the first version of the schematic was done. A lot of that was learning curve and early design changes, finding the appropriate packages, but also just getting the signals where they need to go. I think the next board will go very much faster.
I see the schematic PDF with lots of lovely interfaces on it. ;-) How does one access such a device? Is it all memory mapped ports, or ...?
Basically the idea is that all signals from the Mini Cortex CPUBUS go to the CPLD. There is a 11 bit parallel bus between the microcontroller and the CPLD. In addition the SPI bus of the SD card goes to both the micro and the CPLD. With that, the actual interface topology will be up to the CPLD configuration.
You are right, my idea is that the ports will be memory mapped and actually on the mini cortex that is the only option as the CRU bus does not arrive at the CF connector. This also matches the FPGA design I did on the breadboard - the SD card interface is memory mapped.
There are 8 memory locations that are available. The existing SPI design requires two (one for SPI data read/write, one 1 bit register for SPI chip select and status read). I am planning that on some of the remaining memory locations there would be access to the extension connectors of the CPLD (12 signals - more would exist but I wanted to get the board done and did not want reposition the parts yet again to route more signals out). A couple of registers would be a mailbox interface between the micro and TMS9995. That mailbox would be the means of interprocessor communication, providing access to the vast array of peripherals available on the micro. Of particular interest there are the serial port and the USB port. I also brought out the I2C bus of the micro to some header pins.
The NXP LPC1343 microcontroller is tiny in size but has an impressive amount of stuff in it. It has a 32-bit ARM Cortex M3 processor core running at up 72MHz, 32K of flash ROM and 8K of RAM. It has a very handy boot loader feature, basically you short one pin to ground and it's built in boot ROM makes the chip look like a 32K FAT12 USB memory device, allowing one to drag and drop in a 32K firmware image. I haven't tried that yet, in my test setup I have used a raspberry pi connected to the serial port of the micro and I've used the serial boot loader feature of the chip. In case one wants to use more of the micro controllers features on this board, one can just route the 11 bit bus going to the CPLD thought the CPLD to its extension connectors. I tried to keep things flexible there.
I also wanted to go with these particular chips and packages (XC9572XL and LPC1343) because for both there is a pin compatible upgrade path (XC95144XL with 144 macrocells, and LPC1347 with 64K Flash).
Erik