So today, I did this:

That's a SchmartBoard with an AD5061-2 on it hooked up to my PDIP ATMega328p on the breadboard. I took my firmware test code and uploaded it and got a very, very slow ramp. Even clocked at 20 MHz, it takes ~488 µs to do the bitbang upload into the D/A chip. I'm confident that using the actual SPI functionality of the ATMega I could make it go faster than that, but the truth is, this only has to at most do one D/A operation per second, so it doesn't really matter. After adjusting the firmware to do a divide-by-16 (that is, increment the value by 0x10 instead of 1), I was able to get this:

And, of course, I was able to confirm that the part fits perfectly on the footprint on the board.

The next thing is to test the fan-out chip. I'm going to try and stick one between the oscillator and the controller, but I don't have a lot of doubts that it's going to work. I think I will pull the trigger on the boards.

P.S. Replacing digitalWrite() with direct operations on PORTB results in a 10x improvement. Writing to the DAC now takes 48 microseconds.