First you will need an FPGA development kit, with a minimum of a high gate count device like a Spartan 6 LX150T.  You will also need a basic knowledge of VHDL and the ability to load a bit stream (there are a few programs that can do this out of Xilinx ICE - which is over $4000 for high gate count)

Once you have a bitstream figured out and a board reference design, you will need to lay out or obtain a board layout or talk one of the defunct FPGA Mining developers to give you their IP. Since the device that you are soldering to the board is a 484 pin BGA and each chip is worth over $150 you will want to have these built at a reputable PCB assembler (Advanced Circuits is good).

This is the stage that the whole house of cards falls away for me because to get to this point I will have spent $5000+ (probably more like $10k) and I haven't hashed a single block.  The best bitstream in FPGA was 830 MHash/s and more likely on an amateur board would be 400 MHash/s.  When ASICs are fully released in the next month, difficulty will go many times higher than it is right now (100,000,000+) which means you will MAYBE get 1 coin every year with 4 Spartan-6 LX150s running at full speed and that is if they don't burn out (FPGAs are reliable when cool and slow, not good conditions for mining).

If you want to get into mining you need to go to Butterflylabs or Avalon-ASIC and get in line with your $1500 to get a device.  If you order this week you might receive your miner by June or July.

I hope I don't sound too *negative here, this information is based on my experience over the last couple weeks researching FPGA for mining.

EDIT: It is totally noble to want to advance FPGA, it is just really really really hard.