If it’s to connect a sensor like the Nanovesta board close to the HDR-60, I’d recommend using flex cable, the metal layers are much the same work as laying out a PCB except the material used is now flexible mylar instead of FR4 material or others typically used for rigid boards. If you have a board layout group, you can send them the requirements for what you want done with the flex cable and they can lay it out and have it fabricated. Tell them you want the traces to be 50 ohm per signal side of the differential pairs and the differential pairs, both signals would be next to each other and follow the same path and lengths within a diff pair, and between diff pairs. 100 ohm differential termination on die at the receivers would be best, or an external differential resistor if the FPGA doesn’t have them. The ECP3 does have on die termination capability as does the LatticeSC/M and MachXO2 (but you’ll still need a 200 ohm external termination to parallel with the internal 200 ohm termination).
If it is for use with long cables, be sure to check the loss of the cable vs. frequency over the length of it.
You’ll want the loss to be low over the length, a rough equation for the max cable length would be:
cable -3db frequency for specified length > 2.5/(bit rate)
You’ll also want length matching between signal pairs, this is somewhat a given when using CAT5 cables with connectors.
Coax cables would give the lowest EMI, but they are expensive vs. a cheap CAT 5 cable you can pick up most anywhere.
If it’s ribbon cable between boards, pair up the signals side by side over the cable, and possibly a GND wire between each pair to keep the crosstalk down between pairs.