Seems you don't need/want a 40x40 key keyboard, just 40-key + "Enter" one, and you read in separate lines, each 40 bit long. If you don't want 5-port '51 derivative, you'll need 5 3-state buffers, on a common bus. I can imagine you use switches instead of keys, so the user sets state the state of each bit (0/1) of the keyboard, the states for now are discarded but kept by the mechanical switches. Then the user presses "enter", and 5 pins from one port, in sequence enable each of the buffers, and 8 bits of the sequence appears on other port. Unless you actually want to wire all 1600 switches and enter everything at once, in which case the design wouldn't be that much different (next 5 lines plus 5 8-bit latches to select a row to be read)

But really, entering binary would be a terrible work, and we here won't do the design work for you, only solve technical details.