Depth expansion gives you more words. Each chip owns its own block of addresses:
- The low address bits, as many as one chip needs, go to every chip. They pick the word inside a chip.
- The high address bits go to a decoder. Each decoder output drives one chip select, so the high bits pick which chip.
- The data lines are shared. Only the selected chip drives them; the others sit at high impedance.
Chips needed = wanted words ÷ words per chip.
Decoder bits = system address bits − chip address bits = log₂(number of chips).
Each chip's block starts at a multiple of the chip size. To find which chip holds an address, write it in binary: the top bits are the chip number, and the remaining low bits are the address inside that chip (the offset). Equivalently, chip = address ÷ chip size and offset = the remainder.
This is the decoder expansion idea applied to memory: the decoder outputs are one hot, so exactly one chip ever answers.