Comment by torginus

Well, LFSR RNG-s are pretty efficient in terms of HW space. You take a single-bit wide shift registers of length n, and feed back its output to the beginning XOR-ed with predefined bits in the shift register.