On an efficient VLSI architecture for the multirate additive synthesis of musical tones