Method and Apparatus for delivering audio signals from a source node to a destination node on a network. The apparatus uses a number of switches that transmit prioritized data on a packet network. The switches are coupled to a number of send/receive nodes for sending and receiving digital audio signals on the data network. The audio packet size and the receive buffers are sized to store a minimum possible number of audio samples to minimize latency in processing audio signals arriving at said receive node, but still ensure audio delivery without interruption due to packet data network delay. An additional feature of the invention is recovery of clock synchronization over the same data network by novel arrangement of transmission of timing packets on the network. By sending a multiplicity of packets at irregular intervals a minimum network transit delay can be determined by each of the receive nodes which allows the receive nodes to filter out packet network transit delay error and maintain accurate local clocks.