The Internet is increasingly being used to carry real-time voice traffic. Users of real-time voice services are sensitive to variable audio quality. The quality of packet audio is largely determined by the mouth-to-ear delay and the packet loss. The contribution of this thesis is to provide techniques to improve the packet audio quality: dimensioning links specifically for packet voice communication, modelling the packet audio arrival process at a receiver, measuring connectivity quality in wide area networks, and reducing delays in end systems.