UNISON: A Novel System for Ultra-Low Latency Audio Streaming Over the Internet

Although modern computer networks do provide very high data rates and very low network jitter, consumer-grade systems for low-latency media streaming and audio conferencing still need research. Not only the network itself but also the timing jitter at the sending and receiving endpoints may cause discontinuities in the media stream. Hence, the use of jitter buffering at the receiver side is mandatory. To keep the end-to-end latency low, the average number of samples in the jitter buffer should firstly be as small as possible and secondly, constant over time. In practice, the latter requirement is very difficult to fulfill: As the sender and receiver do not share a common sampling clock, clock drift causes a slight deviation between the number of samples recorded by the sender and the number played by the receiver in a certain time interval. This effect causes frequent buffer under- or overruns if an extremely small jitter buffer is used. In this paper, the authors present their novel audio streaming solution that aims to achieve end-to-end latencies lower than 20 milliseconds and uses a tunable hardware oscillator circuit to compensate clock drift.