Jitter buffer and Opus: good voice quality over the network

In short: In telephony and video calls the arrival times of packets vary (jitter). A jitter buffer smooths this, and codecs like Opus conceal lost packets.

In more detail: Speech must be played back evenly, but packets arrive irregularly. The receiver collects them briefly and plays them back at the right pace.

In Depth

• Jitter: Variation in delay due to changing queues in the network. Measured in milliseconds. See reading a speed test properly.
• Jitter buffer: Holds packets back for about 20 to 200 ms, orders them and plays them evenly. Larger = steadier, but more delay. Adaptive buffers adjust.
• Packet loss: Packets that are too late count as lost. Counter-measures: packet loss concealment, forward error correction, retransmission requests for video.
• Opus: Open codec for speech and music, 6 to 510 kbit/s, delay from 5 ms, common in WebRTC, Discord and Teams. See audio formats.
• Improving: Cable instead of Wi-Fi, QoS for voice traffic, avoid congestion.
• Protocols: RTP carries the media, RTCP reports quality.

See also: UDP