RTPAlso known as Real-time Transport Protocol, RTCP

What is inside an RTP packet

RTP usually runs over UDP. Each packet has a 12-byte header followed by the codec payload:

• Payload type: identifies the codec. Static types from RFC 3551 include 0 = PCMU (G.711 µ-law), 8 = PCMA (A-law), 9 = G.722 and 18 = G.729; codecs such as Opus use dynamic types 96–127.
• Sequence number: increases by one per packet, so loss and reordering can be detected.
• Timestamp: sampling instant, used to rebuild timing and measure jitter.
• SSRC: a random identifier for the stream’s source.

A typical voice stream sends 50 packets per second (20 ms of audio each). With 12 bytes of RTP, 8 of UDP and 20 of IPv4 header, overhead is 40 bytes per packet, which is why low-bitrate codecs still need 24 kbit/s or more on the wire.

RTCP: the control channel

RTP is paired with RTCP (RTP Control Protocol), traditionally on the next odd port. RTCP sender and receiver reports carry packet counts, fraction lost, interarrival jitter and round-trip time, which is where most VoIP quality statistics come from. Encryption is added with SRTP (RFC 3711).

Common RTP problems

Most “no audio” and “one-way audio” faults are RTP faults, not SIP faults: a firewall blocks the negotiated UDP port range, or NAT causes SDP to advertise a private IP address. Asterisk-based systems commonly use 10000–20000 for RTP. An SBC or media relay can anchor RTP to fix this. GSMCalls reports loss, latency and jitter per trunk and device on the platform, which helps locate which leg of a call is degrading.