RTP

RTP (Real-time Transport Protocol) is the standard protocol used by VoIP phone systems to carry the actual audio (and video) of a call, in real time, over an IP network.

RTP vs SIP - who does what

• [[sip|SIP]] handles the signalling layer - call setup, transfer, hold, hangup.
• RTP carries the media - the actual audio you hear, or the video you see.

They work together. SIP says “let’s start a call between these two endpoints, using this codec, with media flowing to these IP addresses”. RTP then carries the audio between those endpoints once SIP has agreed the terms.

How RTP works

• UDP-based - fast but unreliable (no retransmission of lost packets). Acceptable for voice because retransmissions would arrive too late to be useful.
• Sequence numbers - every packet is numbered so the receiver can reassemble in order.
• Timestamps - so the receiver can play audio with correct timing.
• Codec payload - the audio data, compressed using the negotiated [[codec]].

SRTP - Secure RTP

SRTP encrypts the audio payload so eavesdroppers can’t hear the conversation if they capture the packets. Standard in modern business VoIP. 3CX uses SRTP by default for all hosted deployments.

RTP and call quality

Three RTP-related metrics drive [[mos|call quality]]:

• Packet loss - packets that don’t arrive cause audio gaps.
• Jitter - variation in packet arrival times. Buffered out at the receiver but extreme jitter causes drops.
• Latency - one-way delay. Below 150ms is fine; above 300ms is bad.

3CX call detail records report all three so you can spot networks or trunks degrading over time.

RTP and firewalls

RTP uses a range of UDP ports (typically 16384-32768) so the firewall must allow this range outbound. Without proper firewall config, signalling works (SIP completes) but no audio arrives - the classic “one-way audio” problem.

See also

• [[sip]] - partner protocol for signalling
• [[codec]] - what’s inside the RTP payload
• [[mos]] - call quality affected by RTP performance
• [[sbc]] - SBCs sit in the RTP media path