Learn / Field guide

RTP: the audio in a call

RTP carries the actual voice: a small packet every 20 ms in each direction. Sequence numbers reveal lost packets and timestamps reveal jitter, the two things that make calls sound bad.

15 RTP fields, explained

Field names are the ones Wireshark and TShark use, so you can search for them or use them in display filters.

rtp.setup

Wireshark's note about which SIP message set up this audio stream.

rtp.setup-frame

The packet number of the SIP/SDP message that set up this stream.

rtp.setup-method

The protocol that set up this stream, such as SDP.

rtp.version

The RTP version, always 2.

rtp.padding

Whether padding bytes were added to the end of the packet.

rtp.ext

Whether an extension header follows the fixed header.

rtp.cc

How many contributing sources are listed, used when a mixer combines several voices.

rtp.marker

A marker bit, often set on the first packet after silence.

rtp.p_type

The payload type: which codec the audio uses. 0 is G.711 µ-law (PCMU), 8 is G.711 A-law.

rtp.seq

The sequence number. It goes up by one every packet.

Why it matters: Gaps in sequence numbers are lost packets; lost audio sounds like clicks or dropouts.

rtp.extseq

The sequence number extended past 65535, counting wraparounds.

rtp.timestamp

When the audio in this packet was sampled, in clock ticks (8000 per second for G.711).

Why it matters: Comparing timestamps with arrival times gives jitter: how unevenly packets arrive.

rtp.timestamp_ext

The timestamp extended past its 32-bit limit.

rtp.ssrc

The synchronization source: a random ID for this audio stream.

Why it matters: Each direction of a call has its own SSRC, so it identifies one stream.

rtp.payload

The encoded audio itself. 160 bytes is 20 ms of G.711.