The AVB standard requires audio data to be split into packets to be transmitted over ethernet along with meta-information specified by the IEEE 1722 transport protocol. This meta-information incurs an overhead on the bandwidth of each stream of data.

The protocol overhead is detailed in the following table:

Each stream of audio data can contain several multiplexed channels of audio. The higher the number of channels per stream, the more efficient the audio transport is in terms of bandwidth. Note that the IEC 61883-6 standard recommends transmitting single channel streams as stereo with the right channel blank.

After the header in each packet, audio data is stored in AM824 format which pads 24-bit data to 4 byte quadlets. The frame rate is 8kHz so a 48kHz stream has 6 samples per packet, a 96kHz has 12 samples per packet and so on.

The following table shows the bandwidth for streams at different sample rates with different number of channels per stream.

Note that the higher the number of channels per stream, the better the bandwidth usage (e.g. 4 x 8 channels streams uses less bandwidth than 16 x 2 channel streams). The IEEE 1722 standard also specifies that only 75% of available bandwidth can be used for AVB traffic.