When we talk about embedded audio in a SDI video signal, we refer to the digital audio that is included in the frame of the SDI digital video signal. That is, through a single cable or fiber, a video signal plus the audio signal associated with that video is transmitted.
SMPTE 299M-2004 standard describes how a digital audio signal of 24 bits and its control information are included in the ancillary data space of a video frame according to the standard SMPTE 292M.
AES audio data may contain PCM audio data or non- PCM data formatted according to SMPTE 337.
They are organized into groups of 4 channels of digital audio and it can be transmitted up to 4 groups, so then a maximum of 16 channels (8 pairs AES) for sampling rates of 32KHz, 44.1 KHz or 48 KHz. For sample rate of 96 kHz can be transmitted up to 8 channels. The characteristics of the digital audio of all channels must be equal (sampling frequency, sampling phase and timing).
Each audio group has a unique identifier. Audio data packets are embedded in the horizontal ancillary space of Cb / Cr (chrominance) data stream.
Control packets of audio (which give information frame number, sampling frequency, the delay of processing audio, if the channel is active or inactive …) are embedded in the horizontal ancillary space of Y (luminance) data stream.
It may be that in a particular application needed to extract audio from the SDI signal and transmit it separately from the video. This is possible thanks to the audio de-embedder devices.