How Authoring Tools Interleave Audio and Video in a VOB

DVD authoring tools multiplex separate video, audio, and subpicture assets into a single Video Object (VOB) file by interleaving them at the sector level. Because a VOB is an implementation of the MPEG-2 Program Stream, an authoring tool must package each elementary stream into standardized packets and time-align them according to strict buffer constraints. By calculating decode and presentation timestamps against a shared reference clock, the multiplexer schedules video frames, audio samples, and bitmap subtitles into discrete Video Object Units (VOBUs), guaranteeing synchronized, continuous playback without decoder buffer underflow or overflow.

The Foundation: 2048-Byte Sectors and PES Packets

A VOB file is physically arranged into fixed-size sectors of 2,048 bytes (2 KB), matching the native sector size of DVD media. Elementary streams—MPEG-2 video, AC-3 or LPCM audio, and run-length encoded subpicture graphics—are first segmented into Packetized Elementary Stream (PES) packets.

The authoring tool places each PES packet inside a 2048-byte structure known as a Pack. Each Pack begins with a 14-byte Pack Header containing a System Clock Reference (SCR), followed by a PES header, stream-specific private headers, and the raw payload data. Because each Pack generally carries only one type of media, interleaving occurs by alternating these 2 KB sectors on the disc.

Time Stamping and the System Target Decoder (STD)

To determine the exact order of interleaved packets, the authoring tool runs a simulation of the MPEG System Target Decoder (STD). The DVD player contains distinct physical buffers for video, audio, and subpicture streams. The authoring multiplexer must ensure that none of these buffers run dry (underflow) or spill over (overflow).

  1. System Clock Reference (SCR): Defines the exact arrival time of a sector into the player's demultiplexer.
  2. Decode Time Stamp (DTS): Directs the decoder when to remove and decode an access unit (such as a compressed B-frame or P-frame).
  3. Presentation Time Stamp (PTS): Directs the hardware when to display the reconstructed frame, play the audio frame, or show the subtitle bitmap.

Because processing requires time, packets must arrive in the player’s buffers before their PTS. Audio and subpicture data are therefore written to the disc slightly ahead of the video frame they coincide with chronologically.

The Video Object Unit (VOBU)

Interleaving is organized into blocks called Video Object Units (VOBUs). A VOBU typically represents 0.4 to 1.0 seconds of playback and corresponds to an MPEG-2 Group of Pictures (GOP).

Every VOBU strictly adheres to a standard layout:

The Multiplexing Algorithm

When writing a VOBU, the authoring tool’s multiplexer uses a priority scheduler to interleave the sectors:

  1. Write the Navigation Pack: The NV_PCK is placed at the start of the address block.
  2. Allocate Constant Audio: Uncompressed or compressed audio must maintain an unbroken acoustic timeline. Because audio streams have low, constant bitrates relative to video, the authoring tool schedules audio packs at regular, predictable intervals throughout the VOBU to prevent audio buffer starvation.
  3. Insert Bursty Subpictures: Subpictures are asynchronous. A subtitle bitmap or menu highlight appears all at once, requiring a burst of sectors. The tool divides the subpicture data into SP_PCKs and distributes them across the sectors preceding the subtitle's starting PTS, ensuring the player has fully loaded the graphic into memory before rendering.
  4. Fill with Video: Video consumes the majority of the DVD’s 10.08 Mbps maximum multiplex rate. The authoring tool fills all remaining sector slots with video packs. If variable bitrate (VBR) video dips and the total combined bitrate falls below the target sector rate, the authoring tool inserts Padding Packs to maintain a consistent bitstream timing.

By strictly sequencing packets based on decoder arrival deadlines rather than simultaneous display times, the authoring tool ensures that the optical laser reads all disparate streams in a single, continuous track without seeking.