Ingex Archive

A description of the BBC D3 preservation system can be found in the BBC Research white paper File-based Production: Making It Work In Practice (pages 8 to 11).

Archive File Format

The design of the file format to contain the BBC's entire D3 video archive had to fulfil the requirement of long term preservation. Once a programme is transferred from D3 video tape to a file, the expected lifetime of the file is several decades or longer. Although the file storage medium is likely to change, the file itself would not change as it is moved from a legacy file medium to the next file medium.

To help ensure video files would be readable in the distant future, open standards were chosen at all levels of the design. MXF OP-1A was chosen as the file container to store the video, audio and timecode data which would reproduce as closely as possible the information on the D3 video tape. Holding the entire archive of MXF files permanently on hard disks was too expensive so the open standard LTO-3 data tape was chosen as the initial file storage medium.

The OP-1A MXF file contained a sequence of content packages, one for each frame, containing video, audio and timecode, in the following formats:

• A video item of SMPTE 384M uncompressed 4:2:2 video at 8 bits per sample in UYVY format
• An audio item with 4 tracks of SMPTE 382M uncompressed PCM audio at 48kHz and 20 bits per sample
• A system item containing an array of SMPTE 12M timecodes representing the VITC and LTC timecodes read off the video tape

In addition to the essence, the OP-1A MXF file contained the following metadata in the MXF Header:

• Data items from the BBC Infax archive database record for the TV programme, identified by SMPTE Universal Labels registered under the BBC node of SMPTE metadata dictionary
• An array of D3 error records to facilitate future analysis, preservation and restoration
• An array of Photosensitive Epilepsy (PSE) failures if any
• The barcode of the LTO-3 data tape the MXF file is stored on

In designing the way MXF video files were to be stored on data tape, consideration was given to the way archived tapes were to be used. LTO-3 tapes can store 400GB of uncompressed data, corresponding to about 10 uncompressed archive programmes. They are stored on shelves in an archive vault initially, and intended to be retrieved by human operator rather than a robot. A standalone Quality Check (QC) station is used to check the contents of the LTO-3 tapes after transfer, and to make copies of programmes as either video tape copies or file copies. The format of files on the data tape had the following attributes:

• POSIX.1-2001 archive format (also known as pax Interchange format), a superset of the tar format which overcomes the 8 GiB limitation of tar format, to ensure future access to programmes
• No spanning of tapes, only complete MXF files were stored on any single tape
• No tape level compression, to mitigate data loss in the rare event of media errors
• File marks between programmes, to aid identification and improve seeking
• Tape block size of 256 KiB, to optimise LTO-3 data transfer speeds
• An ASCII text index file at the beginning of tape listing each MXF file with its file size and a detailed text description of the tape and MXF structure. A hard disk cache was required to support the construction of the index file.

Example D3 preservation MXF file

• LTA00000101.mxf (21MB).

This example MXF file contains 1 second of material ingested from a D3 tape and is the first file stored on an LTO tape identified with the barcode LTA000001. A fake VTR error event on the first frame and a PSE (Photosensitive epilepsy) failure event on the last frame illustrate how they are included in the file.

A flash movie file,

• LTA00000101.mxf.html

shows the metadata structure contained within the MXF file.