File format conversion has become increasingly important in today's file-based workflow environment. In the past, file format conversion was accomplished with dedicated services that provided their own software and/or hardware to accept the file transfer and convert it to what the station needed.
Today, a TV station may require many different types of file conversions to produce media for on-air playback, content for its Web site and files for archival storage. While the actual file type may be different, the requirements are the same: high-quality video and the smallest file size.
Whenever a video file is converted from one type to another, the issue of video quality must be addressed. Assuming there is no data loss, the parameters of how the video was encoded is the main factor in determining video quality and, thus, its ability to be converted to another file format. Starting with acquisition, files differ in both quality and size.
Much of the future media to be converted will come from analog media, either film or videotape, and the quality of that initial transfer will determine the subsequent quality of future file conversions.
Video file formats are called either “lossless” or “lossy” — any file that uses compression on the video is considered lossy, because it must throw away some picture information to perform the compression and, thus, reduces the file size. You always want to move from lossless toward lossy with as little degradation to the picture as possible.
Files and wrappers
When a video file is created, the raw essence file is made up of the actual data from the codec used to encode the audio and video signals. This could be an MPEG-2 file or one of the many DV formats, but when input to a server or other device, the software that accepts it needs to know certain parameters about the actual audio and video data to be able to play it back. This is where the container comes in, which holds the separate audio and video essence data together and provides the necessary data, including codec used, frame rate, data rate, number of channels, etc., needed to be able to play them back.
One function of the container is to be an interchange format. When this type of container is used with professional video systems, it is called a wrapper. Wrappers have been developed to make it easier to exchange video files and the information associated with them between various systems such as nonlinear editing systems, video servers and some digital VTRs. Wrappers are data files that contain within them the essence, or audio and video files, and the metadata, which is data about or related to the audio and video files. The need for wrappers came about due to the lack of information exchanged during file transfers from one storage system to another. Without wrappers and metadata, the receiving system lacks any information about the content of the files.
The best-known professional wrapper is Material eXchange Format (MXF), developed by SMPTE (377M-2004). MXF was designed for use within broadcast and post-production facilities, making it the first choice for many stations. Although MXF has been widely adopted, it has not been fully implemented in many cases, so its full capabilities have not been used by many broadcasters. (See Figure 1 on page 32.)
Other containers and wrappers include QT (QuickTime), AAF (Advanced Authoring Format), GXF (General eXchange Format, SMPTE 360M), OMF (Open Media Framework) and AVI (Audio Video Interleaved).
The two most common video files in professional use for SD include:
- D10 (SMPTE 356M): Also known as Sony's IMX, this is an MPEG-2 format using only I-frames and eight channels of AES audio.
- DV/DIF (Digital Video/Digital Interface Format): Also known as DV, it is used in AVI-DV, QT-DV, MXF GC DV-DIF.
To apply a new wrapper, the program strips off the original wrapper and creates the new wrapper around the essence data, with all the appropriate data and metadata included. To convert the actual essence file type, or codec, the original encoded data must be unwrapped and examined, and then an algorithm is applied to convert the data to the new codec. How well this is done will determine the quality of the conversion and the quality of the video.
As of today, there is no common file format for HD, because most facilities use the format created during acquisition. Conversion programs can either change the essence file type or the wrapper the essence data is contained within. The latter technique is much easier than the former.
Conversion of the essence data takes time, because it uses a lot of a computer's processor power to perform the complex equations needed to convert from one codec to another. One way to speed up the process is to use multiple computers to distribute the load. A cluster of servers is used to reduce the amount of time required to convert a file and/or to convert multiple files at once. Quicker-than-real-time conversion is possible when using dedicated servers in clusters.
A typical setup would have a hot folder, which resides on a server or a storage device attached to the server. Via the network, any video file that requires conversion would be transferred to this hot folder. When a file appears in the folder, it is read by the conversion software, and the process begins. If the file is corrupt and can't be read, it is transferred to another folder that contains problem files.
Once the conversions are complete, the multiple output files are transferred to their respective folders, which can be seen on the network and transferred to the appropriate devices. Creating multiple types of files at the same time increases efficiencies by delivering copies for the various workflows within the station, including proxies for viewing on the desktop, MXF files for distribution, QuickTime for the video servers and so on. (See Figure 2.)
An important point to keep in mind with any file conversion is to make sure the metadata makes it. Metadata consists of information such as time code, I/O markers, closed-captioning and other VBI data. Different wrappers can hold this data in different forms and will need to be converted during the conversion process as well.
Russell Brown is chief engineer at KMTP-TV in San Francisco and writer of Broadcast Engineering's “Transition to Digital” e-newsletter.