Of course the goal is to do all of our splicing, inserting and branding within the compressed domain.
The general subject of MPEG splicing has been done to death over the last few years. However, this is a subject that won't go away soon. As standards evolve and more practical experience is gained working with MPEG transmission of HDTV and multicast programming, systems and facilities will continue to be faced with new challenges.
A case in point is determining what role MPEG splicing will play in minimizing and avoiding decompression-recompression cycles (and the resultant picture degradation) in the process of manipulating video at the station level. Of course the goal is to do all of our splicing, inserting and branding within the compressed domain. And of course, we can only do that at I frames because all other frames within the GOP are either predictive or bi-directionally interpolated information that reconstruct video frames later on at the decoder. So, what's a master control operator to do? We'll take a look at some present-day solutions and some future considerations.
Today, most commercial stations receive HDTV signals as MPEG-2 bitstreams at 45Mb/s. They download the signal and encode it to baseband HDTV and SDTV. The incoming 45Mb/s bitstream provides enough information for the station's decoder to produce a robust HDTV signal that stands up to the MPEG-to-baseband conversion. This signal is usually handed off to the local routing matrix for further manipulation and on-air control. The baseband video is used for operations that require manipulation on the pixel level, including logo insertion, ad insertion and any editing that takes place. The HDTV signal is then encoded to a 19.4Mb/s MPEG-2 bitstream for transmission.
While the development of MPEG standards and tools for MPEG splicing is desirable, so far baseband conversion has been the standard practice in today's commercial environments because the results are within an acceptable range, at least for now. However, this is most likely a situation that will eventually demand improvement. I firmly believe HD viewers will be here soon, and advertisers will follow.
Think about the energy and talent that high end productions throw at compression and then consider the weakest link in the image path up to the transmitter. While the encoders that are available do a fairly good job, they have the burden of being a one-size-fits-all black box. There are real and perceived differences between the quality of compression that can be obtained by a trained compressionist eyeballing every scene in an expensive compression suite and an automated one-size-fits-all black box solution at the master control switcher. Ad agencies are likely to want to control the compression quality of their commercials by distributing 19.4Mb/s precompressed MPEG video to broadcasters. They might have the expectation or even the requirement to avoid degradation at the station level through the introduction of further decompression-recompression cycles.
One approach to this problem is to splice the pass-through MPEG-2 transport stream downstream from the master control switcher. In this configuration the output of the master control switcher is routed to an encoder. The resultant bitstream is then spliced into the pass-through program feed in a downstream splicer before routing to the transmitter. This approach maintains the integrity of the pass-through feed; however, it can only be achieved at I frames. The spot can be transferred to a local file server as a pre-encoded 19.4 MPEG-2 file. Downstream of your master control switcher the I frames are lined up at predetermined splice points and “spliced” into the on-air bitstream. Now, as an advertiser, I'm fairly confident that the quality is going to be consistent regardless of the transmit path.
The PBS stations are in a different situation, and one that creates a greater immediate challenge. PBS stations are receiving network feeds that are prepackaged at 19.4Mb/s. There are good reasons for this practice. The most pressing is that many PBS stations do not have the budgets to build an extensive internal DTV infrastructure. If they are fed MPEG streams that are already optimized for broadcast, all they need to do is pass the signal through to the transmitter, a much less costly approach.
However, there is still a need to manipulate the program for local branding and other functions. The problem is that the incoming 19.4Mb/s signal does not stand up to the process of decoding to baseband, processing and re-encoding to a new 19.4Mb/s MPEG bitstream anywhere near as well as the incoming 45Mb/s stream in the previous example.
Splicing downstream from the master control switcher is a solution to this problem. This effectively becomes a master encoding step that combines elements created at the station level and output by the master control switcher with elements being fed from other sources and passed through.
Ultimately, the desired requirements for the downstream splicer are that it be visually and syntactically seamless, maintaining the integrity of both the visual continuity and the structure of the bitstream.
There are several manufacturers of MPEG splicers, and products are available that allow you to splice into the transport stream in a number of different ways. Some methods involve the use of markers in metadata that indicate safe splice points in transport streams. One method is to drop to black for a few seconds or freeze a frame of video and then clear the memory buffer and start a new group of pictures with a new I frame. Another manufacturer uses proprietary algorithms to analyze bitstreams on the fly and determine safe splice points in real time. Another method drops to baseband, logs the incoming GOP/macroblock structure, performs operations in baseband, and reapplies the original GOP structure. There are a number of different ways to do it, each with its own pros and cons.
For now, making decisions with regard to MPEG splicing is a matter of careful consideration of both the available technologies and the specific requirements of a given application. With this in mind, it is reasonable to expect significant advances in the design and implementation of MPEG splicing technology as managing DTV signal paths becomes increasingly central to station operations.