MPEG-2 decoders......................................
tvbroadcast.com
Nonconformists
Most MPEG-2 Decoders Have Unique Personalities
By Craig Birkmaier
There's something very comforting about good old composite video. Whetheryou are a professional using coax cables with BNC connectors, or a consumerusing cables with RCA connectors, the chances are pretty good that you canconnect two devices that speak the language of composite video and theywill make pictures.
But this new digital video, based on the international MPEG-2 standard, isdifferent. Most of the products that say they have "MPEG-2 Inside" don'tspeak to one another. Most are boxes you put on top of an analog TV set-forDBS, DVD, DVB, ATSC, or Digital Cable. They still speak analog to theoutside world-composite, S-video, or component.
Some of these boxes have proprietary digital ports to connect to DTV-ready displays with the same proprietary ports. A few are starting to ship with "FireWire," the IEEE-1394 serial interconnect. Digital interconnections have been slow to reach the market for several reasons:
1. Hollywood and the rest of the content-producing world is scared to death what will happen if people can start exchanging MPEG-2 files like they're doing with MP3 files today.
2. If you connect to another product with an MPEG-2 decoder, that decoder must understand how to decode your bitstream.
Things are still a mess on the copy protection front, but some progress has been made. A spec has been created to protect the bitstreams carried via IEEE-1394, and at least a few companies are anxious to start using it.
Although Hollywood is uncomfortable with the concept of "digital cloning," they are beginning to understand that they can make more money faster by learning how to live with the digital monster. The end-game is to release content globally, get your money quickly, and thus eliminate the fertile ground where pirates operate. Cloning can be an advantage if the consumer can buy a relatively inexpensive key to unlock the copy... smart cards and smart networks can make this about as painless as making a long distance telephone call today.
It's the second problem that really has me worried. There's more than ample reason for concern that MPEG-2 bit stream interchange is a looming nightmare. All of the early warning signs are there. Here's the problem...
Learning How To Use The MPEG Toolbox
When I said MPEG-2 is different, I meant really different! Unlike the analog standards we have all grown to understand and love, MPEG-2 is not really a standard... it's a toolbox with which we can build all kinds of cool products. If you read the MPEG-2 standard you will not learn much about how to build a conformable encoder. That's because the process of encoding is not standardized. Conformance is established via the MPEG-2 decoder. If your bitstream conforms to the specifications for a specific MPEG-2 Profile and Level, a decoder conformable to that Profile and Level should be able to produce a useable picture (and audio channels if the MPEG audio tools are used).
And if you really want to follow the letter-of-the-law of MPEG-2 conformance, a decoder that supports a superset of the tools required for a specific Profile and Level should decode bit streams for the superset and all conformable subsets.
MPEG-2 Profiles define a subset of tools that can be used to code video streams over a broad range of spatial and temporal resolutions; however, syntactical restrictions may be imposed. MPEG Levels define specific conformance points in terms of the maximum picture size and frame rates that a decoder is expected to handle. The levels do not limit the possible subsets of frame sizes, aspect ratios, or frame rates, but profiles may. If this sounds all too confusing, perhaps an example will help.
Main Profile at Main Level (MP@ML) can be used to deliver what is called SDTV today. The limits imposed are: up to 720 samples per line; up to 576 lines; 30 frames per second; and a maximum pixel clock for the luminance component of the signal of about 10,368,000 pixels/sec.
Main Profile at High Level (MP@HL) can be used to deliver what is called HDTV today, and a bunch of stuff that lies between SDTV and HDTV, like 480p and 576p. The limits imposed are: up to 1920 samples per line; up to 1088 lines; 60 frames per second; and a maximum pixel clock for the luminance component of the signal of 62,668,800 pixels/sec.
Main Profile has syntactic restrictions that limit: the available frame rates to 23.97, 24, 25, 29.97, 30, 59.94, and 60 frame per second; aspect ratio to 4:3 and 16:9; color resolution to 4:2:0; and other factors including system colorimetry.
To be conformable with MP@ML, a decoder should be capable of handling any line length up to 720 samples per line. To be conformable to the hierarchy of Profiles and Levels established by the MPEG-2 standard, an MP@HL decoder should be able to decode an MP@ML bitstream with 720 samples per line.
Unfortunately, recent tests of several of the first generation ATSC receiver/display systems revealed that a couple could not make pictures with source material encoded with 720 samples per line. What went wrong?
Depending on who you talk to... a lot, or nothing.
The problem is quite simple. MPEG allows other standards bodies or industry trade groups to create standards that place further restrictions on the existing MPEG-2 Profile and Level structure.
Table 3 from ATSC Standard A-53 is a prime example. This table places further restrictions on the formats that a decoder is expected to work with. If a company chooses to implement their decoder to the syntax of the ATSC standard, it may not work with bit streams that are conformable with the full MPEG-2 Profile and Level that is being used. In the example above,several companies build decoders that handle the ATSC limit of 704 samples per line properly, but "broke" when presented with a bitstream with 720 samples per line.
This problem is not unique to the ATSC standard. What is unique are products that are fully conformable to the MPEG-2 Profiles and Levels. In other words, while all of these systems are leveraging the "good name of MPEG-2," they may all be incompatible with one another. At the very least, each must be dealt with as a unique standard. Does this really matter?
Not if you are willing to have a box by every TV display for every possible program service that uses MPEG-2. Who cares if there are a few more remotes on the coffee table, and that we are duplicating many of the same functions inside each box? My guess is that the consumer electronics industry likes it that way? it's as comfortable as those composite video cables.
No Future?
If you thought I was finished, sorry, there's still more bad news. The folks who put the MPEG standard together understood that the specifications might need to be updated from time to time to reflect marketplace realities. So they built a bunch of hooks into the specs to allow the standard to evolve. Scattered throughout the specifications are data fields with reserved bits to accommodate future extensions.
Decoders are not expected to be clairvoyant... i.e. they can't be expected to deal with stuff that hasn't been invented yet. But a decoder that is conformable with the Profile and Level should not break when presented with bits in a reserved extension that it does not understand; the decoder is expected to ignore these bits and keep on working. Alas, there may already be millions of decoders deployed that are non-conformant with respect to "ignoring" reserved extensions to the standard.
Already I have heard manufacturers talk about the need to freeze the MPEG specifications because any changes could cause the products they have deployed to break. Bad idea!
The problems I've just identified are not going away. In fact, they are about to move to center stage, thanks to emerging technologies that are going to put MPEG-2 products to the test. Those technologies are home networking and Personal Video Recorders (including in-home media servers). At the recent Western Cable Show home networking and home gateway products were all the rage, as was the concept of the Personal Video Recorder and/or home media server. A wide range of wired and wireless networking technologies were demonstrated, some of which use existing telephone and power wiring. Philips demonstrated a home gateway with: multiple tuners for each service the home is subscribed to (cable, DBS, DTV, NTSC, etc.); connections for telephone networks and/or ADSL; wired or wireless IP networking; and enough storage to serve multiple programs to networked receivers spread around the home.
Something tells me that those receivers will need MPEG-2 decoders that are conformable to more than one of the many unique flavors of MPEG-2 that exist today. Here's what I think all of this means. You know that old computer you have stored in the closet? The one that became obsolete when you needed a bigger, faster, better PC to run the latest software? Hope there's still room in the closet for some of these early MPEG-2 products. Fortunately, there is a silver lining in the clouds that cover our emerging digital world. We are learning that digital means you can be a nonconformist and still get along with your neighbors. It's that interoperability thing that CAN happen when we are dealing with bits instead of formats.
Hopefully, manufacturers are learning that it may make more sense to conform to the superset of capabilities defined by MPEG-2 Profiles and Levels, than the subsets used by various standards that a decoder may be expected to work with. Hopefully they are learning that conformance testing means dealing with everything that the MPEG-2 standard allows? today and in the future.
And hopefully, the recognition that we have a big problem here may finally force everyone to sit down together and figure out how to use our new digital tools to build a digital infrastructure we can all live with. |
