The MPEG Transition
Progress, for the evolving MPEG-4 encoder marketplace, is all about how low,
in bit rate, one can go, while still making pictures pretty enough for the big screen.
Up until recently, the topic of advanced video compression entered bandwidth conversations as one of the “tools in the toolbox,” alongside switched video, node splitting, and analog spectrum reclamation. Someday, the reasoning went, advanced compression will become a significant way to save shelf space, particularly for bulky content – like that swiftly rising number of HDTV channels.
The promise of MPEG-4 and its many techno-linguistic variations (H.264, AVC, VC-1) is to achieve equivalent picture quality as something compressed with MPEG-2, at half the bit rate.
As “tools in the bandwidth toolkit” go, though, MPEG-4 has long been the apparatus that wasn’t quite ready for primetime, or that didn’t quite fit the tool belt.
That all changed during a breakfast panel at this year’s Society of Cable Telecommunications Engineers annual Cable-Tec Expo, held June 20-22 in Orlando, Fla. As early-bird attendees tucked into coffee and danish, HBO Chief Technology Officer Bob Zitter calmly dished up the network’s HDTV plan:
To exclusively compress its newly-announced, 26-channel lineup using MPEG-4, at an encoding rate of 8 Mbps (7 Mbps payload; 1 Mbps overhead).
The MPEG-4 part wasn’t a big surprise. HDTV signals are as pricey to transmit as they are beautiful to behold. A little extra squish on the front-end literally saves millions in satellite delivery costs.
The encoding rate was interesting, but not earth-shattering. At 8 Mbps, HBO, a premium content provider, is taking a predictably conservative stance. Its moves mirror its “quality-first” stance 15 years ago, when MPEG-2 was the state of the art. (AT&T, by contrast, wants its encoder suppliers as close to 6 Mbps as possible.)
No, the surprise was the exclusive piece. As in, nothing sent in MPEG-2, which is what practically all digital cable (and satellite, for that matter) boxes for sure know how to decode.
A tally of network program announcements show that 66 HDTV channels will be
available for transmission by the end of this year. Next year, another 18.
“In many ways, we’re at the same spot now as we were at the onset of MPEG-2,” Zitter explained, because, in a way, MPEG-4 compression is to high-bandwidth HDTV what MPEG-2 compression was to analog TV. “We have a new technology, that’s going to give us a 50 percent improvement in bandwidth, that we think is ready.”
Here’s how it looks, from the perspective of a program network: For HBO to transmit 26 HDTV channels by satellite, using conventional, MPEG-2 compression, it would need an extra seven or eight satellite transponders – double what it currently uses. (Carriage on a satellite transponder costs networks around $125,000 per month; orbital space for the C-band satellites used to distribute programming across the U.S. is “totally full,” Zitter noted.)
The cable technologists with Zitter on the panel were visibly surprised. (Again, it was not yet 7:30 in the morning.) On the index of facial expressions, the CTOs – Marwan Fawaz, CTO of Charter Communications, John Schanz, EVP/national engineering and technical operations of Comcast, and Jim Ludington, SVP of program management and divisional support for Time Warner Cable – registered somewhere between “did he say exclusively?” and “Uh-oh. This’ll cost me.”
All reacted similarly, saying that if it was up to them, they’d much rather a simulcast of both MPEG-2 and MPEG-4 content.
Zitter reiterated that HBO’s modus operandi, with MPEG-4 now as with MPEG-2 15 years ago, is to proceed only after extensive discussions with its distributors. In the case of cable, he said, the decision reflected clear feedback that operators planned to move to hybrid MPEG 2/4 chips as early as mid-’07 (coincident with the July ’07 removable security mandate). At worst, boxes containing dual-mode chips should start entering the marketplace by early ’08.
Obviously, it’ll take a while for the new, dual-mode boxes to represent even a fraction of the deployed base of MPEG-2-only units. But, a start is a start.
The transcoder market begins
HBO’s moves mean operators will need to step up if they want to carry the network’s 26 HDTV channels. Specifically, they’ll need to step up to their headends. Because the 35 million digital cable boxes in U.S. homes right now don’t know an incoming MPEG-4 signal from a left-handed monkey wrench, something needs to know how to make the conversion back to MPEG-2.
HBO CTO Bob Zitter
That something is a “transcoder” – which, right off the bat, exhibits an unintended case of descriptive whimsy. Because MPEG-2 and MPEG-4 are so markedly different, equipment manufacturers submit, the mechanisms don’t yet affordably exist to simultaneously go from one to the other, on the fly. A transcoder, in today’s realities, doesn’t yet have any “trans.” It’s a decoder, followed by an encoder.
And it’s a crowded equipment marketplace. Consider: Over time, MPEG-4 encoding potentially impacts all HD channels, at all earth stations, in all headends. No wonder there’s no shortage of companies eyeing the advanced compression scene.
At the National Association of Broadcasters convention in Las Vegas in April, the show floor teemed with vendors: Envivio, Harmonic, Modulus (since purchased by Motorola), Scientific Altanta/Cisco, Tandberg TV, and the Grass Valley unit of Thomson, among others, showed up in force to show progress.
Progress, for the evolving MPEG-4 encoder marketplace, is all about how low, in bit rate, one can go, while still making pictures pretty enough for the big screen. One vendor (Grass Valley) boldly proclaimed a 4 Mbps, broadcast-quality encoding rate for HD, not including audio and overhead (which adds 1 Mbps).
Others emphasized their prowess in any of the many additional tools intrinsic to MPEG-4. Generally speaking, MPEG-2 and MPEG-4 are the same only in the sense that both specs tend to yield a device that consists of a motion predictor, and an encoding core.
But if MPEG-2 compression is all about removing unnecessarily duplicative information on a frame-by-frame basis, then MPEG-4 both smartens and sharpens the cutting tool. It adds more precision into motion estimation (quarter-pixel for MPEG-4, vs. half-pixel for MPEG-2), for example, and more flexibility into the size of the cut (by supporting multiple macro-block sizes). Once the duplicative material is removed, the mathematical techniques used to send the remainder of the picture onward are more advanced, MPEG-4 people say.
In an applied sense, an MPEG-4 to MPEG-2 transcoder works by first reverting (unsquishing) the incoming, MPEG-4 compressed signal to digital baseband, then re-encoding (re-squishing) that stream into MPEG-2. (This is already happening in the inverse direction – MPEG-2 to MPEG-4 – for DSL-based video providers, like AT&T, which need to be extraordinarily vigilant about bit rates.)
For transcoder manufacturers, the secret sauce is in the re-encoding. A big design goal is to minimize or eliminate “generation loss,” which is the degradation of a picture, caused by re-processing.
Even a beautifully encoded MPEG-2 picture, when decoded, contains errors. Those errors probably aren’t visible to us humans. On the control room monitor, that decoded, baseband digital signal may look gorgeous. But to the next encoder, those errors are as noticeable as a bumble bee in a bowl of clam chowder. They look and act like noise.
The trick is to hide the baked-in errors, without degrading picture quality; then, send the reconstituted image on its way.
In the weeks following HBO’s MPEG-4 news, the buzz meter on transcoder costs redlined. Estimates still vary widely – from a few hundred dollars per stream, to $12,000 for a three-stream transcoder. Several MSOs said they’re hoping to settle in at $2,500 for a transcoder that can handle three HD streams.
HBO’s MPEG-4 move is notable for one big reason, and important for two big reasons.
It’s notable because it unintentionally favors one type of network distribution. Hint: Not the type used by the home team. AT&T already uses boxes outfitted with MPEG-4 decoder chips – albeit a miniscule fraction of the 40 million cable and satellite customers HBO serves. In theory, that means AT&T can ditch its MPEG-2 to MPEG-4 transcoders, over time and as HD ramps up.
It’s important because it’s probably a portent of where things are going, on a macro, programmer scale. When it comes to new stuff, and especially new high-tech stuff, HBO tends to lead the program network pack. For that reason, it’s plausible to assume that other networks, facing the same distribution cost challenges, will follow suit with MPEG-4-only satellite distribution.
All in all, HBO’s “MPEG-4 exclusively” decision puts a definitive milestone into the developing story that is advanced compression. In a word, the milestone says “go.” By the end of this year, the network plans to have eight or nine feeds up, with the remainder arriving in ’08.
At press time, HBO hadn’t yet announced its choice of encoder supplier. Zitter said he’s evaluating candidates along a systematic prism that includes everything it and its distributors need to get going, without integration asphyxia. That means real-time stream encoding, packaged with security, encryption, multiplexing, upconversion – and the inverse, at the receive end.
“What it means is if you want to carry HBO signature hi-def, you’ll need an MPEG-4 IRD (integrated receiver/decoder),” Zitter said. The IRD will come with an output, he said, that will either be MPEG-4, or transcoded MPEG-2.
“We’re requiring that our vendor make IRDs available for our affiliates to buy,” Zitter explained. “They’re not a replacement, because they (MSOs) don’t have anything right now to receive these new feeds. They’ll have to buy something.” That’d be true whether it was MPEG-4 “or MPEG-17,” he added.
If nothing else, it’s a good time to be an encoder manufacturer. Or anyone without double-digit millions of digital boxes out in the field, that can’t speak MPEG-4.