Audio standards are delaying Digital Video.................................
techweb.cmp.com
Chip makers seek help for MPEG-2 Advanced Audio Coding
By Junko Yoshida
SANTA CLARA, Calif. -- With the Japanese government's endorsement last month of the newly standardized MPEG-2 Advanced Audio Coding (AAC) as the audio coding algorithm for digital TV in Japan, IC vendors are beating the bushes for information on AAC and are seeking partners who will give them access to AAC intellectual property. To date, no one has launched a chip capable of processing AAC, and there is no one-stop shop for licensing the IP.
Now an ISO standard, MPEG-2 AAC is a perceptual audio-coding algorithm that is not backward-compatible with MPEG-1 or MPEG-2 audio. It provides high-quality sound at a rate of 64 kbits/second per channel for multichannel operation.
AAC arrives at a time when a multiplicity of audio-coding algorithms is fragmenting the digital consumer market, bombarding system and chip vendors with changing digital-audio requirements. "A product with multiple [audio] standards is not tomorrow's issue," said Paul Goldberg, vice president of audio products and intellectual properties at chip maker Zoran Corp. (Santa Clara, Calif.). "It's very much today's issue."
AAC applications are not limited to the satellite-based Japanese digital-TV service slated for rollout in 2000; they extend to such new services as delivery of downloadable CD-quality music over the Internet, satellite or cable.
AT&T, for instance, has begun a trial service called "a2b music" over the Internet in collaboration with record companies. Some in the industry even speculate that MPEG-2 AAC may become an integral element for CD-recordable systems, or an additional DVD audio feature when DVD becomes recordable.
IC vendors are examining various digital-audio algorithms. Besides linear pulse-code modulation (PCM), MPEG-1 and 2 and Dolby Digital--the mainstays of today's DVD players--at least a half a dozen more have been proposed for DVD, home-theater and Internet-audio applications, some by such heavyweights as Bell Labs, Sony and Philips.
For chip vendors seeking design wins in DVD systems 18 months from now, the worst-case scenario is a chip that's expected to support not only Dolby Digital, Linear PCM and MPEG-1 and 2, but also AAC and any (or all) of the other audio options. The challenge is to guess which audio-coding algorithms are worth supporting, at what cost and under what kind of silicon architecture.
"Most people in the industry have underestimated the complexity of meeting different digital-audio requirements," said Paul Bundschuh, applications and strategic marketing manager for digital-audio operations at Motorola Inc. While a couple of dozen solutions already exist to decode MPEG-2 video, few audio ICs are capable of decoding multiple audio-coding algorithms.
The MPEG Committee's Audio Subgroup reported just last month that MPEG-2 AAC, tested under the stringent requirements of the ITU-R test methodology, "demonstrated full broadcast-quality audio at 128 kbits/s for stereo, approximately half the bit rate of that needed by the earlier MPEG-1 Layer II codec." The Layer II audio codec is currently used in U.S. and European digital-satellite TV services.
The purpose of the MPEG-2 AAC effort was to achieve an audio coding solution that would be "much more powerful than MPEG-1" and "designed for efficient bandwidth-compression applications," said Peter Kroon, supervisor of the audio-communication group at Bell Labs.
Patent suitors
Chip companies are beginning in earnest to forge partnerships with the essential MPEG-2 AAC algorithm patent holders. Describing MPEG-2 AAC as "a priority for Zoran today," Goldberg said Zoran has been "talking to Frauenhofer Institut [Erlangen, Germany] for the last six months" to work out necessary steps to port MPEG-2 AAC to Zoran's DSPs. The German research institute is one of several companies that claim to have essential IPs for MPEG-2 AAC. Others include Bell Labs, Sony Corp. and Dolby Laboratories (San Francisco).
Of course, IC vendors need to soberly assess exactly how big the MPEG-2 AAC market is going to be and when its growth might reach critical mass. U.S. vendors, however, can't afford to spend too much time fence-sitting. Japanese system vendors have been aggressively developing proprietary consumer digital ICs through their in-house semiconductor arms.
Although the Japanese DTV system remains the current focus of MPEG-2 AAC development, Motorola's Bundschuh regards AAC as a new force that may open the door for broader worldwide applications when MPEG-4 comes online. AAC is defined as a part of the emerging MPEG-4 multimedia representation standard.
Whether MPEG-2 AAC becomes an important element for the recordable DVD standard remains an open question. Considering the heavy Japanese interest in AAC, especially by Sony, some see a possible link.
To complicate matters further, a Bell Labs concept called Perceptual Audio Coding (PAC), which has been proposed for U.S. digital radio and multicasting on the Internet, could compete with AAC for some U.S. applications. Despite a history of close involvement with AAC development, Bell Labs today is promoting its proprietary PAC, claiming that PAC's low-bit-rate, high-quality audio coding laid the basis for AAC. According to Kroon, PAC's coding has substantially evolved, both computationally and in terms of quality, since 1996, when AAC was frozen as a standard.
PAC supports bit rates as low as 8 kbits/s and is commercially available through Elemedia, Lucent Technologies' internal venture. Unlike MPEG-2 AAC, PAC comes with such application-specific features as digital audio broadcasting, resistance against communication errors and distortion, and robust error recovery. PAC at 128 kbits/s requires processing power equivalent to 10 to 15 Mips for decode; encoding requires 40 to 50 Mips. PAC at 16 kbits/s needs only 3 to 4 Mips for decoding and 20 Mips for encoding, according to Deepen Sinha, research staff member at the audio communication group of Bell Labs.
Surrounding issues
If AAC is still a ways off, a more pressing matter for most chip vendors today is how to respond to Japanese consumer-electronics manufacturers' new demand that they integrate a DTS Digital Surround stream output feature into a DVD chip set. Developed by Digital Theater Systems Inc. (Westlake Village, Calif.), DTS Digital Surround is an encode/ decode system that delivers 5.1 channels of master-quality, 20-bit audio. It is derived from the surround-sound technologies the company developed for motion pictures and movie theaters.
The DVD standard does not mandate DTS. But as Hollywood studios have released more DVD movie titles featuring DTS Digital Surround, "consumers' awareness has been going up," said Darren Neuman, director of DVD engineering for LSI Logic Corp. (Milpitas, Calif.). Although DVD players may not have a built-in DTS-decoding capability, manufacturers expect to give them at least an ability to output a DTS bit stream, with decoding in a separate audio/video receiver.
With developers uncertain about which digital-audio coding schemes to run with, many are embracing programmable approaches.
"You need a powerful enough architecture to handle a wide range of audio algorithms," said Zoran's Goldberg. When Goldberg worked for Dolby Labs several years ago, "my assumption was that nothing would stay the same as far as audio is concerned." Zoran's ZR38600, its latest programmable audio processor, is the only audio processor capable of 3-D sound/virtual surround processing simul-taneously with Dolby Digital decoding, he claimed.
The goal for Zoran is to continue to develop a DSP-based audio solution for multistandard products that detect a code in the header, download an appropriate decoder software and decode the stream accordingly, he said.
Bundschuh at Motorola agreed on the basic concept for "a software-configurable hardware architecture." Motorola's recently announced DSP 56362 is the only chip now certified to decode Dolby Digital, DTS and multichannel MPEG-2 audio algorithms. |
