One high-profile sign of the times is the DVB-X initiative. The Digital Video Broadcast Project, a Europe-based industry group, recently disclosed this ambitious project, designed to marry terrestrial digital TV broadcasting technology with a cellular phone network infrastructure. Intended as "a mobility-improved transmission standard for terrestrial digital TV," DVB-X will become "a mandatory spec for handheld devices in the future," said Goeran Wahlberg, Nokia Corp.'s director for concepts and technology.

The DVB group doesn't expect DVB-X to reach the market until 2006. But its appearance underscores the fact that video capabilities are on the road map for everyone from service operators and handset companies to chip vendors. When it comes to actual implementations, however, "video on a handset leads to multiple challenges," observed Randall Fahey, senior consultant with the Forward Concepts research firm.

Beyond the need for high processing power and the impact on power consumption and battery life comes the issue of memory for storage. "What does one do with the resulting video file?" Fahey asked. "Is it being streamed, or downloaded and played back later?"

Next comes the question of whether the handset has enough program memory to contain the decoder/ encoder software. A further complication is the sticky issue of digital-rights management. Fahey warned, "If the handset is capable of storing a downloaded video clip for later playback, how is this handled? Can it be viewed multiple times? Can it be copied?"

All of this adds up to "increased architectural complexity in the system," said Manish Singh, vice president of marketing at chip maker MediaQ. From a welter of existing solutions-ranging from application processors to dedicated DSPs for multimedia acceleration and dedicated hardware acceleration-based media processors-Singh said that "the answers lie in the analysis of the trade-offs: flexibility, energy consumption and system implementation costs."

Even the raft of seemingly similar-looking application processors, many based on the ARM architecture, handle multimedia in very different ways. Clearly, the mobile industry is far from agreement, and the role of multimedia processors or coprocessors in mobile handsets is still being defined. The choices basically boil down to hardware vs. software, with the hardware-heavy approach appearing to gain favor.

On the hardware end, Fahey described the most popular solution as "a RISC core alongside a DSP core," coupled with "hardware acceleration for specific video functions." But another camp, he said, starts with a RISC core, generally an ARM9, and then does "all of the video processing in hardware."

Loosely coupling the hardware accelerator to the processor core can be a viable alternative-at least for companies with video expertise. "In MPEG-4, one of the more processing-intensive functions is motion estimation," said Fahey. "Lots of time and effort have been spent coming up with efficient ways to speed this up" via hardware acceleration.

The fact that the phone industry is "hell-bent on standards" could help drive the hardware-based approach further, said Kathleen Maher, an analyst at Jon Peddie Research, adding, "It looks like MPEG-4 has got the straight path to handhelds." And given the amount of processing power required, "I believe it's going to have to be hardwired in." Maher said that MediaQ, Epson and possibly NeoMagic are eyeing "a standards-based hardwired codec approach, because they'll be able to offer a chip that saves on power and yet delivers the required performance."

Although the standard chosen for the Universal Mobile Telecommunications System is based on MPEG-4, providers in Japan and South Korea with video services today are deploying proprietary codecs. J-Phone is using Office Noa's Nancy codec for its Sha-mail service in Japan, and SK Telecom uses Thin Multimedia's proprietary codec for services in Korea. Such proprietary software-based codecs "require less bandwidth, less memory and less processing horsepower, and can be tweaked to trade off resolution against frame rate, depending on the video content," Fahey explained.

Most experts agree that initial applications will be downloading short video clips, such as news and entertainment. "Download and play is not really an issue," Fahey said. "Once you've waited for the download, and assuming one has sufficient handset memory to store the video in question, it can be played back at just about any frame rate or quality level-assuming the frame rate is scalable in the codec implementation."

Streaming is another question. Video services in Korea today are delivering between 1 and 3 frames per second to SQCIF/QCIF screens. This will increase to between 5 and 10 frames/s as network bandwidth rises and algorithms are enhanced or upgraded.

One application that seems to be crying for video is the camera phone. "Phones with some kind of camera can reach numbers in the range of 300 million by 2006," said Maher. While many believe only a fraction of such devices may be used for video, it's clear that the next step is turning camera phones into camcorder phones.