BINGO!!!!!! Single-chip MPEG-2 codec to emerge
techweb.cmp.com
By Junko Yoshida
MILPITAS, Calif. -- Promising to transfer the power of encoding from broadcasters and Hollywood studios to the lowly consumer, C-Cube Microsystems Inc. will unveil on Monday a newly designed single-chip MPEG-2 codec architecture based on a 32-bit Microsparc processor core licensed two years ago from Sun Microsystems Inc.
The advent of a cost-effective, single-chip MPEG-2 codec could herald an era in which silicon vendors play an increasingly significant role in enabling a new generation of digital-video consumer products, such as rewritable DVD-based video recorders and camcorders. The codec will become "a fundamental technology driver for every platform, including communication, consumer and PC applications," said Alex Balkanski, president and chief executive officer at C-Cube.
C-Cube is not alone as it gears up its engineering resources for this emerging market. LSI Logic Corp. (Milpitas), which withdrew from the MPEG-2 encoder business for professional broadcasting, is planning a comeback with a similar single-chip device for the consumer arena. The low-cost, low-power-consumption MPEG-2 encoder will be out in 1998, said Alain Bismuth, strategic marketing director for DVD consumer products at LSI Logic.
C-Cube's codec, which combines an encoder and decoder on a single chip, is built on a highly scalable and programmable architecture called DVx. It integrates the embedded Microsparc RISC CPU core, C-Cube's own DSP and programmable Motion Estimation co-processor, along with several on-chip interface blocks. Other home-grown technologies include a whole new instruction set for the Microsparc core that C-Cube says makes the DVx architecture truly video friendly.
Manufactured in a 0.35-micron process technology, the DVx silicon is sampling now. It is scheduled for volume shipment in the fourth quarter.
The Microsparc-based codec will serve as a basis for all C-Cube's upcoming encoder/decoder solutions for disparate digital-video applications, effectively replacing the company's own multichip encoding solutions based on the proprietary VideoRISC processors.
Anticipated application areas range from communication/broadcasting, where the codec might go into professional broadcast encoders and two-way video collaboration systems, to PCs and workstations, for capturing video, post-production, editing and authoring. In the consumer market, the chip might find its way into digital-video cameras and DVD-RAM-based home video recorders. "Those are a multibillion-dollar market opportunity that did not exist previously," Balkanski said, "due to the lack of cost-effective codecs."
Besides being smaller and less expensive, the new architecture is also more flexible and extensible than the VideoRISC architecture, C-Cube said. It is capable of encoding all the MPEG-2 profiles, according to Bob Saffari, senior marketing manager of C-Cube's encoder group. That includes 4:2:2 profile@main level, simple profile@main level and main profile@high level, besides the standard MPEG-2 main profile@main level. The 4:2:2 profile is essential for high-quality post-production applications, he said, while the simple profile allows low-latency encoding and decoding, particularly useful for video communications.
Today's DVx codec, packed with 5.5 million transistors, is as big as the Pentium II, according to Saffari. In the encode mode, power dissipation is between 3.5 and 4 W. By moving to 0.25-micron and, later, to 0.18-micron process technologies as they become available, "Our goal is to optimize the memory utilization and reduce the power consumption to 1 to 2 W as a codec," Saffari said. "For camcorder design wins, it's necessary to bring it down to 1.2 to 0.7 W."
C-Cube CEO Balkanski believes that "we are only a generation away" from making consumer applications possible. The company expects that by 1999, the single-chip codec will cost less than $50. Today's version runs $1,500, but that includes all the microcode needed to build a professional broadcast encoder system.
Dataquest Inc. is equally bullish on the emerging MPEG-2 codec market for end-user applications. The San Jose, Calif., market-research firm predicts the market will "very quickly grow to 9 million units by 2000 from the almost non-existing market today," according to Jonathan Cassell, the firm's industry analyst. Dataquest sees about 3.8 million units of MPEG-2 codec being integrated on the PC platform for DVD encode and decode, with another 3.5 million units for consumer video recorders based on optical rewritable media, such as DVD-RAM. Digital set-top boxes capable of two-way video communication will grow to 2 million units, Dataquest said.
Despite possible pitfalls--such as diverging rewritable DVD formats--chip vendors are counting on the digital consumer video market to blossom soon. Noting that the company has retargeted the engineering resources originally engaged in broadcasting applications toward consumer MPEG-2, Bismuth said that LSI Logic expects consumer systems using such a device will first appear in early 1999.
Integration is a key aspect of C-Cube's plan to attack this market. Besides the 100-MHz Microsparc processor, the new chip integrates C-Cube's own programmable DSP and Motion Estimation co-processor. The DSP, running at 1 billion operations per second, performs such video compression tasks as discrete cosine transform, inverse DCT and motion compensation. It also offers functions such as pre-processing filters to reduce the incoming noise for digital video.
In an effort to make an encoding solution as glueless as possible, the chip incorporates several interface blocks. PCI bus mastering makes it easy to integrate the chip with PCs and workstations using a PCI host bus. The on-chip synchronous DRAM controller interface generates the addressing and controlling signals necessary to support the local SDRAM. DVx requires 8 Mbytes of unified memory--four 16-Mbit SDRAMs--for both encoding and decoding.
With the new architecture in place, how is C-Cube going to convince a broad base of system vendors already using a variety of existing codecs to switch to MPEG-2? For example, Motion JPEG is still a prevalent compression method among many non-linear editing systems, while DV or DV Pro codecs are being used in professional video cameras and high-end consumer camcorders for digital acquisition.
Balkanski argues that there are obvious advantages to making the switch. "It's, of course, all cost and performance related," he said. "With a cost-effective, single-chip MPEG-2 codec, it's now possible to do everything from digital acquisition, non-real-time editing to publishing on the MPEG-2-based platform."
Furthermore, many consumer-electronics manufacturers have decided that "a true candidate to replace VCRs is a rewritable storage medium," said Balkanski--not a digital VCR based on a DV codec, a product that has begun to emerge on the consumer market. "The power play is in a rewritable DVD such as DVD-RAM." MPEG-2 is an obvious engine for consumer digital video recorders, he added. |
