MUNICH, Germany — Texas Instruments Inc. is using vertically oriented engineering teams to ensure the success of its fourth-generation wireless design and development efforts.

"We're looking at an ambitious goal [of getting] millions of 4G multimedia sets into the hands of consumers by 2010, and that will take a new, multidisciplinary approach," Dan Shaver, director of communications systems and the SoC R&D Laboratory at TI, said Wednesday (March 8). Shaver's comments came in a keynote speech here at the Design Automation and Test in Europe conference.

The goal is to make a $10 semiconductor device operating at 100 MHz, with 10 million gates and memory.

Shaver said he envisions a system-level design environment to handle the complexity of 4G designs. "You have to establish a startup operation for this vertical team so that there is no handoff of the design" to protoyping, he said. "All team members know exactly the requirements of the project and proceed accordingly to the prototype stage."

Shaver's scheme calls for five system engineers working on 4G algorithms: MIMO, RF, VLSI, physical design and software. Algorithms are sent to system architects, who partition the design into hardware and software tasks and create the prototype for SoC design.

TI's so-called Hollywood chip is a first step toward 4G devices. It was fabricated in 15 months using this flow "from kickoff to tapeout," Shaver said. "It was a first-time silicon success and is now in pilot production."

The mobile-TV chip (the DTV1000/1001) was recently shown at the GSM World Congress.

Chip complexity for emerging markets like 4G wireless will require both designers and software engineers to have applications expertise. "Systems engineers are becoming designers who form the [basis for] small, well-trained, efficient teams for developing the differentiating intellectual property," said Shaver.

"Accommodating the ever-changing mobile standards in a chip of minimum power and maximum range is not easy," he added. TI's approach is based on an orthogonal frequency-division multiplexing computational model, which must comply with multiple international standards.

"That results in 1 billion complex, two-element dot products per second, per antenna—a real challenge," Shaver said.