A two-year-plus slump has seriously hamstrung the economic viability of the semiconductor industry. According to Gartner Dataquest, the worldwide semiconductor market dropped 31.2 percent in 2001 compared to 2000, a record plunge, and grew only 1.9 percent in 2002. While positive, Gartner Dataquest's semiconductor growth forecast for 2003 is only 8.3%, nothing to write home about. For next year, growth is predicted to be a more healthy 23.1%.

Chip manufacturers are looking to identify the next big silicon-fueled high-growth market. Companies that choose correctly stand to leapfrog their competitors as the electronics industry starts to turnaround.

In this article, several semiconductor leaders share their thoughts on what they think the next silicon winners will be and why.

	For more information on SiP technology, check out System-in-Package Technology, Application, and Trends.

Not surprisingly, Tobias considers power consumption an important issue, and not just for portable equipment. A problem with low-power design is the lack of good EDA tools for power analysis and optimization. In fact, at 90 nm Tobias considers the current EDA offerings to be inadequate in general, especially for back-end place-and-route operations.

Tobias does have a vision of how future products will evolve that seems to go against popular opinion—he thinks that, in the near term, there will not be a convergence of the portable appliances you use. In other words, he feels that everyone will still have a laptop (or notebook) computer, digital camera, cell phone, PDA, and so on. Tobias bases his reasoning on the desired form factor of a device. Simply put, it is easier to use a camera when it looks and feels like a camera than when it is embedded into a phone or PDA.

	Is UWB Just Around the Corner? If you're talking about consumer high-definition digital-media applications, UWB is probably on the horizon. If you want to learn about UWB technology and general applicability, there is plenty of literature available on the Internet (for example, see Ultra-Wideband: It's Not Just Hype and Ultra-Wideband Technology for Short- or Medium-Range Wireless Communications). However, UWB will almost assuredly be in the picture for home-entertainment applications—consumers want to be able to handle high-definition video. 802-11b (WiFi) and 802.11g haven't the bandwidth to handle HDTV signals, which need around 35 Mbps support. 802.11a (WiFi-5) has the bandwidth, but uses too much power for handheld-device operation. Another problem with 802.11 technology is that it is an Ethernet derivative for packet-based data-networking applications. This makes 802.11 a poor choice for intensive multimedia applications, since the technology depends on data packets arriving in order and on time. Furthermore, WiFi-5 NIC cards, at over $100 each, are too expensive for home-entertainment consumer applications. However, at a typical bandwidth of 100 Mbps over a 10-meter range, UWB fits the bill. Since 10 meters is too short for whole-house coverage, look for UWB for intra-room networking with a WiFi backbone supporting the entire home. The 'single room' limitation is one of the reasons that many companies developing UWB-based products are looking for ways to increase this range. Several companies are working with the IEEE to develop a UWB standard, but nothing will appear until 2H04 at the earliest. Among the issues contributing to the definition of a UWB standard is that of single-band vs. multi-band implementations within the 7.5 GHz band (3.1 to 10.6 GHz) allocated to UWB equipment by the FCC. Potential UWB product vendors are currently developing products with capabilities to increase both bandwidth and range. Look for high-end DSP chips to be a major factor in the development of future UWB products.

According to Kazi evolving standards require chip reconfigurability, probably via software. He also feels that successful products need to be able to handle several different types of audio and video formats, particularly since the user of these products will want to manipulate as well as play back digital media. Kazi states that power for these living-room applications is not as important as chip integration and the accompanying cost reduction of systems using these chips.

To achieve persuasive computing, Mafie sees wearable computers with wireless capability powered by faster and more flexible processor engines. The computer's silicon will give it capabilities such as true voice and gesture recognition (for multiple persons) to save keystrokes, multi-processing to handle multiple functions, GPS, and lots of Flash memory to handle new programs "on the fly."

SoCMosaic is available as soft silicon IP, RTL code you can synthesize for 180 nm, 130 nm, 90 nm or 65 nm process technologies. The platform, which runs under the Linux OS (with others available on demand), also comprises pre-tested silicon-IP blocks, standardized bus interfaces, an RTL testbench, and high-level, cycle-accurate C models. According to Tobias, SoCMosaic can cut total design time for a customized chip from a typical 18 months to as little as four months.

SoCMosaic contains common peripheral functions such as I/Os, interrupts, counters and serial ports plus the ARM processor. Customers then select silicon IP such as embedded DRAM; system interfaces such as Ethernet, USB, 1394, PCI controllers, and SerDes; and optimized hardware/software application functions such as VoIP, MPEG, and 802.11 (Figure 1). Toshiba also has several analog-IP functions with variants for each block to meet design requirements such as high speed, low power, small area, or low noise.

Figure 1:  SoCMosaic combines several timing and communication cores with an ARM 9 processor, connected via an OCP bus. You customize the SoCMosaic chip by adding additional cores to meet specific application requirements.

The first SoCMosaic platforms will have a single ARM 9 running Linux or another RTOS for low-end networking and consumer applications. A second version of the platform, expected later this year, will support multiple processor cores for high-throughput multimedia and high-end networking applications.

Cuomo does not see process technology as a barrier—ST has 90 nm already available with 65 nm in the works. He also feels that wireless bandwidth is not as big an obstacle as suggested by others, since it is only an issue in real time (several applications, such as receiving video, can be done during relatively long time periods and then played back locally in real time). With the increasing functionality of information appliances, Cuomo does sees power as a very big problem that designers must overcome through both software and hardware development. JPEG and other video functions are 'power hogs', and displays also dissipate relatively large amounts of power. He thus sees a need for substantial improvement in battery technology for powering future consumer handheld devices.

Also important, according to Cuomo, is the need for better 'synchronization' between software and hardware designers, service providers, and manufacturers (the supply chain participants). This will most likely require a new business model between silicon, system, service, and manufacturing partners.

Nomadik uses a distributed-processing architecture to divide the various tasks the platform must handle. Along with an ARM926 core, Nomadik has smart accelerators, low-power CODEC engines that support multiple A/V standards and formats for creating, sending, and receiving images and video clips, as well as audio and video communications (Figure 2). The accelerators operate independently or concurrently with the CPU, depending on task requirements. By handling all A/V functions, including pre- and post-processing, these engines free the ARM CPU for control and program-flow tasks.

Figure 2:  The Nomadik platform combines an ARM 9 core with smart accelerators to handle the coding and decoding of both audio and video signals.

The Nomadik platform supports the OMAPI wireless standard, initially proposed by ST and Texas Instruments late in 2002. The companies' goal with OMAPI is to enhance how multimedia looks and runs on cell phones, PDAs, and other portable appliances. ST and TI based OMAPI on TI's popular OMAP standard, but added a specific set of software and hardware interfaces defining common application peripherals that apply to multimedia applications.

The current version of Nomadik has two smart accelerators for audio and video processing. A C-programmed DSP core powers the audio accelerator while the video accelerator combines software and hardware. Future versions of Nomadik will contain a DSP core and additional smart accelerators for other application-specific functions.

Franz and Rasor believe we need to rethink how we communicate over long distances. Rasor feels that short-range communication, relayed by repeaters and bridges to increase range, may be the way to manage power in handheld communication devices. This communication scenario would take UWB-enabled devices out of the 'in the home only' environment and allow them to communicate with other devices that use current WLAN technology (by FCC decree, you can currently use UWB only indoors or outdoors in handheld devices for peer-to-peer operation) . Rasor also sees future cell-phone operation as a confluence of audio, video, data, and transactions (such as banking and paying bills), making tomorrow's cell phone look much like ST Cuomo's all-in-one consumer appliance.

Frantz also brings out an interesting point that is key to TI's general development efforts. The company seeks to create technology that other companies can then leverage for their own products. This idea helps accelerate the acceptance and use of new technology and products that employ this technology by the end user. Frantz also feels that we need to figure out a way to create the various components (hardware and software) in a system to be able to easily integrate them. In addition, we also need a roadmap for integrating the system towards the goal of a single chip without having to redo the system each step of the way.

According to TI, 1 GHz DSP performance will fuel consumer home products such as a media center for viewing HDTV on a variety of entertainment and productivity-enhancement devices, including TVs, PCs and PDAs, with no wires, while simultaneously recording high-definition content for later viewing. These applications match well with those forecast by Toshiba's Kazi for the intelligent home-entertainment system.

Jim Lipman is a consultant providing marketing, writing, and other electronics industry services, specializing in EDA tools and ASIC/SoC design methodologies. His job experience includes chip-design R&D, marketing, marcom, technical editing, and on-line publishing of technical content for engineers.