Of the developers of SoC solutions for wireless handsets, wireless infrastructure, and high-speed networking, only a minority are considered leaders in terms of in-house expertise in DSP cores.
Most of this DSP technology is proprietary and not available in the open market. Also, applications and software standards are continually evolving, further challenging software developers to keep pace. Semiconductor companies must choose between spending valuable time and money developing-DSP cores and the software to support their applications, or licensing DSP cores externally and leveraging software from third-party providers. OEMs often choose the reduced flexibility of buying SoCs with proprietary DSP architectures. Many feel there is a need in the marketplace for greater availability of modern, high-performance, low power, licensable DSP core intellectual property (IP) for inclusion in communication SoC designs.
One advantage of the IP approach is the flexibility to customise the SoC product by selecting a DSP core that can be customised. Other advantages include the peace of mind that comes from selecting proven technology, including hardware and software that will evolve with changing standards.
Designers can then choose the level on which to differentiate their particular device, whether it's in hardware, software, or both. For these designers, the IP approach allows them to optimise the key parameters of performance, cost, and power consumption, all with a fast time-to-market. This is a shift in design practices for some, but it is a relatively safe one. This same model has been used in the microprocessor industry for years with established IP vendors such as ARM and MIPS.
Semiconductor manufacturers and OEMs also want to customise their SoC product with peripherals, memory, buses, interfaces, and process technology to meet the needs of their target markets. They also need the flexibility to migrate their designs to alternative semiconductor process technologies, or even different foundries, to better address parameters such as economics, performance and power consumption. And they want to do this without spending lots of time and money adapting the DSP core to the new process technology.
A common DSP core platform can enable the use of the same design across different applications, such as using the same high-performance DSP core in the base station as well as the wireless handset. This portability leverages design and development environments, increases code reuse, and helps to reduce engineering time for DSP programming, which in turn improves time-to-market.
DSP core technology vendors that specialise in the needs of this marketplace offer designers the ability to specify the requirements of their DSP core and/or DSP subsystem.
With the IP approach, designers can access state-of-the-art DSP technology, which now can be efficiently programmed using high-level languages such as C and C++. Newgeneration DSP cores that are codesigned with the compiler can effectively achieve minimal cycle counts and high code density. With the freedom to buy the best available DSP technology from multiple semiconductor manufacturers, OEMs have bargaining power to lower costs as well as second source or multi-sourcing options to ensure product quality and on-time delivery.
In the past, if a high-volume wireless handset manufacturer wanted to multi-source from two DSP vendors, their engineeringteams would have to implement the DSP software twice for the SoC product—typically, the two DSP vendors employed different core architectures. Today's DSP IP core environment simplifies this model. If an OEM sources from two semiconductor companies using the same DSP core, the DSP software can be implemented once, which offers significant savings in engineering time and costs.TIME AND MONEY In fact, OEMs and semiconductor manufacturers who need specialised DSP core IP for SoC designs can save two to four years development time by licensing DSP technology, rather than developing it in house. In general, it can cost US$30 to $50 million to develop a competitive DSP core architecture; software development environment including the compiler, assembler, linker, and simulator; and suite of supporting application software. In addition, it's necessary to sustain this level of investment to ensure a competitive roadmap with timely new product introductions.
OEMs and semiconductor manufacturers should also recognise that if they choose a proprietary DSP solution, the cost, quality, and delivery advantages of multi-sourcing would be lost.
To further reduce time-to-market and cost, a good IP provider will bring a system approach with complementary IP blocks and integration. This solution is typically balanced and optimised with the core, providing a means to realise the full potential of the DSP IP core.CHOOSING A DSP IP CORE PROVIDER Once an OEM or semiconductor manufacturer has decided to go with DSP IP, the next step is selecting a vendor. To guarantee that semiconductor manufacturers and OEMs are buying the best technology, they should choose a DSP IP core that is independent without obligation to companies that could compete or otherwise deter full interest in the success of all IP licensees.
Designers should make sure that the DSP IP core provider has a proven and broad product-in-service portfolio. One way an IP vendor can achieve this is with mass-customised DSP IP solutions, i.e. multiple versions of similar cores manufactured with different process technologies. However, it's also important that the IP core meets the needs of the customer without additional overhead for functionality that's not required.
A good DSP IP core provider can offer measurable time-tomarket advantages to its customers.That's because it has leveraged its relationships with third-party vendor partners, including RTOS, integrated development environments (IDE), application software, and system solutions.
This combined effort will result in market momentum that will help ensure the long-term viability of the architecture and guarantee that customers will always have a software-compatible roadmap for the future. For instance, a DSP architecture that can be leveraged across multiple markets, such as the SC1000 family from StarCore, allows third-party vendors like Enea OSE, Green Hills, Ittiam, Lineo, Metrowerks, Signals and Software, Surf, and Trinity Convergence to support the success of the same DSP IP for multiple SOC vendors.
This can significantly improve the availability of software IP for SoC applications.
As time moves on, code-intensive applications will make reprogramming DSP solutions in assembly even less appealing to the engineering budget. For instance, a typical 2G wireless handset requires approximately 150,000 lines of DSP C code, while a feature-rich 3G handset can require up to 10 times that amount. If the DSP IP provider has a strong C/C++ compiler that minimises the cycles and memory used for key communications applications, it can make these coding challenges easier.
Next-generation applications require SoC solutions that feature complex, high-performance, low-cost, low-power, memory-optimized solutions with an improved time-to-market. Quality development tools, a scalable DSP core, broad application software support, superior systems design assistance, and a solid roadmap of DSP solutions that meet all of the latest design challenges are what OEMs and semiconductor companies will require.
For the wireless market, that includes support of 384kb/s to 2Mb/s data rates, network access to enable multimedia, still-picture and streaming video capabilities, low power consumption, and DSP performance of 3000+ MIPS. For wireline applications, this translates to support for 1 to 6Mb/s data rates, interactive online experiences, home networks, low system costs and similar DSP performance.
The DSP IP core vendor should also provide systems, which include application and low-level software integrated with the hardware.
The support of software for future wireless environments is significant and will continue to grow. A good DSP IP provider will provide solutions that provide an optimised balance of hardware and software, while leaving "windows" where semiconductor companies and OEMs will be able to still differentiate their solutions.
The road to next-generation, DSP-enabled designs is complicated by increased device complexity; cost, power, and size constraints; software; and time-to-market pressures. By licensing a proven DSP core IP solution, designers can leverage state-of-the-art hardware and software, maintain focus on their core competencies while enjoying investment savings. DSP IP complemented with optimised application software and development tools provides OEMs and semiconductor companies with a competitive advantage and the ability to put products in the market earlier.