To provide improved efficiency and convenience, disparate technologies are quickly converging on portable devices such as PDAs, cell phones and MP3 players to create integrated, supersmart Internet appliances that offer broad functionality. Bluetooth is leading the way in the convergence market by offering wireless connectivity between personal devices and existing wired or wireless data networks. This will allow access to multimedia data and Internet on the move, and ultimately, revolutionize the way we communicate.
Bluetooth is being promoted as a cable replacement technology, but it is much more than that. By allowing personal devices like your cell phone and lap top to communicate with each other, data will always be consistent across these devices and remote control of personal devices will be possible. More importantly, the automatic discovery-without user intervention-of public-access devices like printers or LAN access points that are available in meeting rooms or public spaces, will enable a new level of transparent ease of use.
Bluetooth capability can be grafted as a block onto existing products for the short term, but to compete long term and establish itself as the new wireless connectivity standard, Bluetooth will have to be deeply integrated into pre-existing functions in target appliances. In addition, Bluetooth must be introduced rapidly and smoothly across a wide range of telecommunications and computing devices if it is to become the prevailing wireless LAN technology.
The market for Bluetooth silicon is expected to exceed $2 billion by 2005, and there are tremendous business implications for Bluetooth-enabled portable devices that offer multiple functionality. But Bluetooth can only succeed as a standard for wireless connectivity if it can be cost-effectively integrated in a variety of handheld and wireless devices.
The current market for portable devices is segmented into entertainment, such as CD and MP3 players; mobile connectivity devices, such as mobile phones, pagers and CPDP modems; and business/professional devices, such as PDAs and notebook PCs. New device classes that integrate the functionality needed for personal entertainment, mobile telephony, internet access and computing into the same handheld device call for wireless connectivity that is easy to deploy to a wide range of peripheral devices and other networks.
A common method of integrating multiple functions into a single portable device is sourcing the missing pieces of the desired functionality from outside venders and integrating fully functional modules with loosely coupled interfaces to the host system. This approach has numerous problems, however. Modules, because they are predefined, limit how much you can reduce the footprint and power of the end product. They typically contain large and bulky subsystems that introduce redundancy as more modules are added. Because each module operates on its own circuitry, the end product does not benefit from sharing resources.
Additionally, plugging modules into an existing design means that the designer must sacrifice the flexibility of tight integration. Typical costs for Bluetooth modules in Q2/3 of 2000 will be between $20 and $30, much higher than the goal of $5 Bluetooth SIG set in 1998.
Bluetooth will be the first mass-market wireless technology to be used in CMOS for budget reasons, and the solutions that are established here will be extended to other wireless technologies. A key issue in finding the lowest-cost implementation of a device incorporating Bluetooth functionality in CMOS or any other process technology is system partitioning. This concerns both how the radio and baseband are partitioned from each other, as well as how the baseband is partitioned from any other system-on-chip (SoC) ASIC present in the device.
Using a digital interface with low-voltage signaling is one way to partition the radio from the baseband. Such an interface lets the IC designer optimize the internal architecture and circuit topology of the RF or baseband IC while ensuring compliance with the interface and, hence, interoperability with radios or basebands from other suppliers. A standard digital interface facilitates multiple sourcing for radios and basebands, which encourages competition and enables OEMs and systems integrators to choose the lowest-cost radio that meets their requirements.
A more integrated solution is to put the baseband IP and processor onto the same CMOS die as the radio. Using this strategy, it may be possible to achieve a very low die cost compared with the more traditional choice of partitioning the radio into a separate BiCMOS or SiGe BiCMOS die, then implementing a separate CMOS baseband ASIC.
It is important to note that using IP in an ASIC solution with a digital interface between radio and baseband does not preclude the integration of both functions onto the same die. If anything, integrating in two sequential stages (i.e., proving radio and baseband separately in CMOS silicon and then integrating) is less risky than attempting to overcome the combined "RF CMOS" and "RF/Baseband CMOS" challenge in one step. And reduced risk could produce earlier revenue and market share from first-generation product, supporting second-generation development.
An alternative approach that leverages synthesizable IP blocks as RTL code rather than separate modules lets designers take advantage of tightly coupled interfaces for a finer granularity of integration, resulting in greater flexibility and a smaller form factor. The net result is a low-cost device with more differentiated functionality that can be easily customized for a given application. In any case, it's crucial that the system designer takes a top-level view of how to combine the technology blocks, and keeps an open mind about what approach will produce the best product at the lowest cost, which in turn will hinge on the most cost-effective SoC for the application.
Since combined functionality in a single Bluetooth subsystem requires interoperability between products from different vendors, it's important to have a single standard interface that lets the IC designer optimize the internal architecture and circuit topology of the RF or baseband IC while ensuring that it complies with the interface. A standard interface offers interoperability between the manufacturer's device and radios or basebands from other suppliers, which gives OEMs the advantage of multiple sources. That produces competition, which in turn ensures low-cost solutions.
For some applications, it's logical to integrate RF and baseband functions onto a single Bluetooth ASIC with a host interface that can be added to a pre-existing larger system via a universal serial bus or UART host interface. This system partition has the attendant difficulties of integrating a 32-bit RISC processor, memory and a large amount of digital logic onto the same substrate as a CMOS RF radio. This solution is convenient in terms of concentrating all of the functionality onto a single, high-volume device, which can relieve the OEM from some, if not all, of the responsibility for Bluetooth certification. This approach is also low-risk from the OEMs' point of view, since they can add Bluetooth functionality to an existing system without having detailed knowledge of RF or Bluetooth.
In the long run, being able to partition functions that will deploy a node in the most cost-effective way possible is what will propel Bluetooth into the market. For the handheld devices consumers will carry in their pockets in the next few years, separating the RF from the digital core SoC ASIC capitalizes on the low-cost 0.18-micron (and lower) geometry technologies that are now entering mass production. A pure digital ASIC will benefit most quickly from advances in process technology and testing to ensure the lowest-cost solutions for these platforms without significant re-engineering between product generations.
Integration of all digital computing and baseband functions onto a single 0.18-micron or sub-0.18-micron digital SoC ASIC, as well as the radio functions onto one or more external radio RFICs, offers many benefits for the semiconductor manufacturer or system OEM. An obvious advantage of a digital radio interface is that it avoids many of the pitfalls of existing GSM radios, such as having to generate multiple revisions of the radio to work with different basebands as the baseband process technologies scale between process generations. A digital interface with low-voltage signaling means that a radio can support multiple generations of baseband, even where transitions between process generations occur.
Integrating a baseband IP block for Bluetooth onto a system ASIC is equivalent to adding 200,000 gates, including baseband IP and radio interface, 32-bit RISC, RAM, ROM and peripherals. Once this design exists, it scales from about 8 mm2 in 0.25 micron to 4 mm2 in 0.18 micron and so on, without requiring any modification to the basic design.
Once the digital baseband and the system ASIC are integrated onto the same substrate, there's no need for the baseband to have a USB or a UART to support the HCI. That means the gate count for integration can be reduced by the equivalent of 50,000 gates, taking account of firmware and hardware optimizations. This corresponds to an additional 25 percent reduction.
Finally, as real-time operating system technology improves by including real-time time slicing, it will become possible to eliminate the extra processor required to run the baseband protocol software and to run the protocol stack on the main SoC CPU.
This approach currently is not employed because the real-time performance of embedded operating systems is not good enough to switch between running the protocol stack and user applications without adversely affecting the performance of either or both.
Eliminating the extra CPU and sharing resources could eliminate another 50,000 to 100,000 gates from the incremental gate count required to deploy the baseband functions on the SoC, with the stack firmware RAM and ROM requirements being absorbed into the large RAM and ROM on the SoC ASIC.
In each of these scenarios, not only is the die size reduced, but the total IP content of the SoC is reduced, resulting in lower licensing and royalty fees to the silicon supplier or system OEM. The net result is that integrating Bluetooth as IP rather than as a module, can actually reduce system costs by lowering the price of silicon through the elimination of redundant interfaces. This, more than any other reason, is why Bluetooth will become the pre-eminent host interface for consumer equipment in the next two to three years.