Integration by function puts more features, interfaces into handsets

For the past decade or so, wireless telephone designers, with the support of the semiconductor industry, have relied on increasing device integration to pack ever-greater functionality into small, sleek, low-cost wireless handsets. In general, their integration strategy has been tailored to IC technologies suited to particular groups of functions. High-performance BiCMOS, silicon germ anium (SiGe) or even gallium arsenide have been the standard for the radios. High-voltage wafer processes with good analog characteristics have been applied to the analog and power management areas. Dense logic and specialized memory technologies have been used for the data-processing functions.

This technology-based strategy served wireless designers well through the 1990s, when most handsets incorporated only a single band of operation, typically around 800 to 900 MHz, with some offering both analog (AMPS) and digital modulation support. By the end of the decade, though, dual-band handsets became common, and multimode handsets — that is, handsets that support more than one multiple-access technology — appeared on the scene. Today, integration of Bluetooth, GPS, WLAN and multimode wireless electronics is the challenge. Along with a dramatic increase in the number of air interfaces to be supported and concomitant complexity, the applications available through those technologies bring new de mands for high-performance data processing.

Can an integration strategy based on like semiconductor technologies continue to serve as wireless handsets evolve further toward multimedia communications appliances? Can OEMs rely on that strategy as they develop diverse products with differentiated feature sets? Can a technology-based strategy take us into the era of true system-on-chip wireless handsets? This article explores those questions. It discusses the limitations of technology-based integration and examines an alternative strategy available today: function-based integration.

Today's wireless handsets must include high-performance radio electronics to process signals from the Global System for Mobile Communications (GSM) band at 800 MHz up to the industrial, scientific and medical radio band for Bluetooth and wireless local-area network (WLAN) support at 2.4 GHz.

System requirements

Power-management electronics in handsets must provide high-voltag e support up to a few times the battery voltage so that no breakdowns occur during battery charging. Analog electronics are required to provide a high-quality audio channel and A/D and D/A conversion of the radio channel signals. Very high-performance logic and dense memory also are required for processing baseband signals, maintaining protocol and processing user applications such as games, video, music and positioning services.

From a wafer process perspective, it might seem that the most direct step forward would be simply to pull all the devices onto like wafer process technologies into common integrated circuits. The result, clearly, would be a radio interface IC, in a SiGe process, that would provide multiband, multi-mode radio support; an analog/power-management IC in analog CMOS that would provide analog and power-management support for the full system; and a large logic function that would provide processors, logic and possibly, memory, to meet the full system's needs.

It is this pa th that wireless handset designers and chip makers have followed in years past. But the strategy involves inherent limitations that may make it inappropriate as we move forward.

For one thing, it requires a large number of interfaces between the ICs to accommodate signal routing. For another, no innovation that includes logic processing close to the radio front end can reduce power or external passive devices because radio, analog, power management and logic are segregated onto separate ICs.

For OEMs, the lack of flexibility in that approach may be an enormous disadvantage. Integrating all of the radios, for example, because they rely on like wafer technologies, means that fixed functions are, in effect, built into every handset. To produce a low-end handset with limited functionality — such as one with GSM and Bluetooth, but not WLAN — may require developers to bypass radios available on the device. The greater disadvantage, though, arises when new and emerging technologies require the OEM to add radios in order to remain competitive. In many cases, advances of that kind would require redesign not only of the integrated radio IC, but also of the entire system.

Perhaps the greatest disadvantage inherent in this strategy is that it, inevitably, brings integration to a dead end. As the drive toward the true system-on-chip continues, incompatible wafer technologies simply will not permit further integration. Major system overhauls will be required to migrate components from one wafer technology to another so that radios, logic, memory, analog and power-management support and data processing can reside on the same device.

Overcoming those limitations requires an integration strategy that's dramatically different from what many semiconductor designers have used to date.

Function integation

The most promising appro ach integrates, in CMOS silicon, all of the activities needed by a particular function. For example, a complete quad-band General Packet Radio Service transceiver, including all baseband analog and RF functions, may be integrated into a single chip. A GSM module contains all of the radio, logic, A/D and D/A converters and so on needed for GSM operation. Similarly, Bluetooth, WLAN and GPS are also supported with their own single-chip solutions.

Integration in this manner provides considerable benefit. To begin with, it overcomes the interface complexities and power-management problems inherent in a technology-based strategy. In each IC, the logic and radio functions are in close proximity so that the benefits of digital processing can be applied to the radio, reducing its power and area. All the signal interfaces along the signal-processing chain are included inside each of the ICs. Pin counts of the ICs are reasonable and few sensitive external nodes exist.

Since the functions are fully conta ined on a single IC, the product offering is modular. A variety of handset configurations can be realized without unnecessary cost or wasted functionality. Only needed functions are included in each design.

The next step in integration — combining the functionally integrated ICs into complete system ICs — can be achieved in a routine fashion as high-volume market segments emerge that warrant it. No new basic technology is required once the technology needed to support functional integration in CMOS has been put in place.

The major concern with the integration of handset electronics by common function is the challenge of developing a wafer process technology and a circuit technology that cost-effectively allow such high levels of integration. Wafer reticle counts must remain very close to baseline logic CMOS to avoid excessive wafer cost. At the same time, circuit technology must leverage CMOS logic capability to reduce analog and RF performance requirements. This is by no means a s mall challenge, as literally hundreds of radio performance requirements must be worked through in detail.

Recently, however, innovations in circuit design, advanced 130-nanometer wafer processing and system architecture have enabled such integration. Power, performance and cost are all favorable to multi-IC implementations on the market today.

With function-based integration, a wide variety of handsets offering different combinations of air interfaces and user features will soon be possible. From there, as the ever-evolving integration of electronics continues, high-volume market segments will be identified for further integration of full combinations of functions on a single silicon IC.