Out-of-the-box thinking is clearly needed if mobile phone design is to enable the ubiquitous services that future subscribers are going to expect. Many industry observers believe the coming wave of enabling technology will be based on microelectromechanical system (MEMS) design.

Indeed, MEMS devices have recently proved their usefulness in high-volume consumer market applications as diverse as microphones and gaming consoles. We are reaching a point where no system will be complete without the integration of MEMS functionality. In this way, MEMS is the new analog, the element every system has to have to make it functional, flexible and able to interface with the outside world.

While Moore's Law describes the progression of transistor density and computing power, the integration of MEMS will act as a multiplication factor and put many of the functions that have previously required hybrid implementation directly on the chip.

Today's strongest trend in RF design is the drive toward configurable/band-free radio and antenna design. It is increasingly beneficial and ultimately necessary to make RF components digitally reconfigurable, so that frequencies and impedance levels can be accurately and digitally controlled, and continuously optimized for best system performance. Such a reconfigurable front end is able to switch among frequencies and communication standards literally in the blink of an eye, while reusing the same signal path.

By combining MEMS technology with mainstream semiconductor manufacturing processes to build a low-loss RF capacitor element with on-the-fly digital tunability and cost efficiency, WiSpry has enabled dynamic RF technology that, in turn, enables a true software-defined radio, where the RF front end is digitally controlled by the baseband and all the standard-specific functions are loaded as DSP programs. Once the front end has become digitally tunable, much of the RF engineering work moves into the software domain, greatly reducing the number and cost of hardware (re)designs and the time spent manually tuning a circuit.

The programmable front end may be used across multiple platforms and even provides some "future proofing," as new responses may be loaded into the platform's firmware.

Standards glut

Today, most wireless standards use two frequency masks for transmitting and receiving data--also called frequency duplexing--within the frequency band stipulated by the spectrum allocation plan. Because of regional differences in spectrum allocation, as well as the rapid evolution and sheer number of competing wireless communication standards worldwide, there has been a multiplicative effect on the number of frequencies a global cellular phone platform must support. The drive to use the radio spectrum as efficiently as possible and to support new services in previously unused spectrum gaps is driving this trend.

However, the technical requirements that devices must fulfill in order to gain network entry have not changed. In fact, high-performance individual hardware solutions for the RF front end are required to provide the necessary selectivity, linearity and isolation while minimizing insertion loss and power consumption in the circuit.

As a typical example, with seven bands integrated into a phone, at least five separate sets of RF components (chains)--including multiple antennas--are required, in addition to an eight-throw switch or higher to select the desired band of operation.

With the first cellular phones, radio designs were single band, and subscribers were excited about being able to place a call away from their desk. RF designers had one frequency band with which their design had to work.

As technology quickly progressed, however, dual-band phones suddenly became a requirement in order to support an increased number of users. As those users started to travel with their phones, triple-, quad- and penta-band phone designs quickly became the norm, adding to the designer's headache.

As further bands have been added, the additive RF design approach has become progressively less tenable. The simplest headache is the expansion of size, cost and complexity.

Adding band coverage is a nearly linear progression. It is sublinear as, first, the switching solution has continually improved in tandem with increases in the throw count. Second, as was discussed earlier, technology continues to enable the individual band elements to be smaller and less costly in each successive generation, and third, many of the individual elements are now combined into modules, thus reducing overhead but not solving the fundamental issue.

Today there is a widespread realization within the cellular phone industry that simply continuing on this path is not feasible. Beyond the complexity, size and cost implications, the multiple-chain approach imposes a fundamental performance limitation.

Each chain has somewhat different impedance characteristics for the band of interest. If each chain had an individual antenna, the overall chain could be optimized. However, individual antennas are neither space- nor cost-effective and can have significant cross-coupling. Thus, the chains are forced to combine into a single path using switches and filters.

Even assuming perfect switches, it becomes increasingly difficult to maintain high performance for all bands as each new band is added, since compromises must be made in the shared circuitry.

Also, since every component in the chain has its specific fixed-frequency response, the band edge performance is typically suboptimal.

Single-chain solution

All of the above issues can be avoided if the RF front-end components are tunable. A single chain can then be optimized specifically for the channels currently in use.

The benefits of the single-chain approach are widely appreciated--as are the challenges remaining in the way of implementation.

Research into tunable front-end components has been progressing for a couple of decades, but only now are the requisite technologies maturing. The traditional stumbling blocks have been size, cost, repeatability, reliability and performance. Each of these has been addressed in part by previous work, but WiSpry is the first to bring to market a complete solution that is suitable for high-volume production at a price point that is compatible with the cell phone industry.

By pioneering the integration of high-Q MEMS capacitor elements into a mainstream RF CMOS process technology, WiSpry brings together the benefits of a high-volume, low-cost process with the advantages of high-performance RF MEMS technology. Individual capacitor elements are integrated on-chip as tiny parallel plate capacitors with a digitally variable air gap. Individual shunt or series elements are combined into capacitance cells and then into arrays that can contain any combination of individual cells, resulting in a digital capacitor that is well behaved and free of higher modes, with capacitance ratios (max/min) greater than 10 and a quality factor (Q-value) well over 200 at 1 GHz.

The manufacturing of this device benefits from the latest advancements in CMOS semiconductor process technology. WiSpry is using a fabless model that integrates the programmable digital capacitor technology monolithically on mainstream 8-inch RF CMOS wafers that can be produced in extremely high volumes, thereby eliminating the size and cost concerns traditionally associated with high-performance MEMS technology.

The process flow also includes wafer-level encapsulation so that finished wafers from the foundry can be utilized directly in conventional, automated back-end processing (such as bumping, thinning, dicing, packaging and test), making for a highly reliable end product that can be produced in a traditional RF semiconductor manufacturing flow.

No external circuitry

So how does the component work, and what must the designer need provide in order to use the technology?

There is no need for external circuitry, as the component truly works like a high-Q capacitor with integrated serial interface. All the support functions for the MEMS elements are integrated on the chip.

By loading a digital word that contains the desired setting of the digital capacitor elements over the serial bus, the internal logic and driver circuits immediately set the capacitor bank to the selected value.

This programming can be repeated at high rates to create dynamic RF functionality, which has a significant number of interesting applications.

As programmable chips are integrated into custom modules together with other high-Q integrated passive and active components and support circuitry, they form a platform that WiSpry will use to provide programmability to the complete RF front end.

The work starts t the antenna, with frequency agility functions and mismatch tuning, and will address the rest of the RF chain in a natural sequence.

Marten Seth (Marten.Seth@wispry.com) is director of product planning and business development at WiSpry, a fabless semiconductor company based in Irvine, Calif.

Arthur Morris (art.morris@wispry.com) is chief technology officer at WiSpry. The company is pioneering the use of RF MEMS in industry-standard CMOS technology and is currently sampling its first products to key customers.

 See related chart