Moore's Law and Analog Technology - A Power Amplifier Case Study

Hans-Otto Scheck, Principal Engineer, Nokia

Moore's law is a well accepted and accurate tool to describe the advancement of digital semiconductor technologies. For analog technologies like radio technology, different roles apply. This article describes the development of analog technologies by the example of RF power amplifiers. The impact of Moore's law but also significant differences between digital and analog technologies are described.

Moore's law and the IT world.

The well known and often cited Moore's law doesn't present a law of nature; it's rather an empiric observation. Gordon Moore (Fairchild Semiconductor) predicted 1965 a doubling of the number of transistors per digital chip at about every two years.  His early observation of the trend in IT proved to be correct over the last four decades and is expected to be valid for the coming decade.

The very basic function of digital circuits is the ability to store information. Though the processing of information is the essential part of any computer or signal processor, it can be considered as secondary at this point. To store information, a certain amount of energy is needed. The energy might either be used initially to write a non-volatile memory or might require some continuous energy to keep the information. Important is the fact, that ever smaller energies are needed to store a certain bit of information, and as a consequence, to process the bits. Scaling is the most dominant factor for Moore's law, although the actual technology development is much more complex than just scaling. The essential physical parameter driving Moore's law is and remains the reduction of energy per information unit. Gene Frantz from Texas Instruments reformulated Moore's law into ”The power consumption per digital operation is halved every 18 months” (figure 1). With this step he introduced the power consumption as an essential parameter. Power consumption is a common figure of merit for both digital and analog technologies.

Why Moore's Law isn't Applicable to Analog Technologies

As outlined above, digital technology is about information storage (and processing): Moore's law reflects the ability to reduce the amount of electrons needed to store and read a single bit. Analog (radio) technology is about the transport of information, e.g. the ability to detect a bit at a certain distance. The minimum energy needed is described by the Shannon theorem:

According to Shannon we can trade capacity (and spectral efficiency) against signal energy. As the radio spectrum is a limited resource, system capacity becomes a function of available energy. Shannon's theorem is the analog technology's hard limiting factor, similar as Heisenberg's uncertainty limit is the ultimate limit for information storage. In order to judge the potential of future developments it is an important factor how far the respective technologies are from their absolute limits.

Development of Analog Technologies

When analyzing the performance of a technology, it always has to be seen in a larger eco-technical context. While Moore's law predicted continuous growth over decades, such simple relations rarely exist. Analyzing radio systems, there is a more profound performance factor beyond spectral efficiency, the limited efficiency of the DC to RF conversion. The transmitter's power amplifier is usually the dominant factor of power consumption. Therefore, radio development can be measured by power amplifier efficiency. Initially GSM amplifiers were below 30% efficiency, which was improved over time with better semiconductors and circuit optimization to around 30%.The spectral efficiency of GSM was relative poor, which wasn't a limited factor due to the low penetration of mobile phones in the early 90's. With the success of GSM, spectral efficiency became a more dominant factor than amplifier efficiency. With the introduction of EDGE and WCDMA, the PA efficiency dropped because of the higher order modulation, introduced to improve spectral efficiency on the cost of DC efficiency.

The first WCDMA power amplifiers had efficiencies below 10%, which spurred the development of new amplifier architectures. Today's high-end (digitally) linearized WCDMA power amplifiers reach close to 40% efficiency. With a realistic limit of 80-90% efficiency doubling the efficiency, equivalent to one digital generation step, would bring us to the very limits. We can't expect to reach the limit within a single development step or in 18 months but rather see a smoother asymptotical approach to the limit as described in the figure 2 below. Or new requirements push us back again and set the starting point for a new asymptotic line. Wideband amplifiers, which can handle a channel bandwidth of 100MHz and more, are most likely the next challenge in RF design.

The different pace of digital and analog technology development is a particular challenge in the design of a cellular base station with its long life cycles. Once the mechanical design is fixed, new sub-modules have to be fit to the given form factor, which leads to increasingly inefficient solutions until the expensive step of a complete redesign is unavoidable. One solution is the modularization with well specified interfaces to separate the radio modules and signal processing units. A well designed radio module might even allow the support of different radio standards, as long as they use the same radio frequencies.

An important step to support modularity and optimally utilize technology advances despite different development cycles has been achieved with the OBSAI interfaces. OBSAI has specified a range of BTS interfaces, including an optical interface between a system unit and multiple radio units. For a list of OBSAI members and further information go to: http://www.obsai.com/

Author Biography

As Principal Engineer for radio architectures at Nokia’s Networks business division, Hans-Otto gives technical direction for cellular radio architecture development, prepares cellular infrastructure hardware vision, radio architecture roadmaps and mixed signal strategy. He joined Nokia’s Research Center as a research manager for radio system implementations in 1994 and moved to the Networks unit in 2001. Before joining Nokia he worked in several research labs in Germany and Finland in the fields of wind energy, satellite navigation, bi-static radar and microwave circuit design. He holds 8 patents in the field of telecommunications with several applications currently pending.