GSM systems have achieved strong market-penetration worldwide. Of the 817 million mobile phones manufactured around the world in 2005, no less than 575 million operate on the GSM standard. When GSM was introduced, the operators used only the 900 MHz band. A few years later, this was supplemented by the DCS band at 1800 MHz. These frequency ranges continue to be used in Europe and Asia, whereas the GSM system became of interest in the USA with the availability of the PCS band at 1900 MHz. Tri-band GSM telephones are the first of their type with global operability to be implemented within a single standard. As the 850 MHz band is now additionally available in the USA, GSM currently operates in four frequency ranges.

In response to customer demand for higher data rates, the suppliers have extended GSM to GPRS services as well as to EDGE. In the GSM system environment, the 3^rd Generation Partnership Project—comprising the European Telecommunications Standards Institute together with other standardization bodies—has defined UMTS as the global standard. This now permits data rates that are three times higher than with EDGE.

Mobile phones will be equipped with more and more functions in the future. These will include streaming of audio and video data via the Internet as well as games. The phones are no longer differentiated on the basis of technical performance. In fact, the end-users decide what a telephone can do, what it will look like and what it costs. This means that costs and dimensions remain the determining factors for the suppliers of components and subsystems such as the RF filter section.

More additional benefits, more processing power

One of the latest changes in the requirements profile for mobile phones comes from the operators of the mobile phone networks: from pure operating and stand-by times to supplementary uses and autonomy in all services. In the applications usual today, in which the user profile must consider not only voice and SMS but also e-mail, games, WAP and cameras, the power consumption of every utilization mode must be examined individually in order to determine the maximum time of use.

The processing power of a telephone operating with CDMA protocols (CDMA-2000 and WCDMA) depends strongly on the selection of the components located in the immediate vicinity of the antenna. The first component directly behind the antenna is the duplexer. It separates the phone’s transmit frequencies (output power up to about 1 W) from the incoming receive frequencies that have an input power in the region of only 10^-14 W at a minimum level of 110 dBm.

As the component located closest to the antenna, the duplexer has a major impact on the phone’s RF performance. In the past, the telephone specifications could be satisfied only with large ceramic duplexers. For physical reasons, these components had certain minimum dimensions in order to assure high performance. The introduction of SAW and BAW technologies has changed the situation: the requirements for significantly better electrical properties and compact dimensions can now be satisfied simultaneously.

Thus in recent years, developers were able to extend the operating time from three to six hours, the stand-by time from six to ten days, and the signal-to-noise ratio from 103 to 108 dB. At the same time, new features such as GPS, push-to-talk, integrated camera and more-compact overall dimensions required further miniaturization of the duplexers. The dimensions have now been reduced to a mere 3.0 x 2.5 x 0.6 mm³, whereas predecessor types required more than twice the area and a greater insertion height.

Progress thanks to integration

Whereas the digital part of the telephone is already highly integrated, the RF front end consists of many passive and active components mounted on a circuit board. At the end of the 1990s, developers integrated the antenna interface in a first step by using an antenna switch module (ASM) to switch between different frequency bands and modes. About three years ago, more highly integrated front-end modules (FEMs) began to replace the ASM. FEMs are a combination of SAW receive filters, ASM and several dozen components required for matching. Most of the substrates used to implement these modules are manufactured in low temperature co-fired ceramic (LTCC) technology.

Compared to the FR4 material or laminates, LTCC elements are high-precision, low-loss substrates that allow many passive components to be integrated in a compact volume. Thus diplexers, which separate the 1-GHz and 2-GHz frequency bands, can be implemented with minimum dimensions in LTCC technology.

The trend to a major reduction of insertion size—with simultaneously unchanged or even improved performance of the RF filters—requires precise software tools to simulate all relevant effects and interactions. It used to be sufficient to predict the acoustic behavior of the SAW chip. However, higher operating frequencies up to 2.5 GHz and specifications with tight framework conditions up to 6 GHz also require electromagnetic effects (EM) to be included in the considerations. The combination of precise acoustic simulation tools and EM software simulation packages enables designers to predict and optimize the performance of SAW filters as well as front-end modules on a SAW basis.

Today, mobile phones are lifestyle products whose focus of use is shifting from pure telephony to data and entertainment applications. In addition, increased user mobility requires the support of a greater number of communication standards and protocols.

SAW and BAW filters can handle these challenges by allowing a large number of band selection filters to be realized. Specific solutions here also include band-rejection filters, for example to prevent DVB reception in the GSM transmit band. Several filter functions can simultaneously be combined in a single package.

Improving the SAW design by precise modeling

Because performance requirements are rising greatly in line with the filter specifications, the simulations must describe the attenuation in the passband to within a few tenths of a dB and in the blocking band to within a few dB. The matching must be predicted at levels of 10 dB down to an accuracy of half a dB. Supplementary functions such as electrical separation and impedance conversion have now been integrated in the components and must also be simulated.

Taken together, the specifications and complexity of the components have developed to become an even greater challenge, whereas the time to market maturity of the product is simultaneously becoming ever shorter. That is why development engineers need the support of CAD systems to assure an efficient, fast and seamless design process. There is an additional need for simulation tools allowing the development of filter designs that can be reliably implemented with a minimum number of sample production runs (Fig. 1).

Two variants are essentially used today in modeling and simulation: specialized procedures such as COM (coupling of modes) and P-matrix models on the one hand, and generic tools such as the finite element or boundary element method on the other.

In contrast to acoustic modeling, the EM modeling of SAW components is a relatively new procedure that is used particularly in low-loss SAW filters in the GHz range.

	FIGURE 1: DESIGN ROUTINE
SAW components can be simultaneously modeled and optimized.

This is done with the aid of a P-matrix description of the SAW properties. This design flow can be used equally well for BAW components by replacing the SAW modeling with detailed three-dimensional BAW modeling of the electro-acoustic properties.

The great effectiveness of this method will be illustrated on the basis of a SAW duplexer for the new US mobile phone band at 850 MHz, all of whose terminals are referred to ground in unipolar mode and are matched to 50 Ω. Fig. 2 shows the transfer functions for the transmit and receive range. The duplexer is accommodated in a nine-pin SAW package of the CSSP (chip-sized SAW package) type and has dimensions of only 3.0 x 2.5 x 1.0 mm³.

	FIGURE 2: TRANSFER FUNCTIONS FOR THE TRANSMIT AND RECEIVE RANGE
Comparison of measurement (blue line), 2.5-D simulation (red line) and 3-D simulation (green line).

An accurate comparison shows that the simulation based on a three-dimensional EM description of the package shows better agreement with the measurements than one based on the 2.5-D description.

Duplexers as the key criterion

In view of the requirements on CDMA handsets, it is clear that improved duplexer performance plays a significant part in these communications systems.

Duplexers for the 800 MHz band are implemented largely in SAW technology, whereas PCS at 1900 MHz or WCDMA duplexers for 2100 MHz can be built in either SAW or BAW technology. Duplexers for the PCS band in particular have required substantial improvements in their SAW characteristics, such as a steeper transition from the passband to the stopband and improved temperature behavior.

The duplexer performance significantly affects the total performance of a mobile phone. Thus functions such as the simultaneous use of GPS on the basis of suitable transmit suppression at GPS frequencies or an extension of the band allocation can already be implemented.

The performance of the duplexers has a direct impact on the battery stand-by time in the mobile phones and thus on their operating time. This is due to the insertion loss, which must therefore be compensated for by amplifiers. Finally, duplexers also influence the effective range of the base stations via the receive insertion loss. The receive filters in GSM systems have a particularly great impact on the overall performance of the system. Receive filters with low insertion loss are of special significance for the sensitivity and stability of the receive chain, especially with the use of RF-ICs on a CMOS basis.

The resistive losses in the terminals at the acoustic level represent an important secondary effect in SAW components. This fact is of special importance in low-loss RF filters because the losses limit the attainable insertion loss. Fig. 3 compares two simulations of a conventional DMS track (dual mode SAW filter) with three transducers and a resonant series circuit at its input. The red dotted line was simulated with a resistance of 0 Ω, whereas the solid black line shows a simulation that includes the resistance of the metal fingers.

	FIGURE 3: COMPARING TWO SIMULATIONS OF A CONVENTIONAL DMS TRACK
Simulation of a DMS filter at 2 GHz with resistance (green line) and without resistance (red line)

The method used to improve these RF properties is defined by two steps: the first is to correctly model all the physical effects required to describe the loss mechanism (see Fig. 4). This is indispensable to allow the design improvements to provide the required results. In recent years, the use of these modeling methods has led to a significant improvement of the insertion loss as shown in Fig. 5. EPCOS has developed two new models for the resistive losses in the metal fingers. These new models take into account both inhomogeneous current distributions and the current flow between the fingers connected to the same distribution bus.

	FIGURE 4: MEASUREMENT AND SIMULATION OF A DMS FILTER
Measurement (blue line) and simulation of a DMS filter at 2 GHz with a simple (red line) and comprehensive (green line) model for the finger resistance at acoustic level

	FIGURE 5: DRAMATIC REDUCTION OF INSERTION LOSS
Development of the maximum insertion loss in the passband for DMS filters at 2 GHz

New packaging technology for further miniaturization

Fig. 6 shows that the filters within the RF section of a mobile phone take up a significant part of the space on the RF stage of its circuit board. The latest trends to integrate the filter functions in front-end modules—trends that accommodate the power amplifier, the switching function or the transceiver IC—increase the pressure on further miniaturization of the package.

	FIGURE. 6: STRUCTURE OF A DISCRETE RF FRONT END
Four SAW receive filters for GSM are accommodated in the immediate vicinity of the RF ASIC (circled in yellow).

The third generation of the CSSP technology developed by EPCOS for SAW and BAW filters is now available. It can be used to manufacture SAW filters with dimensions of only 1.4 x 1.1 x 0.4 mm³. Fig. 7 shows a cross-section of a flip-chip package in this technology.

	FIGURE 7: BASIC CROSS-SECTION OF A CSSP PACKAGE
Cross-section of a flip-chip package in CSSP technology

The chips are soldered onto a high temperature co-fired ceramic (HTCC) substrate with the aid of lead-free solder balls made of SnAgCu, each with a diameter of 100 µm. By laminating a polymer film onto the chips, a cavity is formed for the propagation of the surface acoustic waves. In order to ensure good adhesion between film and intermediate ceramic substrate as well as the chip and also to ensure a constant film geometry in volume production, EPCOS has developed a special high-temperature process that operates under high pressure. The film at the edge of the component is removed prior to metallization in order to prevent moisture from penetrating the package. The package is terminated with copper and nickel layers totaling 20 µm in thickness. The individual components are then separated in a dicing process.

Fig. 8 summarizes the recent progress made in the miniaturization of SAW and BAW filters. In the long term, technology developments will also open up additional ways of packaging SAW chips within the scope of a standard IC process.

Authors:

Ulrich Bauernschmitt: Executive Vice President Operations Front-End SAW

Christian Block: Vice President and Chief Technology Officer SAW Components

Dr. Clemens Ruppel: Vice President Research and Development Patents and Standards