Trends for Portable Wireless Applications

This article explores key issues related to the development and manufacture of commercial radios and discusses various trends associated with the commercial radio business. The paper focuses on the RF portion of the radio, and specifically on the requirements for cellular phones, digital pagers and other commercial communication applications. Device technologies that are addressed include silicon bipolar junction transistors (BJT), silicon MOSFETs, GaAs MESFETs, high electron mobility transistors (HEMT), and both III-V semiconductor-based and SiGe-based heterojunction bipolar transistors (HBT). Cost, manufacturability and cycle time constraints are discussed in terms of their effect on technical developments.

David Halchin
Rockwell Semiconductor Systems
Newbury Park, CA

Michael Golio
Rockwell Collins
Cedar Rapids, IA

While commercial wireless electronics has evolved from military systems, the general requirements and characteristics of these two applications differ greatly. Bandwidths for the commercial receive and transmit channels are extremely narrow in frequency as regulated by the appropriate regulatory commission of the government, for example, the Federal Communications Commission in the US. Typical volumes for portable communication products are very large, with sales on the order of millions per year as opposed to hundreds or thousands in a product lifetime. The cost of a commercial unit is typically priced in the tens or hundreds of dollars, while the military counterpart is usually many times this amount. In addition, minimizing the power consumption of a commercial portable product is of paramount importance, while this concern is not usually a primary driver for the corresponding military application.

The wireless market comprises an ever-expanding number of systems both in operation and in the planning stage. Most of these systems operate at frequencies less than 5 GHz. Table 1 lists some of these systems. The two products representing the largest volume applications are pagers operating at 200 and 900 MHz, and analog cellular phones operating near 900 MHz, that is, the Advanced Mobile Phone Service (AMPS). Another wireless system expected to be produced in large volume for the past five years is wireless local area networks (LAN).1 Once predicted to be the future of interconnectivity, this market has simply not materialized due mainly to the strong position of inexpensive coaxial-based systems. Finally, a system that is being pursued by a few companies is direct-to-satellite communication networks.

Table I: Commercial Wireless Products

Application

Frequency (GHz)

Cell phones
(AMPS, CDMA)

0.9

Pagers

0.2 and 0.9

GSM, PCN, DECT, PHS

0.9 to 2.4

Wireless LANs

2.4

Iridium

1.6

Mobile/Trunking systems

0.9

System Trends


Several general trends exist in system architecture that are present in most commercial wireless systems. The first of these trends is toward lower supply voltages and lower power consumption. Figure 1 shows the observed trends in battery voltage, from 12 V in the first portable radios to approximately 3 V 60 years later in modern phones. Alert engineers with management potential will note that the battery soon will be tossed away as the goal of 0 V is reached in the not-too-distant future. True visionaries will extrapolate beyond this zero voltage point to predict actual power generation from wireless products in the years that follow. In spite of such trends and predictions, the supply voltage probably will remain around 3 V for a few years due to the growing use of lithium ion batteries. This battery technology provides a solution that can be molded to form fit nearly any battery profile, can be recharged many times (no memory) and provides for fairly large energy storage.

Fig. 1: The supply voltage trend from 1930 to present.

A second important trend is increasing power-added efficiency (PAE) of power amplifiers (PA) even as the supply voltage decreases. Unfortunately, high efficiency at low voltages is difficult to obtain due to both the finite on-resistance of an active device reducing the output voltage swing, and the requirement for a steeper load line to maintain output power. A steep load line is not compatible with high efficiency operation since it requires a larger device current. This effect is shown in Figure 2 .

Fig. 2: Typical high efficiency load lines for 5.8 and 3.5 V supplies, and the 3.5 V load line to achieve the 5.8 V-equivalent output power.

Another system architecture trend is the move toward digital modulation schemes from the more standard analog system. This movement toward digital modulation schemes is anticipated to help alleviate the overcrowding of the 900 MHz cellular band in major metropolitan areas since digital systems allow for an increase in caller density. There is an added side benefit of digital modulation schemes in that security during phone operation comes automatically. However, digital phones require a marked increase in the linearity of the PA and, therefore, the output device. This increase in linearity is usually achieved by decreasing the efficiency of the PA. That is, to obtain good linearity, a device is biased toward class A operation. To maintain high efficiency, the device bias is class B or AB. Figure 3 shows the two different bias regions.

Fig. 3: A typical IV plane indicating the different regions for high efficiency vs. high linearity biasing.

Currently, a large increase is beginning in digital systems for phones in both Europe (Global System for Mobile communications (GSM)) and the Pacific Rim (personal digital cellular and personal handyphone system (PHS)). Alternatively, in the US, analog phone systems are still dominant and will remain that way for the next few years due to the massive infrastructure in place presently.

There is also a move to higher frequencies as the lower frequency bands become more cluttered. However, the majority of present and projected systems are still below 5 GHz. This modest increase in frequency still has an impact on the device design.

Another system designed to relieve overcrowding and to reduce power consumption is the microcell approach. This system comprises smaller cells than the current analog cell systems, thus requiring more base stations. However, at any given instant in time a phone would be much closer to a base station and therefore would require much less transmit power. This lower transmit power requirement results in a decrease in power consumption and, therefore, an increase in talk time. The microcell system, shown in Figure 4 , indicates the honeycomb pattern of a cell system and the inset of a microcell system.

Fig. 4: An analog cellular cell system showing the implementation of a microcell system.

A final trend for system architectures is the possibility of direct-to-satellite systems. Currently, several of these systems are under development with Motorola's IRIDIUMTM system likely to be the first operational system. Several concerns exist as to the practicality of a direct-to-satellite system for a handheld portable radio, including the large power requirements for the handset transmitter to contact a satellite and the low noise figure required for the front-end receiver of the phone to acquire the low level satellite signals. These requirements will greatly push the capabilities of the active devices in the phone.

The Low Voltage Trend


There are several reasons for the trend toward lower voltage batteries and lower power consumption in a portable product. One of the main reasons for this trend is that the consumer desires both longer talk and standby times, meaning that the power consumption of the individual components must be reduced to reduce the overall power consumption of the phone. Another motivation to reduce the phone's power supply is a direct result of the subscriber's desire for a phone that is compact, lightweight, and requires little or no maintenance (never need to change a battery). At present, the single largest volume and weight component in a phone is the battery. Therefore, a major impact on the mass and volume of the phone can be made by reducing its supply voltage and reducing power consumption simultaneously.

Device Issues


The device requirements for commercial radios vary significantly from system to system and from circuit to circuit. For example, digital pager requirements are distinct from cellular phone requirements. Similarly, the device requirements for an output stage power amplifier are very different from those of a front-end low noise amplifier (LNA). For this reason, there is no one device technology that can be used for all applications. Likewise, no rigid set of rules can be applied to determine the optimum device for every power amplifier, LNA, mixer or transfer switch. However, each device technology exhibits distinct properties that can be viewed as an advantage or disadvantage for each application. Similarly, certain general requirements and desires exist for each particular application. The process of listing such device properties and general requirements can aid both the radio and device designer.

Prior to examining how each device technology might be applied to the various circuit functions and consumer systems, it is useful to note some general issues. The primary driver for almost all commercial parts is cost. Device technologies that make use of straight forward, low cost processing and low cost materials have a decided competitive advantage over higher cost device technologies. In the commercial world, it is often preferable to offer performance identical to competing technologies at a lower cost than to offer improved performance at the same cost. This high premium on cost often gives silicon BJTs and MOSFETs an edge over III-V-based MESFETs, HEMTs or HBTs. III-V semiconductors are those devices with constituent elements that compose the materials in columns three and five on the periodic table, such as GaAs, InP or AlGaAs.

There is a desire among phone manufacturers to use only parts with single-polarity power supply requirements. The negative supply voltage requirement results in an additional power drain (to run the negative supply charge pump), additional cost and additional size. All of these additions are unwanted. For this reason, devices that require only a single-polarity supply voltage such as bipolar transistors (Si, SiGe and III-V HBTs), MOSFETs, or enhancement-mode MESFETs and HEMTs are preferred to parts that require a negative supply. Similarly, devices that do not require a drain switch between the battery and the PA have a potential cost advantage over those that do. This advantage favors bipolar transistors (Si, SiGe and III-V HBTs) and MOSFETs.

For small-signal front-end and driver amplifier applications, ease of integration, reduction of parasitic loss and tight parametric control of turn-on voltage are important properties for device technologies to possess. Because of the semi-insulating properties of GaAs and InP, III-V semiconductor devices hold a slight advantage over most silicon devices in terms of integration. Similarly, the favorable transport properties of III-V devices offer the potential for lower loss due to parasitic resistance. Bipolar technologies offer the advantage of tightly controlled turn-on voltage as compared to MOSFETs, MESFETs or HEMTs. This feature is an important consideration as supply voltages and power dissipation continue to be reduced in newer system designs. Bipolar transistors have a fundamental advantage over FETs in this respect because the base-to-emitter turn-on voltage Vbe is determined by band gap, which is essentially fixed over all process variations. In contrast, all FET technologies depend on control of doping densities and surface processing for control of the threshold voltage Vth .

For power applications, high breakdown voltage and reduced on-resistance are essential. In general, increasing critical breakdown voltages in all device technologies involves a reduction in some other desirable quantity so that no single device technology has a clear advantage. This situation could change in favor of III-V materials once commercial RF allocations move to higher bands (> 2 to 5 GHz). Low on-resistance favors III-V transistors due to higher low field mobility.

It is important to note that the specifications driving device choice are not issues that are monitored or controlled routinely during device manufacturing. In particular, linearity, stability, ruggedness and 1/f noise each can be the determining factor in the choice of which device is used in a particular radio application. With the exception of 1/f noise, there is little understanding of what device properties should be considered to optimize these device characteristics. In the case of 1/f noise, silicon BJTs are the clear performance leader.

In order to define a framework that can be used to discuss general device requirements, it is useful to consider the general block diagram of a commercial radio, shown in Figure 5 . In the case of portable pagers, only the receive side of the radio needs to be considered. This article focuses only on the RF portion of the radio.

Fig. 5: A commercial radio generic block diagram.

Depletion-mode MESFETs and HEMTs


Today, depletion-mode MESFETs and HEMTs provide the best PAE available for 1 to 3 W output power levels, 3 to 6 V supply voltages and frequency ranges greater than 800 MHz. This property alone has led to the displacement of silicon devices and the introduction of many commercially available high tier cellular phones incorporating GaAs MESFET power amplifier final stages. This situation exists even though these devices require a negative supply and a drain switch. However, competition from silicon RF MOSFETs has not subsided and GaAs parts have not been able to penetrate the lower tier radios, primarily for cost reasons.

GaAs MESFETs have also found their way into many radios that have RF transfer switch requirements. Many of the mobile/trunking systems are examples of such radios. In addition, some of the full-featured cellular phones are now available with a switched antenna port. The GaAs MESFET is an ideal candidate for this function once a negative bias supply is required for the radio.

As mentioned previously, an additional property of all III-V-based devices is their good integration capability. In practice, integration levels in most commercial phones are extremely low as hybrids outperform MMICs and can be less expensive. However, there is somewhat of a trend to change this situation in high tier radios and some integrated driver amplifiers and front-end electronics are appearing now in commercial products.

Emerging digital cellular systems are placing a premium on PA linearity. Although many device technologies are competing for these slots and no clear technical leader has yet emerged, depletion-mode MESFETs and HEMTs are strong contenders for this market. Poor 1/f noise has left depletion-mode MESFETs and HEMTs out of the design of oscillators for nearly all commercial radios. This situation is not likely to change in the near future.

Cost remains an issue with III-V-based devices and especially with heterostructure devices. This is a primary reason why essentially no use has been made of depletion-mode MESFETs and HEMTs in portable pagers today. The negative supply requirement and threshold control issues also work against depletion-mode parts for these applications. Table 2 lists a controversial summary of some of the critical characteristics of depletion-mode MESFETs and HEMTs, as well as other key device technologies.2

Table II: III-V Semiconductor Device Characteristics

Property

D-mode
FET/HEMT

E-mode
FET/HEMT

III-V HBT

Si BJT

SiGe HBT

MOSFET

Cost

moderate
and
decreasing

moderate
and
decreasing

high and
decreasing

low and
mature

moderate
and
decreasing

low to
moderate

Single
polarity
supply

no

yes

yes

yes

yes

yes

Integration
capability

excellent

excellent

excellent

OK at low
frequency

OK at low
frequency

OK at low
frequency

Parasitic
loss

very good

very good

very good

modest

modest

modest

modest

Turn-on
voltage
control

modest

modest

very good

excellent

excellent

good

 

PAE

excellent

OK at low
power

very good

poor

moderate

very good

 

Linearity
for digital
systems

very good
for epi-MESFET

N/A because
of power
capability

very good

N/A because
of PAE
deficiency

?

?

 

Enhancement-mode MESFETs and HEMTs


To overcome the negative supply requirement, enhancement-mode MESFETs and HEMTs are considered for many commercial radio applications. These parts share the same excellent integration capability and high PAE attributes as their depletion-mode counterparts, but are limited in maximum output power to relatively low levels. This limitation will affect which applications can be filled by the technology. Unfortunately, enhancement-mode MESFETs and HEMTs share the same cost concerns as their depletion-mode relatives.

Enhancement-mode parts have not been used effectively for most switch applications. This failure is due to inadequate isolation in the off mode as well as to the parts' power handling capability.

As advanced pagers with stiffer performance requirements and call back capability are introduced, enhancement-mode MESFETs and HEMTs will be considered for these radios. The high volume, low cost nature of this business will place severe demands on these technologies.

Ion-implanted vs. Epitaxially Grown MESFETs
Although most of the MESFET devices that have found their way into commercial products today are fabricated using ion implantation, epitaxially grown MESFETs offer significant potential as a commercial radio part. The traditional reasons for choosing ion implantation include cost as well as current and threshold control. By leveraging the significant advances made in epitaxial growth techniques over the past 15 to 20 years, epitaxial MESFETs can now be fabricated with cost and parametric control comparable to or better than implanted parts. Parametric control is gained using epitaxial buffer and etch stop layers. The improved control is translated into improved yield, which impacts cost directly. Further cost reduction is realized by the decrease in required processing steps (implant and anneal) and by size reduction, which can be realized due to higher obtainable current densities.

Epitaxial MESFET performance is improved significantly over performance achievable with ion-implanted devices in almost all respects. Because epitaxial material exhibits better transport properties, reduced damage and higher activation than implanted material, epitaxial MESFETs (epi-MESFET) achieve higher power and current densities. Improved transport also translates into reduced parasitic resistance. The ability to grow a back-side high band gap buffer layer provides epi-MESFETs with dramatically improved carrier confinement, which leads to superior transconductance and output conductance. Epi-MESFETs have also exhibited ultra-linear power amplification with extremely low adjacent-channel interference characteristics.3

Silicon BJTs
Silicon BJTs offer the lowest cost alternative for a large number of commercial radio applications. For this reason they have been used almost exclusively in portable pagers. BJTs have also dominated front-end electronics and VCOs for most radios below 2 GHz.

BJTs offer the greatest technology maturity of the device options available to radio designers. This property can be used to significant advantage during new product development since a greater knowledge of the process usually leads to a reduced number of required learning cycles. The only circuit application from which BJT technology has been essentially eliminated is final stage power amplifiers.

Although the integration capability of Si substrate material is inferior compared to III-V substrates, the bipolar complimentary metal-oxide semiconductor (BiCMOS) processes have been used to significant advantage for many applications. Integration of RF devices on silicon will become more difficult as application frequencies continue to increase. A dominance of commercial applications can be expected to continue until III-V-based transistors move further down the cost curve, or until the frequency limits of silicon BJTs are exceeded. However, establishing what that frequency limit is has been a difficult task as BJT technology continues to advance.

One method of extending silicon BJT performance is through the use of SiGe HBTs. The comments made regarding Si BJTs apply generally to SiGe HBTs with the critical distinctions that cost is added to the starting material, and the technology is not nearly as mature.

The key advantage of SiGe HBTs is the extended frequency performance. Although this feature may play an important role in radios with frequency allocations greater than 2 to 5 GHz, it is not critical to today's commercial systems.

Silicon MOSFETs
RF MOSFETs are the major competitor to GaAs MESFETs for high tier PA applications and tend to dominate mid and low tier power circuitry. The MOSFETs have the advantages of requiring no negative supply and no drain switch. In addition, starting silicon wafer cost is significantly lower than that of GaAs. The challenge to the advanced MOSFET device designer is to contain processing costs, reduce chip size and ensure that an overall device cost advantage is maintained. Scaling MOSFETs to operate as power devices with a supply of less than 3 V is also a challenge.

RF circuit integration is not as easy for MOSFET technologies as for GaAs, but has been accomplished. As frequencies increase, the GaAs integration advantage becomes more difficult to overcome.

III-V HBTs
Of all the device technologies discussed, III-V HBTs are the least mature. This fact alone is why the devices are not more prevalent in the marketplace, as III-V HBTs offer a number of potential technical advantages. First, the devices require neither a negative supply nor a drain switch. Because the substrates are III-V semiconductor based, the integration capability is excellent. PAE values that have been demonstrated by HBTs are competitive and there is some indication that HBTs offer improved linear power properties. This improved linear power could be important in the development of emerging digital cellular systems. III-V HBTs' 1/f noise performance is superior to that of a III-V FET, while their noise figure is higher.

Modeling Issues


The cost of the modeling effort affects only the development cycle of commercial products with essentially no impact on recurring costs. For high volume products, nonrecurring costs are insignificant so that modeling capability does not have an important direct impact on commercial product costs. However, the availability of accurate models does have a significant impact on product development cycle time. It is this cycle time that is often the critical distinguishing feature in the competition for commercial products. Clearly, mature technologies with complete, accurate, validated model libraries have a significant advantage over emerging technologies. Specific modeling issues do exist for every device technology (from inadequacies of the Gummel-Poon model for Si BJTs to the complete absence of built-in models developed specifically for HBTs). Despite this fact, it is the absence of an adequate effort by device manufacturers to utilize existing characterization, parameter extraction and modeling methodologies that is often the most severe issue facing designers. The commercial semiconductor manufacturer who invests in a well-focused and up-front modeling effort will have a clear competitive advantage over those who do not.

Measurement Issues


Several issues exist regarding the measurement of devices and circuits utilized in wireless systems. One issue involves the measurement of devices designed to operate at very low currents, such as those in a typical receiver application. These devices usually operate at DC currents much less than 1 mA. The output power at the test set ports of a network analyzer with a typical calibration is on the order of -20 dBm. This power level produces an RF current in a device that is comparable to the device's DC bias current. In fact, for a device operating in the hundreds of microamps, the input power level at the device plane needs to be on the order of -40 dBm or less for the measurement to be noninvasive. Of course, this input power level reduces the dynamic range of the measurement system and thus reduces the measurement accuracy.

One of the most common measurements performed on a power device is load pull. For a given bias point, this measurement is used to determine the device's optimal match condition for a predetermined parameter (for example, maximum output power). In this measurement, the load impedance presented to the device is varied and the device performance is measured at each of these impedances with the device input usually matched for maximum power transfer. A set of contour lines of a constant parameter can then be plotted to determine an optimal load state for that parameter (for example, output power). There are several problems with using this approach to determine the best match conditions. Due to poor scaling between wafer-level and power devices, the load pull should be performed on a device sized for a particular application. For a wireless PA, a typical power level is 1 to 2 W. This PA level is in contrast to a wafer-level measurement that is limited to a few hundred milliwatts for thermal considerations. Therefore, the load pull measurement usually is performed on a packaged part mounted in a test fixture. Measuring a packaged device requires a good model for the fixture/package combination, which is not always easy to obtain. Load pull measurements can be extremely time consuming and need to be repeated for every potential quiescent bias point. Finally, the impedances presented to both the input and the output of the device at the fundamental and harmonic frequencies can greatly affect the device's output power, efficiency and linearity. While all load pull systems control the impedance at the fundamental frequency, most do not control the harmonic terminations. This means that as a load pull is performed, the harmonic impedances are varying from load state to load state, thus confounding the interpretation of the data.

Harmonic terminations also greatly impact the results obtained from linearity measurements. These results include intermodulation distortion, adjacent-channel power and noise power ratio. Since all of these results are a measure of a device's deviation from linear behavior (fundamental only), the termination of the generated nonlinear energy (harmonics) directly affects the measurement outcome. This condition makes it extremely difficult to perform a unique measurement and, therefore, to determine the best match conditions for a given application. Also, the harmonic effects mean that comparing measurements performed on various device technologies at different locations is very questionable, which again raises the question of how the design engineer can determine which device technology offers the best linearity.

Another measurement required for a PA is the determination of a device's ruggedness and large-signal stability. This measurement involves the intentional mismatch (a large SWR) of the output of a PA while operating under full RF power. The object is to determine if the PA can survive this stress and remain reasonably stable during stress. This test is usually performed manually and, therefore, is time consuming. In addition, the procedure tends to be a survival test and does not provide a great deal of information as to what caused the device to fail or to oscillate.

A final measurement issue concerns the volume RF testing of devices and circuits. Many phone manufacturers require full RF testing of the individual devices in the output and receiver prior to acceptance. This testing could involve either the complete receiver or transmitter in the case of a MMIC or hybrid module, or the individual discrete components in the case of a chip and wire approach. This testing tends to be very time consuming and expensive in a rack and stack system, or, alternatively, very expensive in a high speed test system as the tester and handler costs often exceed $1 M. This testing can add a significant cost to each part, which easily can be a large fraction of the processing costs. However, the cost of a high speed tester will decrease as the volume increases, thus making it the best choice for large volume, high throughput manufacturing.

Conclusion


A broad range of systems and applications defines the commercial radio business today. For each system there are several conflicting constraints and emerging trends affecting the technology decisions related to the design and manufacture of these parts. In addition, each application has different performance requirements that make selection of an ultimate device technology impossible. Further, both the systems and the devices continue to evolve so that the appropriate device selection for today's radio may not be optimum tomorrow. In this confusing and changing technical environment, device and radio engineers must constantly re-examine their device selections for each application.

Acknowledgment


This paper was presented originally at the 1996 Wireless Workshop, which was held September 29-October 2, 1996 in Sedona, AZ.

References

  1. Ali, Golio and Maki, "Cost-effective MMICs for WLANs: Fact or Fantasy?" 1995 International Microwave Symposium Panel Session, May 1995.
  2. Michael Golio, "Device and Material Technologies for Commercial Communications," Proceedings of the 1995 International Topical Meeting on Nomadic Microwave Technologies and Techniques for Mobile Communications and Detection , Arcachon, France, November 1995, pp. 3Ð9.
  3. Golio and El-Ghazaly, "Emerging RF Issues for Modern Personal Commercial Communications Systems: Power Amplifiers for Portable Cellular Applications," Thirteenth National Radio Science Conference (NRSC 96), Cairo, Egypt, March 1996.