RF and Microwave Power Amplifier and Transmitter Technologies

Home , Ampliphase, Linear amplifier, RF power amplifier, Transmitter

RF and Microwave Power Amplifier and Transmitter Technologies — Part 1

By Frederick H. Raab, Peter Asbeck, Steve Cripps, Peter B. Kenington, Zoya B. Popovic, Nick Pothecary, John F. Sevic and Nathan O. Sokal

F and microwave lope tracking, outphasing, and Doherty. With this issue, we begin a power amplifiers Linearity can be improved through techniques four-part series of articles Rand transmitters such as feedback, feedforward, and predistor- that offer a comprehensive are used in a wide variety tion. Also discussed are some recent develop- overview of power amplifier of applications including ments that may find use in the near future. technologies. Part 1 covers wireless communication, A power amplifier (PA) is a circuit for con- the key topics of amplifier jamming, imaging, radar, verting DC input power into a significant linearity, efficiency and and RF heating. This amount of RF/microwave output power. In available RF power devices article provides an intro- most cases, a PA is not just a small-signal duction and historical amplifier driven into saturation. There exists background for the subject, and begins the a great variety of different power amplifiers, technical discussion with material on signals, and most employ techniques beyond simple linearity, efficiency, and RF-power devices. At linear amplification. the end, there is a convenient summary of the A transmitter contains one or more power acronyms used—this will be provided with all amplifiers, as well as ancillary circuits such as four installments. Author affiliations and con- signal generators, frequency converters, mod- tact information are also provided at the end ulators, signal processors, linearizers, and of each part. power supplies. The classic architecture employs progressively larger PAs to boost a 1. INTRODUCTION low-level signal to the desired output power. The generation of significant power at RF However, a wide variety of different architec- and microwave frequencies is required not tures in essence disassemble and then only in wireless communications, but also in reassemble the signal to permit amplification applications such as jamming, imaging, RF with higher efficiency and linearity. heating, and miniature DC/DC converters. Modern applications are highly varied. Each application has its own unique require- Frequencies from VLF through millimeter ments for frequency, bandwidth, load, power, wave are used for communication, navigation, efficiency, linearity, and cost. RF power can be and broadcasting. Output powers vary from 10 generated by a wide variety of techniques mW in short-range unlicensed wireless sys- using a wide variety of devices. The basic tems to 1 MW in long-range broadcast trans- techniques for RF power amplification via mitters. Almost every conceivable type of mod- classes A, B, C, D, E, and F are reviewed and ulation is being used in one system or anoth- illustrated by examples from HF through Ka er. PAs and transmitters also find use in sys- band. Power amplifiers can be combined into tems such as radar, RF heating, plasmas, laser transmitters in a similarly wide variety of drivers, magnetic-resonance imaging, and architectures, including linear, Kahn, enve- miniature DC/DC converters.

This series of articles is an expanded version of the paper, “Power Amplifiers and Transmitters for RF and Microwave” by the same authors, which appeared in the the 50th anniversary issue of the IEEE Transactions on Microwave Theory and Techniques, March 2002. © 2002 IEEE. Reprinted with permission.

22 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

No single technique for power generated by a single alternator. One PAs to operate over two decades of amplification nor any single trans- such transmitter (SAQ) remains bandwidth without tuning. Because mitter architecture is best for all operable at Grimeton, Sweden. solid-state devices are temperature- applications. Many of the basic tech- sensitive, bias stabilization circuits niques that are now coming into use Vacuum Tubes were developed for linear PAs. It also were devised decades ago, but have With the advent of the DeForest became possible to implement a vari- only recently been made practical audion in 1907, the thermoionic vac- ety of feedback and control tech- because of advances in RF-power uum tube offered a means of elec- niques through the variety of op- devices and supporting circuitry such tronically generating and controlling amps and ICs. as digital signal processing (DSP). RF signals. Tubes such as the RCA Solid-state RF-power devices UV-204 (1920) allowed the transmis- were offered in packaged or chip 2. HISTORICAL DEVELOPMENT sion of pure CW signals and facilitat- form. A single package might include The development of RF power ed the transition to higher frequen- a number of small devices. Power out- amplifiers and transmitters can be cies of operation. puts as high as 600 W were available divided into four eras: Younger readers may find it con- from a single packaged push-pull venient to think of a vacuum tube as device (MRF157). The designer basi- Spark, Arc, and Alternator a glass-encapsulated high-voltage cally selected the packaged device In the early days of wireless com- FET with heater. Many of the con- that best fit the requirements. How munication (from 1895 to the mid cepts for modern electronics, includ- the transistors were made was 1920s), RF power was generated by ing class-A, -B, and -C power ampli- regarded as a bit of sorcery that spark, arc, and alternator techniques. fiers, originated early in the vacuum- occurred in the semiconductor houses The original RF-power device, the tube era. PAs of this era were charac- and was not a great concern to the spark gap, charges a capacitor to a terized by operation from high volt- ordinary circuit designer. high voltage, usually from the AC ages into high-impedance loads and mains. A discharge (spark) through by tuned output networks. The basic Custom/Integrated Transistors the gap then rings the capacitor, tun- circuits remained relatively un- The late 1980s and 1990s saw a ing inductor, and antenna, causing changed throughout most of the era. proliferation variety of new solid- radiation of a damped sinusoid. Vacuum tube transmitters were state devices including HEMT, Spark-gap transmitters were rela- dominant from the late 1920s pHEMT, HFET, and HBT, using a tively inexpensive and capable of through the mid 1970s. They remain variety of new materials such as InP, generating 500 W to 5 kW from LF to in use today in some high power SiC, and GaN, and offering amplifica- MF [1]. applications, where they offer a rela- tion at frequencies to 100 GHz or The arc transmitter, largely tively inexpensive and rugged means more. Many such devices can be oper- attributed to Poulsen, was a contem- of generating 10 kW or more of RF ated only from relatively low voltages. porary of the spark transmitter. The power. However, many current applications arc exhibits a negative-resistance need only relatively low power. The characteristic which allows it to oper- Discrete Transistors combination of digital signal process- ate as a CW oscillator (with some Discrete solid state RF-power ing and microprocessor control allows fuzziness). The arc is actually extin- devices began to appear at the end of widespread use of complicated feed- guished and reignited once per RF the 1960s with the introduction of sil- back and predistortion techniques to cycle, aided by a magnetic field and icon bipolar transistors such as the improve efficiency and linearity. hydrogen ions from alcohol dripped 2N6093 (75 W HF SSB) by RCA. Many of the newer RF-power into the arc chamber. Arc transmit- Power MOSFETs for HF and VHF devices are available only on a made- ters were capable of generating as appeared in 1974 with the VMP-4 by to-order basis. Basically, the designer much as 1 MW at LF [2]. Siliconix. GaAs MESFETs introduced selects a semiconductor process and The alternator is basically an AC in the late 1970s offered solid state then specifies the size (e.g., gate generator with a large number of power at the lower microwave fre- periphery). This facilitates tailoring poles. Early RF alternators by Tesla quencies. the device to a specific power level, as and Fessenden were capable of oper- The introduction of solid-state well as incorporating it into an RFIC ation at LF, and a technique devel- RF-power devices brought the use of or MMIC. oped by Alexanderson extended the lower voltages, higher currents, and operation to LF [3]. The frequency relatively low load resistances. 3. LINEARITY was controlled by adjusting the rota- Ferrite-loaded transmission line The need for linearity is one of the tion speed and up to 200 kW could be transformers enabled HF and VHF principal drivers in the design of

24 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 1 · SRRC data pulses. Figure 2 · RF waveforms for SRRC and multicarrier signals. modern power amplifiers. Linear tively low data rates and a relatively Depending on the application, the amplification is required when the uncrowded spectrum. signals can have different ampli- signal contains both amplitude and Modern digital signals such as tudes, different modulations, and phase modulation. It can be accom- QPSK or QAM are typically generat- irregular frequency spacing. plished either by a chain of linear ed by modulating both I and Q sub- In a number of applications PAs or a combination of nonlinear carriers. The requirements for both including HF modems, digital audio PAs. Nonlinearities distort the signal high data rates and efficient utiliza- broadcasting, and high-definition being amplified, resulting in splatter tion of the increasingly crowded spec- television, it is more convenient to into adjacent channels and errors in trum necessitates the use of shaped use a large number of carriers with detection. data pulses. The most widely used low data rates than a single carrier Signals such as CW, FM, classical method is based upon a raised-cosine with a high data rate. The motiva- FSK, and GMSK (used in GSM) have channel spectrum, which has zero tions include simplification of the constant envelopes (amplitudes) and intersymbol interference during modulation/demodulation hardware, therefore do not require linear ampli- detection and can be made arbitrari- equalization, and dealing with multi- fication. Full-carrier amplitude mod- ly close to rectangular [4]. A raised- path propagation. Such Orthogonal ulation is best produced by high level cosine channel spectrum is achieved Frequency Division Multiplex amplitude modulation of the final RF by using a square-root raised-cosine (OFDM) techniques [5] employ carri- PA. Classic signals that require lin- (SRRC) filter in both the transmitter ers with the same amplitude and ear amplification include single side- and receiver. The resultant SRRC modulation, separated in frequency band (SSB) and vestigal-sideband data pulses (Figure 1) are shaped so that modulation products from one (NTSC) television. Modern signals somewhat like sinc functions which carrier are zero at the frequencies of that require linear amplification are truncated after several cycles. At the other carriers. include shaped-pulse data modula- any given time, several different data Regardless of the characteristics tion and multiple carriers. pulses contribute to the I and Q mod- of the individual carriers, the resul- ulation waveforms. The resultant tant composite signal (Figure 2) has Shaped Data Pulses modulated carrier (Figure 2) has simultaneous amplitude and phase Classic FSK and PSK use abrupt simultaneous amplitude and phase modulation. The peak-to-average frequency or phase transitions, or modulation with a peak-to-average ratio is typically in the range of 8 to equivalently rectangular data pulses. ratio of 3 to 6 dB. 13 dB. The resultant RF signals have constant amplitude and can therefore be Multiple Carriers and OFDM Nonlinearity amplified by nonlinear PAs with good Applications such as cellular base Nonlinearities cause imperfect efficiency. However, the resultant stations, satellite repeaters, and reproduction of the amplified signal, sinc-function spectrum spreads sig- multi-beam “active-phased-array” resulting in distortion and splatter. nal energy over a fairly wide band- transmitters require the simultane- Amplitude nonlinearity causes the width. This was satisfactory for rela- ous amplification of multiple signals. instantaneous output amplitude or

26 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

EDGE ACPR Offsets I The traditional measure of linearity is the carrier-to- ERROR VECTOR intermodulation (C/I) ratio. The PA is driven with two or more carriers (tones) of equal amplitudes.

ACTUAL VECTOR Nonlinearities cause the produc- INTENDED VECTOR

Normalized Magnitude (dB) tion of intermodulation products at Q Frequency Offset from Carrier (MHz) frequencies corresponding to sums Figure 3 · ACPR offsets and bandwidths. Figure 4 · Error vector. and differences of multiples of the carrier frequencies envelope to differ in shape from the act upon the instantaneous signal [6]. The amplitude of the third-order corresponding input. Such nonlinear- voltage or envelope. However, memo- or maximum intermodulation distor- ities are the variable gain or satura- ry effects can also occur in high- tion (IMD) product is compared to tion in a transistor or amplifier. power PAs because of thermal effects that of the carriers to obtain the C/I. Amplitude-to-phase conversion is a and charge storage. Thermal effects A typical linear PA has a C/I of 30 dB phase shift associated with the signal are somewhat more noticeable in III- or better. amplitude and causes the introduc- V semiconductors because of lower Noise-Power Ratio (NPR) is a tra- tion of unwanted phase modulation thermal conductivity, while charge- ditional method of measuring the lin- into the output signal. Amplitude-to- storage effects are more prevalent in earity of PAs for broadband and phase conversion is often associated overdriven BJT PAs. noise-like signals. The PA is driven with voltage-dependent capacitances with Gaussian noise with a notch in in the transistors. While imperfect Measurement of Linearity one segment of its spectrum. frequency response also distorts a Linearity is characterized, mea- Nonlinearities cause power to appear signal, it is a linear process and sured, and specified by various tech- in the notch. NPR is the ratio of the therefore does not generate out-of- niques depending upon the specific notch power to the total signal power. band signals. signal and application. The linearity Adjacent Channel Power Ratio Amplitude nonlinearity and of RF PAs is typically characterized (ACPR) characterizes how nonlinear- amplitude-to-phase conversion are by C/I, NPR, ACPR, and EVM ity affects adjacent channels and is described by transfer functions that (defined below). widely used with modern shaped- pulse digital signals such as NADC STANDARD Offset 1 Offset 2 BW Integration EVM and CDMA. Basically, ACPR is the (kHz) Filter (peak/rms) ratio of the power in a specified band NADC [13] ±30 kHz ±60 kHz 32.8 kHz RRC 25%/12% outside the signal bandwidth to the –26 dBc –45 dBc α=0.35 rms power in the signal (Figure 3). PHS [14] ±600 kHz ±900 kHz 37.5 kHz RRC 25%/12% In some cases, the actual power spec- –50 dBc –55 dBc α=0.50 trum S(f) is weighted by the frequen- EDGE [15] ±400 kHz ±600 kHz 30 kHz None 22%/7.0% cy response H(f) of the pulse-shaping –58 dBc –66 dBc filter; i.e. (eq. 1) TETRA [16] 25 kHz 50 kHz 25 kHz RRC 30%/10%

–60 dBc –70 dBc α=0.35 −+ ffBWco / 2 2 IS-95 CDMA [17] 885 kHz 1980 kHz 30 kHz None N/A ∫ Hf( ) Sfdf( ) –45 dBc –55 dBc ffBW−− / 2 ACPR = co lower fU W-CDMA (3G-PP) 5.00 MHz 10.0 MHz 4.68 MHz RRC 25%/N/A 2 ∫ Hf( ) Sfdf( ) [18] –33 dB –43 dB α=0.22 fL Table 1 · ACPR and EVM requirements of various systems.

28 High Frequency Electronics Figure 5 · Envelope PDFs. Figure 6 · Power-output PDFs.

where fc is the center frequency, B is i.e. (Pout – PDR)/Pin. PAE gives a rea- bution function (CDF), which gives the bandwidth, fo is the offset, and fL sonable indication of PA performance the probability that the envelope and fU are the band edges. The when gain is high; however, it can does not exceed a specified ampli- weighting, frequency offsets, and become negative for low gains. An tude. CW, FM, and GSM signals have required ACPRs vary with applica- overall efficiency such as Pout/(Pin + constant envelopes and are therefore tion as shown in Table 1. ACPR can PDR) is useable in all situations. This always at peak output. SRRC data be specified for either upper or lower definition can be varied to include modulation produces PDFs that are sideband. In many cases, two differ- driver DC input power, the power concentrated primarily in the upper ent ACPRs for two different frequen- consumed by supporting circuits, and half of the voltage range and have cy offsets are specified. ACPR2, based anything else of interest. peak-to-average ratios on the order of upon the outer band, is sometimes 3 to 6 dB. Multiple carriers [8] pro- called “Alternate Channel Power Average Efficiency duce random-phasor sums much like Ratio.” The instantaneous efficiency is random noise and therefore have Error Vector Magnitude (EVM) is the efficiency at one specific output Rayleigh-distributed envelopes; i.e., a convenient measure of how nonlin- level. For most PAs, the instanta- ξ ξ earity interferes with the detection neous efficiency is highest at the p(E) = 2E exp(–V2 ) (4) process. EVM is defined (Figure 4) as peak output power (PEP) and the distance between the desired and decreases as output decreases. Peak-to-average ratio ξ is typically actual signal vectors, normalized to a Signals with time-varying ampli- between 6 and 13 dB. fraction of the signal amplitude. tudes (amplitude modulation) there- The average input and output Often, both peak and rms errors are fore produce time-varying efficien- powers are found by integrating the specified (Table 1). cies. A useful measure of performance product of their variation with ampli- is then the average efficiency, which tude and the PDF of the envelope. 4. EFFICIENCY is defined [7] as the ratio of the aver- Two cases are of special interest. Efficiency, like linearity, is a criti- age output power to the average DC- When the DC input current is con- cal factor in PA design. Three defini- input power: stant (class-A bias), the DC input tions of efficiency are commonly used. power is also constant. The average η η ξ Drain efficiency is defined as the AVG = PoutAVG/PinAVG (3) efficiency is then PEP/ . If the DC ratio of RF output power to DC input input current (hence power) is pro- power: This concept can be used with any portional to the envelope (as in class- of the three definitions of efficiency. B), the average efficiency is (4/π)1/2 η η = Pout/Pin (2) The probability-density function PEP, for a Rayleigh-distributed sig- (PDF) of the envelope gives the rela- nal. Thus for a multicarrier signal Power-added efficiency (PAE) tive amount of time an envelope with a 10 dB peak-to-average ratio, incorporates the RF drive power by spends at various amplitudes (Figure ideal class-A and B PAs with PEP subtracting it from the output power; 5). Also used is the cumulative distri- efficiencies of 50 and 78.5 percent,

May 2003 29 High Frequency Design RF POWER AMPLIFIERS respectively, have average efficiencies are npn or n-channel types because which is very useful in switching- of only 5 and 28 percent, respectively. the greater mobility of electrons (vs. mode operation with reactive loads. holes) results in better operation at Vertical RF power MOSFETs are Back-Off higher frequencies. useable through VHF and UHF. The need to conserve battery Gemini-packaged devices can deliver power and to avoid interference to Bipolar Junction Transistor (BJT) up to 1 kW at HF and 100s of watts other users operating on the same The Si BJT is the original solid- at VHF.VMOS devices typically oper- frequency necessitates the transmis- state RF power device, originating in ate from 12, 28, or 50-V supplies, sion of signals whose peak ampli- the 1960s. Since the BJT is a vertical although some devices are capable of tudes well below the peak output device, obtaining a high collector- operation from 100 V or more. power of the transmitter. Since peak breakdown voltage is relatively sim- power is needed only in the worst- ple and the power density is very Laterally Diffused MOS (LDMOS) case links, the “back-off” is typically high. Si BJTs typically operate from LDMOS is especially useful at in the range of 10 to 20 dB. 28 V supplies and remain in use at UHF and lower microwave frequen- For a single-carrier mobile trans- frequencies up to 5 GHz, especially in cies because direct grounding of its mitter, back-off rather than envelope high-power (1 kW) pulsed applica- source eliminates bond-wire induc- PDF is dominant in determining the tions such as radar. While Si RF tance that produces negative feed- average power consumption and power devices have higher gain at back and reduces gain at high fre- average efficiency. The PDF of the high frequencies, their fundamental quencies. This also eliminates the transmitting power (Figure 4) properties are basically those of ordi- need for the BeO insulating layer depends not only upon the distance, nary bipolar transistors. The positive commonly used in other RF-power but also upon factors such as attenu- temperature coefficient of BJTs tends MOSFETs. ation by buildings, multipath, and to allow current hogging, hot-spot- LDMOS devices typically operate orientation of the mobile antenna [8], ting, and thermal runaway, necessi- from 28-V supplies and are currently [9], [10]. To facilitate prediction of the tating carefully regulated base bias. available with power outputs of 120 power consumption, the envelope and Since RF power BJTs are generally W at 2 GHz. They are relatively low back-off PDFs can be combined [11]. composed of multiple, small BJTs in cost compared to other devices for (emitter sites), emitter ballasting this frequency range and are current- 5. RF POWER TRANSISTORS (resistance) is generally employed to ly the device of choice for use in high- A wide variety of active devices is force even division of the current power transmitters at 900 MHz and 2 currently available for use in RF- within a given package. GHz. power amplifiers, and RF-power transistors are available in packaged, Metal-Oxide-Silicon Field-Effect Junction FET (JFET) die, and grown-to-order forms. Transistor (MOSFET) JFETs for power applications are Packaged devices are used at fre- MOSFETs are constructed with often called Static Induction quencies up to X band, and are domi- insulated gates. Topologies with both Transistors (SITs). Impressive power nant for high power and at VHF and vertical and later current flow are and efficiency have been obtained lower frequencies. A given package used in RF applications, and most are from RF JFETs based upon Si, SiGe, can contain one or more die connect- produced by a double-diffusion pro- and SiC at frequencies through UHF. ed in parallel and can also include cess. Because the insulated gate con- However, the JFET has never become internal matching for a specific fre- ducts no DC current, MOSFETs are as popular as other RF-power FETs. quency of operation. Dice (chips) can very easily biased. be wire-bonded directly into a circuit The negative temperature coeffi- GaAs MEtal Semiconductor FET to minimize the effects of the package cient of a MOSFET causes its drain (GaAs MESFET) and are used up to 20 GHz. In current to decrease with tempera- GaAs MESFETs are JFETs based MMICs, the RF-power device is ture. This prevents thermal runaway upon GaAs and a Schottky gate junc- grown to order, allowing its size and and allows multiple MOSFETs to be tion. They have higher mobility than other characteristics to be optimized connected in parallel without ballast- do Si devices and are therefore capa- for the particular application. This ing. The absence of base-charge stor- ble of operating efficiently at higher form of construction is essential for age time allows fast switching and frequencies. GaAs MESFETs are upper-microwave and millimeter- also eliminates a mechanism for sub- widely used for the production of wave frequencies to minimize the harmonic oscillation. An overdriven microwave power, with capabilities of effects of strays and interconnects. (saturated) MOSFET can conduct up 200 W at 2 GHz and 40 W at 20 Virtually all RF power transistors drain current in either direction, GHz in packaged devices. These

30 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS devices have relatively low break- although technically an “HFET” has and efficiency than the GaAs down voltages compared to a doped channel that provides the pHEMT, with the PA efficiency begin- MOSFETs or JFETs and are typical- carriers (instead of the heterojunc- ning to drop at 60 GHz. However, it ly operated from supply voltages tion). The acronyms “HFET and has a lower breakdown voltage (typi- (drain biases) of 5 to 10 V. Most “HJFET” (HeteroJunction FET) cally 7 V) and must therefore be oper- MESFETs are depletion-mode appear to be used interchangeably. ated from a relatively low drain-volt- devices and require a negative gate GaAs HEMTs/HFETs with fT as age supply (e.g., 2 V). This results in bias, although some enhance-mode high as 158 GHz are reported. PAs lower output per device and possibly devices that operate with a positive based upon these HEMTs exhibit 15- loss in the combiners required to bias have been developed. Linearity W outputs at 12 GHz with a power- achieve a specified output power. is often poor due to input capacitance added efficiency (PAE) of 50 percent. Nonetheless, the InP HEMT general- variation with voltage; the output Outputs of 100 W are available at S ly has a factor-of-two efficiency capacitance is also often strongly band from packaged devices. advantage over the pHEMT and bias- and frequency-dependent. GaAs HEMT. Pseudomorphic HEMT InP HEMTs have been fabricated Heterojunction FET (HFET) / High- The pseudomorphic HEMT with fmax as high as 600 GHz (0.1 Electron-Mobility Transistor (HEMT) (pHEMT) further improves upon the µm gate length), and amplification HFETs and HEMTs improve upon basic HEMT by employing an has been demonstrated at frequen- the MESFET geometry by separating InGaAs channel. The increased cies as high as 190 GHz. The efficien- the Schottky and channel functions. mobility of In with respect to GaAs cy does not begin to drop until about Added to the basic MESFET struc- increases the bandgap discontinuity 60 GHz. Power levels range from 100 ture is a heterojunction consisting of and therefore the number of carriers to 500 mW per chip. an n-doped AlGaAs Schottky layer, in the two-dimensional electron gas. an undoped AlGaAs spacer, and an The lattice mismatch between the Metamorphic HEMT (mHEMT) undoped GaAs channel. The disconti- GaInAs channel and GaAs substrate The mHEMT allows channels nuity in the band gaps of AlGaAs and is also increased, however, and this with high-In content to be built on GaAs causes a thin layer of electrons limits the In content to about 22 per- GaAs substrates. The higher electron (“two-dimensional electron gas or 2- cent. mobility and higher peak saturation DEG”) to form below the gate at the The efficiency of PAs using velocity result in higher gain than is interface of the AlGaAs and GaAs pHEMTs does not begin to drop until possible in a pHEMT. mHEMTs are layers. Separation of the donors from about 45 GHz and pHEMTs are use- generally limited to low-power appli- the mobile electrons reduces colli- able to frequencies as high as 80 cations by their relatively low break- sions in the channel, improving the GHz. Power outputs vary from 40 W down voltage (<3 V). However, an mobility, and hence high-frequency at L band to 100 mW at V band. mHEMT capable of 6-V operation response, by a factor of about two. While pHEMTs are normally grown and a power output of 0.5 W has been AlGaAs has crystal-lattice proper- to order, a packaged device pHEMT recently reported. ties similar to those of GaAs, and this has recently become available. makes it possible to produce a poten- Heterojunction Bipolar Transistor tial difference without lattice stress. InP HEMT (HBT) The GaAs buffer contributes to a rel- The InP HEMT places an HBTs are typically based upon atively high breakdown voltage. AlInAs/GaInAs heterojunction on an the compound-semiconductor materi- Their fabrication employs advanced InP substrate. The lattices are more al AlGaAs/GaAs. The AlGaAs emitter epitaxial technologies (Molecular closely matched, which allows an In is made as narrow as possible to min- Beam Epitaxy or Metal Organic content of up to about 53 percent. imize base resistance. The base is a Chemical Vapor Deposition) which This results in increased mobility, thin layer of p GaAs. The barrier is tends to increase their cost. which in turn results in increased created by heterojunction (AlGaAs/ The GaAs HEMT is known in the electron velocity, increased conduc- GaAs) rather than the doping. The literature by a wide variety of differ- tion-band discontinuity, increased base can therefore be doped heavily ent names, including MODFET two-dimensional electron gas, and to minimize its resistance. Base sheet (Modulation-Doped FET), TEGFET higher transconductance. The ther- resistance is typically two orders of (Two-dimensional Electron-Gas mal resistance is 40 percent lower magnitude lower than that of an FET), and SDFET (Selectively Doped than that of a comparable device ordinary BJT, and the frequency of FET). It is also commonly called an built on a GaAs substrate. operation is accordingly higher. The HFET (Heterostructure FET), The InP HEMT has higher gain current flow is (in contrast to a MES-

32 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

FET) vertical so that surface imper- power densities of 10 W/mm, which is Radio Transmitters and Carrier fections have less effect upon perfor- ten times that of a GaAs MESFET. Currents, Scranton PA: International mance. The use of a semi-insulating The high thermal conductivity of the Textbook Company, 1928. substrate and the higher electron SiC substrate is particularly useful 4. J. B. Groe and L. E. Larson, mobility result in reduced parasitics. in high-power applications. The high- CDMA Mobile Radio Design, The DC curves are somewhat similar er operating voltage and associated Norwood, MA: Artech House, 2000. to those of a conventional BJT, but higher load impedance greatly sim- 5. R. van Nee and R. Prasad, often contain a saturation resistance plify output networks and power OFDM for Wireless Multimedia as well as saturation voltage. combining. SiC MESFETs typically Communications, Norwood, MA: Currently available AlGaAs/GaAs operate from a 48-V supply. Devices Artech House, 2000. HBTs are capable of producing sever- with outputs of 10 W are currently 6. H. L. Krauss, C. W. Bostian, and al watts and are widely used in wire- available, and outputs of 60 W or F. H. Raab, Solid State Radio less handsets, GaAs HBTs are also more have been demonstrated exper- Engineering, New York: Wiley, 1980. widely used in MMIC circuits at fre- imentally. The cost of SiC devices is 7. F. H. Raab, “Average efficiency quencies up to X band and can oper- at presently about ten times that of of power amplifiers,” Proc. RF ate in PAs at frequencies as high as Si LDMOS. Technology Expo '86, Anaheim, CA, 20 GHz. pp. 474-486, Jan. 30-Feb. 1, 1986. GaN HEMT 8. N. Pothecary, “Feedforward lin- SiGe HBT GaN offers the same high break- ear power amplifiers,” in Workshop The use of SiGe rather than Si in down voltage of SiC, but even higher Notes WFB, Int’l. Microwave Symp., the base of the HBT both increases mobility. Its thermal conductivity is, Boston, MA, June 16, 2001. the maximum operating frequency however, lower, hence GaN devices 9. J. F. Sevic, “Statistical charac- and decreases the base resistance. must be built substrate such as SiC terization of RF power amplifier effi- However, they are generally less effi- or diamond. While the GaN HEMT ciency for wireless communication cient than GaAs HBTs and can have offers the promise of a high-power, systems,” Proc. Wireless Commun. lower breakdown voltages. One high-voltage device operating at fre- Conf., Boulder, CO, pp. 1-4, Aug. 1997. experimental SiGe HBT is capable of quencies of 10 GHz or more, it is still 10. G. Hanington, P.-F. Chen, P. M. delivering over 200 W at L band. in an experimental state. Power out- Asbeck, and L. E. Larson, “High-effi- puts of 8 W at 10 GHz with 30 per- ciency power amplifier using dynam- InP HBT cent efficiency have been demonstrat- ic power-supply voltage for CDMA The use of InP in an HBT further ed. applications,” IEEE Trans. boosts mobility and therefore the Microwave Theory Tech., vol. 47, no. 8, high frequency response. In addition, Monolithic Microwave Integrated pp. 1471-1476, Aug. 1999. InP HBTs have lower turn-on and Circuit (MMIC) 11. I. Kipnis, “Refining CDMA knee voltages, resulting in higher MMICs integrate RF power mobile-phone power control,” gain and efficiency. InP HBTs for RF- devices and matching/decoupling ele- Microwaves & RF, vol. 39, no. 6, pp. power applications incorporate two ments such as on-chip inductors, 71-76, June 2000. heterojunctions (AlInAs/GaInAs and capacitors, resistors, and transmis- 12. J. Staudinger, “Applying GaInAs/InP). The InP in the collector sion lines. The proximity of these ele- switched gain stage concepts to increases the breakdown voltage, ments to the RF-power devices is improve efficiency and linearity for allowing higher output power. To essential for input, output, and inter- mobile CDMA power amplification,” date, outputs of about 0.5 W at 20 stage matching at microwave and Microwave Journal, vol. 43, no. 9, pp. GHz have been demonstrated, but it millimeter-wave frequencies. 152-162, Sept. 2000. is anticipated that operation to 50 or 60 GHz will be possible. References Table 1 References 1. W. J. Bryon, “Arcs and sparks, 13. “Mobile station – base station SiC MESFET Part 1,” Communications Quarterly, interoperability standard for dual- The wide band gap of SiC results vol. 4, no. 2, pp. 27-43, Spring 1994. mode cellular system,” ANSI-136 in both high mobility and high break- 2. W. J. Bryon, “The arc method of Standard, Telecommun. Industries down voltage. An SiC MESFET can producing continuous waves,” Assoc., 2000. therefore have a frequency response Communications Quarterly, vol. 8, no. 14. “Digital cellular communica- comparable to that of a GaAs MES- 3, pp. 47-65, Summer 1998. tion systems,” RCR STD-27, Research FET, but breakdown voltages double 3. K. M. MacIlvain and W. H. and Development Center for Radio that of Si LDMOS. This results in Freedman, Radio Library, Vol. III: Systems (RCR), April 1991.

34 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Acronyms Used in Part 1

AC Alternating Current MESFET MEtal Semiconductor FET ACPR Adjacent-Channel Power Ratio mHEMT Metamorphic HEMT BJT Bipolar-Junction Transistor MMIC Microwave Monolithic Integrated Circuit C/I Carrier-to-Intermodulation MOSFET Metal-Oxide-Silicon Field-Effect Transistor CDF Cumulative Distribution Function NADC North American Digital Cellular CDMA Code-Division Multiple Access NPR Noise-Power Ratio CW Continuous Wave NTSC National Television Standards Committee DC Direct Current OFDM Orthogonal Frequency-Division Multiplex DSP Digital Signal Processing PA Power Amplifier EVM Error-Vector Magnitude PAE Power-Added Efficiency FET Field-Effect Transistor PDF Probability-Density Function FSK Frequency-Shift Keying GMSK Gaussian Minimum Shift Keying PEP Peak-Envelope Power GSM Global System for Mobile communication pHEMT Pseudomorphic HEMT HBT Heterojunction bipolar transistor PSK Phase-Shift Keying HEMT High Electron-Mobility Transistor QAM Quadrature Amplitude Modulation HFET Heterojunction FET (also HJFET) QPSK Quadrature Phase Shift Keying IC Integrated Circuit RF Radio Frequency JFET Junction Field-Effect Transistor SRRC Square-Root Raised Cosine LDMOS Laterally Diffused MOS (FET) SSB Single SideBand

15. “Digital cellular telecommunications system 2. To maintain continuity, all figures, tables, equations (phase 2+), radio transmission and reception,” GSM 5.05 and references will be numbered sequentially throughout Standard, v. 8.4.1, European Telecommun. Standards the entire series. Inst., 1999. 3. Like all articles in High Frequency Electronics, this 16. “Terrestrial trunked radio (TETRA) voice+data air series will be archived and available for downloading (for interface,” TETRA Draft Standard, European personal use by individuals only) online at the magazine Telecommun. Standards Inst., 1999. website: www.highfrequencyelectronics.com 17. “Mobile station – base station interoperability standard for dual-mode wideband spread-spectrum cellu- Author Information lar system,” TIA/EIA IS-95 Interim Standard, The authors of this series of articles are: Frederick H. Telecommun. Industries Assoc., July 1993. Raab (lead author), Green Mountain Radio Research, e- 18. “UE radio transmission and reception (FDD),” TS mail: [email protected]; Peter Asbeck, University of 25.101, v. 3.4.1, Third Generation Partnership Project, California at San Diego; Steve Cripps, Hywave Technical Specification Group, 1999. Associates; Peter B. Kenington, Andrew Corporation; Zoya B. Popovic, University of Colorado; Nick Pothecary, Series Notes Consultant; John F. Sevic, California Eastern 1. The remaining three parts of this series will appear Laboratories; and Nathan O. Sokal, Design Automation. in successive issues of High Frequency Electronics (July, Readers desiring more information should contact the September and November 2003 issues). lead author.

RF and Microwave Power Amplifier and Transmitter Technologies — Part 2

By Frederick H. Raab, Peter Asbeck, Steve Cripps, Peter B. Kenington, Zoya B. Popovich, Nick Pothecary, John F. Sevic and Nathan O. Sokal

art 1 of this series Our multi-part series on introduced basic power amplifier tech- Pconcepts, discussed nologies and applications the characteristics of sig- continues with a review of nals to be amplified, and amplifier configurations, gave background infor- classes of operation, mation on RF power device characterization devices. Part 2 reviews and example applications the basic techniques, ratings, and implementation methods for power amplifiers operating at HF through microwave frequencies.

6a. BASIC TECHNIQUES FOR Figure 7 · A single-ended power amplifier. RF POWER AMPLIFICATION RF power amplifiers are commonly desig- nated as classes A, B, C, D, E, and F [19]. All Consequently, the drain voltage and current but class A employ various nonlinear, switch- waveforms are (ideally) both sinusoidal. The ing, and wave-shaping techniques. Classes of power output of an ideal class-A PA is operation differ not in only the method of 2 operation and efficiency, but also in their Po = Vom / 2R (5) power-output capability. The power-output

capability (“transistor utilization factor”) is where output voltage Vom on load R cannot defined as output power per transistor nor- exceed supply voltage VDD. The DC-power malized for peak drain voltage and current of input is constant and the efficiency of an ideal 1 V and 1 A, respectively. The basic topologies PA is 50 percent at PEP. Consequently, the (Figures 7, 8 and 9) are single-ended, trans- instantaneous efficiency is proportional to the former-coupled, and complementary. The power output and the average efficiency is drain voltage and current waveforms of select- inversely proportional to the peak-to-average ed ideal PAs are shown in Figure 10. ratio (e.g., 5 percent for x = 10 dB). The utilization factor is 1/8. Class A For amplification of amplitude-modulated In class A, the quiescent current is large signals, the quiescent current can be varied in enough that the transistor remains at all proportion to the instantaneous signal enve- times in the active region and acts as a cur- lope. While the efficiency at PEP is rent source, controlled by the drive. unchanged, the efficiency for lower ampli-

22 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

turn proportional to the RF-output current. Consequently, the instantaneous efficiency of a class-B PA varies with the output voltage and for an ideal PA reaches π/4 (78.5 percent) at PEP. For low-level signals, class B is significantly more efficient than class A, and its average efficiency can be several times that of class A at high peak-to-average ratios (e.g., 28 vs. 5 percent for ξ = 10 dB). The utilization factor is the same 0.125 of class A. Figure 8 · Transformer-coupled In practice, the quiescent current push-pull PA. is on the order of 10 percent of the peak drain current and adjusted to minimize crossover distortion caused by transistor nonlinearities at low outputs. Class B is generally used in a push-pull configuration so that the two drain-currents add together to produce a sine-wave output. At HF and VHF, the transformer-coupled push-pull topology (Figure 8) is generally used to allow broadband operation with minimum filtering. The use of the complementary topology Figure 9 · Complementary PA. Figure 10 · Wavefrorms for ideal PAs. (Figure 9) has generally been limited to audio, LF, and MF applications by the lack of suitable p-channel tran- tudes is considerably improved. In an or saturation voltage of the transis- sistors. However, this topology is FET PA, the implementation tor. It is also degraded by the pres- attractive for IC implementation and requires little more than variation of ence of load reactance, which in has recently been investigated for the gate-bias voltage. essence requires the PA to generate low-power applications at frequen- The amplification process in class more output voltage or current to cies to 1 GHz [20]. A is inherently linear, hence increas- deliver the same power to the load. ing the quiescent current or decreas- Class C ing the signal level monotonically Class B In the classical (true) class-C PA, decreases IMD and harmonic levels. The gate bias in a class-B PA is the gate is biased below threshold so Since both positive and negative set at the threshold of conduction so that the transistor is active for less excursions of the drive affect the that (ideally) the quiescent drain cur- than half of the RF cycle (Figure 10). drain current, it has the highest gain rent is zero. As a result, the transis- Linearity is lost, but efficiency is of any PA. The absence of harmonics tor is active half of the time and the increased. The efficiency can be in the amplification process allows drain current is a half sinusoid. increased arbitrarily toward 100 per- class A to be used at frequencies close Since the amplitude of the drain cur- cent by decreasing the conduction to the maximum capability (fmax) of rent is proportional to drive ampli- angle toward zero. Unfortunately, the transistor. However, the efficiency tude and the shape of the drain-cur- this causes the output power (utiliza- is low. Class-A PAs are therefore typ- rent waveform is fixed, class-B pro- tion factor) to decrease toward zero ically used in applications requiring vides linear amplification. and the drive power to increase low power, high linearity, high gain, The power output of a class-B PA toward infinity. A typical compromise broadband operation, or high-fre- is controlled by the drive level and is a conduction angle of 150° and an quency operation. varies as given by eq. (5). The DC- ideal efficiency of 85 percent. The efficiency of real class-A PAs input current is, however, proportion- The output filter of a true class-C is degraded by the on-state resistance al to the drain current which is in PA is a parallel-tuned type that

24 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS bypasses the harmonic components with frequency. Class F of the drain current to ground with- Class-D PAs with power outputs Class F boosts both efficiency and out generating harmonic voltages. of 100 W to 1 kW are readily imple- output by using harmonic resonators When driven into saturation, effi- mented at HF, but are seldom used in the output network to shape the ciency is stabilized and the output above lower VHF because of losses drain waveforms. The voltage wave- voltage locked to supply voltage, associated with the drain capaci- form includes one or more odd har- allowing linear high-level amplitude tance. Recently, however, experimen- monics and approximates a square modulation. tal class-D PAs have been tested with wave, while the current includes even Classical class C is widely used in frequencies of operation as high as 1 harmonics and approximates a half high-power vacuum-tube transmit- GHz [22]. sine wave. Alternately (“inverse class ters. It is, however, little used in F”), the voltage can approximate a solid-state PAs because it requires Class E half sine wave and the current a low drain resistances, making imple- Class E employs a single transis- square wave. As the number of har- mentation of parallel-tuned output tor operated as a switch. The drain- monics increases, the efficiency of an filters difficult. With BJTs, it is also voltage waveform is the result of the ideal PA increases from the 50 per- difficult to set up bias and drive to sum of the DC and RF currents cent (class A) toward unity (class D) produce a true class-C collector-cur- charging the drain-shunt capaci- and the utilization factor increases rent waveform. The use of a series- tance. In optimum class E, the drain from 1/8 (class A) toward 1/2π (class tuned output filter results in a voltage drops to zero and has zero D) [29]. mixed-mode class-C operation that is slope just as the transistor turns on. The required harmonics can in more like mistuned class E than true The result is an ideal efficiency of 100 principle be produced by current- class C. percent, elimination of the losses source operation of the transistor. associated with charging the drain However, in practice the transistor is Class D capacitance in class D, reduction of driven into saturation during part of Class-D PAs use two or more tran- switching losses, and good tolerance the RF cycle and the harmonics are sistors as switches to generate a of component variation. produced by a self-regulating mecha- square drain-voltage waveform. A Optimum class-E operation nism similar to that of saturating series-tuned output filter passes only requires a drain shunt susceptance class C. Use of a harmonic voltage the fundamental-frequency compo- 0.1836/R and a drain series reac- requires creating a high impedance nent to the load, resulting in power tance 1.15R and delivers a power out- (3 to 10 times the load impedance) at π2 2 2 outputs of (8/ )VDD /R and put of 0.577VDD /R for an ideal PA the drain, while use of a harmonic π2 2 (2/ )VDD /R for the transformer-cou- [23]. The utilization factor is 0.098. current requires a low impedance pled and complementary configura- Variations in load impedance and (1/3 to 1/10 of the load impedance). tions, respectively. Current is drawn shunt susceptance cause the PA to While class F requires a more com- only through the transistor that is deviate from optimum operation [24, plex output filter than other PAs, the on, resulting in a 100-percent effi- 25], but the degradations in perfor- impedances must be correct at only a ciency for an ideal PA. The utilization mance are generally no worse than few specific frequencies. Lumped-ele- factor (1/2π = 0.159) is the highest of those for class A and B. ment traps are used at lower fre- any PA (27 percent higher than that The capability for efficient opera- quencies and transmission lines are of class A or B). A unique aspect of tion in the presence of significant used at microwave frequencies. class D (with infinitely fast switch- drain capacitance makes class E use- Typically, a shorting stub is placed a ing) is that efficiency is not degraded ful in a number of applications. One quarter or half-wavelength away by the presence of reactance in the example is high-efficiency HF PAs from the drain. Since the stubs for load. with power levels to 1 kW based upon different harmonics interact and the Practical class-D PAs suffer from low-cost MOSFETs intended for open or short must be created at a losses due to saturation, switching switching rather than RF use [26]. “virtual drain” ahead of the drain speed, and drain capacitance. Finite Another example is the switching- capacitance and bond-wire induc- switching speed causes the transis- mode operation at frequencies as tance, implementation of suitable tors to be in their active regions while high as K band [27]. The class-DE PA networks is a bit of an art. conducting current. Drain capaci- [28] similarly uses dead-space Nonetheless, class-F PAs are success- tances must be charged and dis- between the times when its two tran- fully implemented from MF through charged once per RF cycle. The asso- sistors are on to allow the load net- Ka band. ciated power loss is proportional to work to charge/discharge the drain A variety of modes of operation in- 3 VDD /2 [21] and increases directly capacitances. between class C, E, and F are possi-

26 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

parameter may produce a new set of contours. A variety of different parameters can be plotted during a load-pull analysis, including not only power and efficiency, but also distortion and stability. Harmonic impedances as well as drive impedances are also sometimes varied. A load-pull system consists essentially of a test fixture, provided with biasing capabilities, and a pair of low- loss, accurately resettable tuners, usually of precision mechanical con- Figure 12 · Example load-pull construction. A load-pull characteriza- tours for a 0.5-W, 836 MHz PA. tion procedure consists essentially of (Courtesy Focus Microwaves and measuring the power of a device, to a dBm Engineering) given specification (e.g., the 1-dB compression point) as a function of impedance. Data are measured at a tance lines on the Smith chart. The large number of impedances and Figure 11 · Contant power contours impedances for a specified maximum plotted on a Smith chart. Such plots and transformation. current analogously follow a series- are, of course, critically dependent on resistance line. For an ideal PA, the the accurate calibration of the tuners, resultant constant-power contour is both in terms of impedance and loss- ble. The maximum achievable effi- football-shaped as shown in Figure es. Such calibration is, in turn, highly ciency [30] depends upon the number 11. dependent on the repeatability of the of harmonics, (0.5, 0.707, 0.8165, In a real PA, the ideal drain is tuners. 0.8656, 0.9045 for 1 through 5 har- embedded behind the drain capaci- Precision mechanical tuners, with monics, respectively). The utilization tance and bond-wire/package induc- micrometer-style adjusters, were the factor depends upon the harmonic tance. Transformation of the ideal traditional apparatus for load-pull impedances and is highest for ideal drain impedance through these ele- analysis. More recently, a new gener- class-F operation. ments causes the constant-power ation of electronic tuners has contours to become rotated and dis- emerged that tune through the use 6b. LOAD-PULL torted [31]. With the addition of sec- varactors or transmission lines CHARACTERIZATION ond-order effects, the contours switched by pin diodes. Such elec- RF-power transistors are charac- become elliptical. A set of power con- tronic tuners [32] have the advantage terized by breakdown voltages and tours for a given PA somewhat of almost perfect repeatability and saturated drain currents. The combi- resembles a set of contours for a con- high tuning speed, but have much nation of the resultant maximum jugate match. However, a true conju- higher losses and require highly com- drain voltage and maximum drain gate match produces circular con- plex calibration routines. Mechanical current dictates a range of load tours. With a power amplifier, the tuners are more difficult to control impedances into which useful power process is more correctly viewed as using a computer, and move very can be delivered, as well as an loading to produce a desired power slowly from one impedance setting to impedance for delivery of the maxi- output. As shown in the example of another. mum power. The load impedance for Figure 12, the power and efficiency In an active load-pull system, a maximum power results in drain contours are not necessarily aligned, second power source, synchronized in voltage and current excursions from nor do maximum power and maxi- frequency and phase with the device near zero to nearly the maximum mum efficiency necessarily occur for input excitation, is coupled into the rated values. the same load impedance. Sets of output of the device. By controlling The load impedances correspond- such “load-pull” contours are widely the amplitude and phase of the ing to delivery of a given amount of used to facilitate design trade-offs. injected signal, a wide range of RF power with a specified maximum Load-pull analyses are generally impedances can be simulated at the drain voltage lie along parallel-resis- iterative in nature, as changing one output of the test device [33]. Such a

28 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS system eliminates the expensive nents and are used primarily as and “shorted” means no more 1/10 to tuners, but creates a substantial cali- chokes and by-passes. Matching, tun- 1/3 of the fundamental-frequency bration challenge of its own. The wide ing, and filtering at microwave fre- impedance [FR17]. availability of turn-key load-pull sys- quencies are therefore accomplished A wide variety of class-F PAs have tems has generally reduced the appli- with distributed (transmission-line) been implemented at UHF and cation of active load-pull to situations networks. Proper operation of power microwave frequencies [36-41]. where mechanical or electronic tun- amplifiers at microwave frequencies Generally, only one or two harmonic ing becomes impractical (e.g., mil- is achieved by providing the required impedances are controlled. In the X- limeter-wave frequencies). drain-load impedance at the funda- band PA from [42], for example, the mental and a number of harmonic output circuit provides a match at the 6c. STABILITY frequencies. fundamental and a short circuit at The stability of a small-signal RF the second harmonic. The third-har- amplifier is ensured by deriving a set Class F monic impedance is high, but not of S-parameters from using mea- Class-F operation is specified in explicitly adjusted to be open. The 3- sured data or a linear model, and terms of harmonic impedances, so it dB bandwidth of such an output net- then establishing the value of the k- is relatively easy to see how trans- work is about 20 percent, and the effi- factor stability parameter. If the k- mission-line networks are used. ciency remains within 10 percent of factor is greater than unity, at the Methods for using transmission lines its maximum value over a bandwidth frequency and bias level in question, in conjunction with lumped-element of approximately 10 to 15 percent. then expressions for matching tuned circuits appear in the original Dielectric resonators can be used impedances at input and output can paper by Tyler [34]. In modern in lieu of lumped-element traps in be evaluated to give a perfect conju- microwave implementation, however, class-F PAs. Power outputs of 40 W gate match for the device. Amplifier it is generally necessary to use trans- have been obtained at 11 GHz with design in this context is mainly a mission lines exclusively. In addition, efficiencies of 77 percent [43]. matter of designing matching net- the required impedances must be works which present the prescribed produced at a virtual ideal drain that Class E impedances over the necessary speci- is separated from the output network The drain-shunt capacitance and fied bandwidth. If the k factor is less by drain capacitance, bond-wire/lead series inductive reactance required than unity, negative feedback or lossy inductance. for optimum class-E operation result matching must be employed in order Typically, a transmission line in a drain impedance of R + j0.725R to maintain an unconditionally stable between the drain and the load pro- at the fundamental frequency, design. vides the fundamental-frequency –j1.7846R at the second harmonic, A third case is relevant to PA drain impedance of the desired value. and proportionately smaller capaci- design at higher microwave frequen- A stub that is a quarter wavelength tive reactances at higher harmonics. cies. There are cases where a device at the harmonic of interest and open At microwave frequencies, class-E has a very high k-factor value, but at one end provides a short circuit at operation is approximated by provid- very low gain in conjugate matched the opposite end. The stub is placed ing the drain with the fundamental- condition. The physical cause of this along the main transmission line at frequency impedance and preferably can be traced to a device which has either a quarter or a half wavelength one or more of the harmonic gain roll-off due to carrier-mobility from the drain to create either an impedances [44]. effects, rather than parasitics. In open or a short circuit at the drain An example of a microwave such cases, introduction of some posi- [35]. The supply voltage is fed to the approximation of class E that pro- tive feedback reduces the k-factor drain through a half-wavelength line vides the correct fundamental and and increases the gain in conjugately bypassed on the power-supply end or second-harmonic impedances [44] is matched conditions, while maintain- alternately by a lumped-element shown in Figure 13. Line l2 is a quar- ing unconditional stability. This tech- choke. When multiple stubs are used, ter-wavelength long at the second nique was much used in the early era the stub for the highest controlled harmonic so that the open circuit at of vacuum-tube electronics, especially harmonic is placed nearest the drain. its end is transformed to a short at in IF amplifiers. Stubs for lower harmonics are placed plane AA'. Line l1 in combination progressively further away and their with L and C is designed to be also a 6d. MICROWAVE IMPLEMENTATION lengths and impedances are adjusted quarter wavelength to translate the At microwave frequencies, lumped to allow for interactions. Typically, short at AA' to an open at the tran- elements (capacitors, inductors) “open” means three to ten times the sistor drain. The lines l1 to l4 provide become unsuitable as tuning compo- fundamental-frequency impedance, the desired impedance at the funda-

30 High Frequency Electronics Figure 13 · Idealized microwave class-E PA circuit. Figure 14 · Example X-band class-E PA. mental. The implementation using an power, but class F has about 15 per- HF/VHF Single Sideband FLK052 MESFET is shown in Figure cent higher efficiency. Class E has the One of the first applications of 14 produces 0.68 W at X band with a highest efficiency. Gain compression RF-power transistors was linear drain efficiency of 72 percent and occurs at a lower power level for class amplification of HF single-sideband PAE of 60 percent [42]. E than for class F. For a given effi- signals. Many PAs developed by Methods exist for providing the ciency, class F produces more power. Helge Granberg have been widely proper impedances through the For the same maximum output adapted for this purpose [56, 57]. The fourth harmonic [45]. However, the power, the third order intermodula- 300-W PA for 2 to 30 MHz uses a pair harmonic impedances are not critical tion products are about 10 dB lower of Motorola MRF422 Si NPN transis- [30], and many variations are there- for class F than for class E. Lower- tors in a push-pull configuration. The fore possible. Since the transistor power PAs implemented with smaller PA operates in class AB push-pull often has little or no gain at the high- RF power devices tend to be more from a 28-V supply and achieves a er harmonic frequencies, those efficient than PAs implemented with collector efficiency of about 45 per- impedances often have little or no larger devices [42]. cent (CW) and a two-tone IMD ratio effect upon performance. A single- of about –30 dBc. The 1-kW amplifier stub match is often sufficient to pro- Millimeter-Wave PAs is based upon a push-pull pair of vide the desired impedance at the Solid-state PAs for millimeter- MRF154 MOSFETs and operates fundamental while simultaneously wave (mm-W) frequencies (30 to 100 from a 50-V supply. Over the frequen- providing an adequately high GHz) are predominantly monolithic. cy range of 2 to 50 MHz it achieves a impedance at the second harmonic, Most Ka-band PAs are based upon drain efficiency of about 58 percent thus eliminating the need for an pHEMT devices, while most W-band (CW) with an IMD rating of –30 dBc. extra stub and reducing a portion of PAs are based upon InP HEMTs. the losses associated with it. Most Some use is also made of HBTs at the 13.56-MHz ISM Power Sources microwave class-E amplifiers operate lower mm-W frequencies. Class A is High-power signals at 13.56 MHz in a suboptimum mode [46]. used for maximum gain. Typical per- are needed for a wide variety of Demonstrated capabilities range formance characteristics include 4 W Industrial, Scientific, and Medical from 16 W with 80-percent efficiency with 30-percent PAE at Ka band [53], (ISM) applications such as plasma at UHF (LDMOS) to 100 mW with 250 mW with 25-percent PAE at Q generation, RF heating, and semicon- 60-percent efficiency at 10 GHz [47], band [54], and 200 mW with 10-per- ductor processing. A 400-W class-E [48], [44], [49], [50], [51]. Optical sam- cent PAE at W band [55]. Devices for PA uses an International Rectifier pling of the waveforms [52] has veri- operation at mm-W are inherently IRFP450LC MOSFET (normally fied that these PAs do indeed operate small, so large power outputs are used for low-frequency switching- in class E. obtained by combining the outputs of mode DC power supplies) operates multiple low-power amplifiers in cor- from a 120-V supply and achieves a Comparison porate or spatial power combiners. drain efficiency of 86 percent [58, 26]. PAs configured for classes A (AB), Industrial 13.56-MHz RF power gen- E, and F are compared experimental- 6e. EXAMPLE APPLICATIONS erators using class-E output stages ly in [50] with the following conclu- The following examples illustrate have been manufactured since 1992 sions. Classes AB and F have essen- the wide variety of power amplifiers by Dressler Hochfrequenztechnik tially the same saturated output in use today: (Stolberg, Germany) and Advanced

July 2003 31 High Frequency Design RF POWER AMPLIFIERS

Figure 16 · Output section of a 50- Figure 15 · 3-kW high efficiency PA for 13.56 ISM-band operation. kW AM broadcast transmitter. (Courtesy Advanced Energy) (Courtesy Harris)

Energy Industries (Ft. Collins, CO). MF AM Broadcast Transmitters 900-MHz Cellular-Telephone They typically use RF-power Since the 1980s, AM broadcast Handset MOSFETs with 500- to 900-V break- transmitters (530 to 1710 kHz) have Most 900-MHz CDMA handsets down voltages made by Directed been made with class-D and -E RF- use power-amplifier modules from Energy or Advanced Power output stages. Amplitude modulation vendors such as Conexant and RF Technology and produce output pow- is produced by varying the supply Micro Devices. These modules typi- ers of 500 W to with 3 kW with drain voltage of the RF PA with a high-effi- cally contain a single GaAs-HBT efficiencies of about 90 percent. The ciency amplitude modulator. RFIC that includes a single-ended Advanced Energy Industries amplifi- Transmitters made by Harris class-AB PA. Recently developed PA er (Figure 15) uses thick-film-hybrid (Mason, Ohio) produce peak-envelope modules also include a silicon control circuits to reduce size. This allows output powers of 58, 86, 150, 300, and IC that provides the base-bias refer- placement inside the clean-room 550 kW (unmodulated carrier powers ence voltage and can be commanded facilities of semiconductor-manufac- of 10, 15, 25, 50, and 100 kW). The to adjust the output-transistor base turing plants, eliminating the need 100-kW transmitter combines the bias to optimize efficiency while for long runs of coaxial cable from an output power from 1152 transistors. maintaining acceptably low amplifier RF-power generator installed outside The output stages can use either distortion. over the full ranges of the clean-room. bipolars or MOSFETs, typically oper- temperature and output power.A typ- ate in class DE from a 300-V supply, ical module (Figure 17) produces 28 VHF FM Broadcast Transmitter and achieve an efficiency of 98 per- dBm (631 mW) at full output with a FM-broadcast transmitters (88 to cent. The output section of the Harris PAE of 35 to 50 percent. 108 MHz) with power outputs from 3DX50 transmitter is shown in 50 W to 10 kW are manufactured by Figure 16. Cellular-Telephone Base Broadcast Electronics (Quincy, Transmitters made by Broadcast Station Transmitter Illinois). These transmitters use up to Electronics (Quincy, IL) use class-E The Spectrian MCPA 3060 cellu- 32 power-combined PAs based upon RF-output stages based upon lar base-station transmitter for 1840- Motorola MRF151G MOSFETs. The APT6015LVR MOSFETs operating 1870 MHz CDMA systems provides PAs operate in class C from a 44-V from 130-V maximum supply volt- up to 60-W output while transmitting supply and achieve a drain efficiency ages. They achieve drain efficiencies a signal that may include as many 9 of 80 percent. Typically, about 6 per- of about 94 percent with peak-enve- modulated carriers. IMD is mini- cent of the output power is dissipated lope output powers from 4.4 to 44 kW. mized by linearizing a class-AB main in the power combiners, harmonic- The 44-kW AM-10A transmitter com- amplifier with both adaptive predis- suppression filter, and lightning-pro- bines outputs from 40 individual out- tortion and adaptive feed-forward tection circuit. put stages. cancellation. The adaptive control

32 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

for next-generation wireless communications,” Microwave J., vol. 42, no. 2, pp. 22-42, Feb. 1999. 23. N. O. Sokal and A. D. Sokal, “Class E—a new class of high efficiency tuned single-ended switching power amplifiers,” IEEE J. Solid-State Circuits, vol. SC-10, no. 3, pp. 168-176, June 1975. 24. F. H. Raab, “Effects of circuit variations on the class E tuned power Figure 17 · Internal view of a dual- amplifier,” IEEE J. Solid State Circuits, band (GSM/DCS) PA module for vol. SC-13, no. 2, pp. 239-247, April cellular telephone handsets. Figure 18 · Thick-film hybrid S-band 1978. (Courtesy RF Micro Devices) PA module. (Courtesy UltraRF) 25. F. H. Raab, “Effects of VSWR upon the class-E RF-power amplifier,” Proc. RF Expo East ’88, Philadelphia, system adjusts operation as needed GaAs MMIC Power Amplifier PA, pp. 299-309, Oct. 25-27, 1988. to compensate for changes due to A MMIC PA for use from 8 to 14 26. J. F. Davis and D. B. Rutledge, “A temperature, time, and output power. GHz is shown in Figure 19. This low-cost class-E power amplifier with The required adjustments are amplifier is fabricated with GaAs sine-wave drive,” Int. Microwave Symp. derived from continuous measure- HBTs and intended for used in Digest, vol. 2, pp. 1113-1116, Baltimore, ments of the system response to a phased-array radar. It produces a 3- MD, June 7-11, 1998. spread-spectrum pilot test signal. W output with a PAE of approxi- 27. T. B. Mader and Z. B. Popovic, The amplifier consumes a maximum mately 40 percent [59]. “The transmission-line high-efficiency of 810 W from a 27-V supply. class-E amplifier,” IEEE Microwave References and Guided Wave Letters, vol. 5, no. 9, S-Band Hybrid Power Module 19. H. L. Krauss, C. W. Bostian, and pp. 290-292, Sept. 1995. A thick-film-hybrid power-ampli- F. H. Raab, Solid State Radio 28. D. C. Hamill, “Class DE invert- fier module made by UltraRF (now Engineering, New York: Wiley, 1980. ers and rectifiers for DC-DC conver- Cree Microwave) for 1805 to 1880 20. R. Gupta and D. J. Allstot, “Fully sion,” PESC96 Record, vol. 1, pp. 854- MHz DCS and 1930-1960 MHz PCS monolithic CMOS RF power amplifiers: 860, June 1996. is shown in Figure 18. It uses four Recent advances,” IEEE Communi- 29. F. H. Raab, “Maximum efficiency 140-mm LDMOS FETs operating cations Mag., vol. 37, no. 4, pp. 94-98, and output of class-F power amplifiers,” from a 26-V drain supply. The indi- April 1999. IEEE Trans. Microwave Theory Tech., vidual PAs have 11-dB power gain 21. F. H. Raab and D. J. Rupp, “HF vol. 47, no. 6, pp. 1162-1166, June 2001. and are quadrature-combined to pro- power amplifier operates in both class 30. F. H. Raab, “Class-E, -C, and -F duce a 100-W PEP output. The aver- B and class D,” Proc. RF Expo West ’93, power amplifiers based upon a finite age output power is 40 W for EDGE San Jose, CA, pp. 114-124, March 17-19, number of harmonics,” IEEE Trans. and 7 W for CDMA, with an ACPR of 1993. Microwave Theory Tech., vol. 47, no. 8, –57 dBc for EDGE and –45 dBc for 22. P.Asbeck, J. Mink, T. Itoh, and G. pp. 1462-1468, Aug. 2001. CDMA. The construction is based Haddad, “Device and circuit approaches 31. S. C. Cripps, RF Power upon 0.02-in. thick film with silver Amplifiers for Wireless Communi- metalization. Acronyms Used in Part 2 cation, Norwood, MA: Artech, 1999. BJT Bipolar Junction 32. “A load pull system with har- Transistor monic tuning,” Microwave J., pp. 128- DSP Digital Signal 132, March 1986. Processor 33. B. Hughes, A. Ferrero, and A. IC Integrated Circuit Cognata, “Accurate on-wafer power and IMD Intermodulation harmonic measurements of microwave Distortion amplifiers and devices,” IEEE Int. MOSFET Metal Oxide Silicon Microwave Symp. Digest, Albuquerque, Figure 19 · MMIC PA for X- and K- FET NM, pp. 1019-1022, June 1-5, 1992. bands. 34. V. J. Tyler, “A new high-efficiency

34 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS high power amplifier,” The Marconi 45. A. J. Wilkinson and J. K. A. sources,” Int. Microwave Symp. Digest, Review, vol. 21, no. 130, pp. 96-109, Fall Everard, “Transmission line load net- vol. 2, pp. 955-958, Boston, MA, June 1958. work topology for class E amplifiers,” 13-15, 2000. 35. A. V. Grebennikov, “Circuit IEEE Trans. Microwave Theory Tech., 56. H. Granberg, “Get 300 watts design technique for high-efficiency vol. 47, no. 6, pp. 1202-1210, June 2001. PEP linear across 2 to 30 MHz from class-F amplifiers,” Int. Microwave 46. F. H. Raab, “Suboptimum opera- this push-pull amplifier,” Bulletin Symp. Digest, vol. 2, pp. 771-774, tion of class-E power amplifiers,” Proc. EB27A, Motorola Semiconductor Boston, MA, June 13-15, 2000. RF Technology Expo., Santa Clara, CA, Products, Phoenix, Feb. 1980. 36. P. Colantonio, F. Giannini, G. pp. 85-98, Feb. 1989. 57. H. Granberg, “A compact 1-kW Leuzzi, and E. Limiti, “On the class-F 47. S. Li, “UHF and X-band class-E 2-50 MHz solid-state linear amplifier,” power-amplifier design,” RF and Micro- amplifiers,” Ph.D. Thesis, California QEX, no. 101, pp. 3-8, July 1990. Also wave Computer-Aided Engin-eering, Institute of Technology, Pasadena, 1999. AR347, Motorola Semiconductor vol. 32, no. 2, pp. 129-149, March 1999. 48. F. J. Ortega-Gonzalez, J. L. Products, Feb. Oct. 1990. 37. A. N. Rudiakova and V. G. Jimenez-Martin, A. Asensio-Lopez, G. 58. N. O. Sokal, “Class-E RF power Krizhanovski, “Driving waveforms for Torregrosa-Penalva, “High-efficiency amplifiers ... ,” QEX, No. 204, pp. 9-20, class-F power amplifiers,” Int. load-pull harmonic controled class-E Jan./Feb. 2001. Microwave Symp. Digest, vol. 1, pp. 473- power amplifier,” IEEE Microwave 59. M. Salib, A. Gupta, A. Ezis, M. 476, Boston, MA, June 13-15, 2000. Guided Wave Lett., vol. 8, no. 10, pp. Lee, and M. Murphy, “A robust 3W high 38. A. Inoue, T. Heima, A. Ohta, R. 348-350, Oct. 1998. efficiency 8-14 GHz GaAs/AlGaAs het- Hattori, and Y. Mitsui, “Analysis of 49. E. Bryerton, “High-efficiency erojunction bipolar transistor power class-F and inverse class-F amplifiers,” switched-mode microwave circuits,” amplifier,” Int. Microwave Symp. Int. Microwave Symp. Digest, vol. 2, pp. Ph.D. dissertation, Univ. of Colorado, Digest, vol. 2, pp. 581-584, Baltimore, 775-778, Boston, MA, June 13-15, 2000. Boulder, June 1999. MD, June 7-11, 1998. 39. F. van Rijs et al., “Influence of 50. T. B. Mader, E. W. Bryerton, M. output impedance on power added effi- Markovic, M. Forman, and Z. Popovic, Author Information ciency of Si-bipolar power transistors,” “Switched-mode high-efficiency micro- The authors of this series of arti- Int. Microwave Symp. Digest, vol. 3, pp. wave power amplifiers in a free-space cles are: Frederick H. Raab (lead 1945-1948, Boston, MA, June 13-15, power-combiner array,” IEEE Trans. author), Green Mountain Radio 2000. Microwave Theory Tech., vol. 46, no. 10, Research, e-mail: [email protected]; 40. F. Huin, C. Duvanaud, V. Serru, pt. I, pp. 1391-1398, Oct. 1998. Peter Asbeck, University of F. Robin, and E. Leclerc, “A single supply 51. M. D. Weiss and Z. Popovic, “A 10 California at San Diego; Steve very high power and efficiency integrat- GHz high-efficiency active antenna,” Cripps, Hywave Associates; Peter B. ed PHEMT amplifier for GSM applica- Int. Microwave Symp. Digest, vol. 2, pp. Kenington, Andrew Corporation; tions,” Proc. 2000 RFIC Symp., Boston, 663-666, Anaheim, CA, June 14-17, Zoya B. Popovic, University of MA, CD-ROM, June 11-13, 2000. 1999. Colorado; Nick Pothecary, 41. B. Ingruber et al., 52. M. Weiss, M. Crites, E. Bryerton, Consultant; John F. Sevic, California “Rectangularly driven class-A harmon- J. Whitacker, and Z. Popovic, “"Time Eastern Laboratories; and Nathan O. ic-control amplifier,” IEEE Trans. domain optical sampling of nonlinear Sokal, Design Automation. Readers Microwave Theory Tech., pt. 1, vol. 46, microwave amplifiers and multipliers,” desiring more information should no. 11, pp. 1667-1672, Nov. 1998. IEEE Trans. Microwave Theory Tech., contact the lead author. 42. E. W. Bryerton, M. D. Weiss, and vol. 47, no.12, pp. 2599-2604, Dec. 1999. Z. Popovic, “Efficiency of chip-level ver- 53. J. J. Komiak, W. Kong, P. C. Notes sus external power combining,” IEEE Chao, and K. Nichols, “Fully monolithic 1. In Part 1 of this series (May Trans. Microwave Theory Tech., vol. 47, 4 watt high efficiency Ka-band power 2003 issue), the references contained no. 8, pp. 1482-1485, Aug. 1999. amplifier,” Int. Microwave Symp. in Table 1 were not numbered cor- 43 S. Toyoda, “Push-pull power Digest, vol. 3, pp. 947-950, Anaheim, rectly. The archived version has been amplifiers in the X band,” Int. CA, June 14-17, 1999. corrected and may be downloaded Microwave Symp. Digest, vol. 3, pp. 54. S.-W. Chen et al., “A 60-GHz from: www.highfrequencyelectronics. 1433-1436, Denver, CO, June 8-13, 1997. high-efficiency monolithic power ampli- com — click on “Archives,” select 44. T. B. Mader and Z. Popovic, “The fier using 0.1-µm pHEMTs,” IEEE “May 2003 — Vol. 2 No. 3” then click transmission-line high-efficiency class- Microwave and Guided Wave Lett., on the article title. E amplifier,” IEEE Micro-wave and vol.5, pp. 201-203, June 1995. 2. This series has been extended Guided Wave Lett., vol. 5, no. 9, pp. 290- 55. D. L. Ingram et al., “Compact W- to five parts, to be published in succe- 292, Sept. 1995. band solid-state MMIC high power sive issues through January 2004.

36 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS From September 2003 High Frequency Electronics Copyright © 2003 Summit Technical Media, LLC RF and Microwave Power Amplifier and Transmitter Technologies — Part 3

By Frederick H. Raab, Peter Asbeck, Steve Cripps, Peter B. Kenington, Zoya B. Popovich, Nick Pothecary, John F. Sevic and Nathan O. Sokal

he building blocks

Transmitter architectures is used in transmit- Mixer the subject of this install- Tters are not only RF/ ment of our continuing power amplifiers, but a Baseband Exciter series on power amplifiers, variety of other circuit with an emphasis on elements including oscil- 3-stage PA designs that can meet lators, mixers, low-level LO RX today’s linearity and high amplifiers, filters, match- efficiency requirements ing networks, combiners, and circulators. The Figure 20 · A conventional transmitter. arrangement of building blocks is known as the architecture of a transmitter. The classic transmitter architecture is based upon linear The outputs of a driver stage must be PAs and power combiners. More recently, matched to the input of the following stage transmitters are being based upon a variety of much as the final amplifier is matched to the different architectures including stage load. The matching tolerance for maintaining bypassing, Kahn, envelope tracking, outphas- power level can be significantly lower than ing, and Doherty. Many of these are actually that for gain [60], hence the 1-dB load-pull fairly old techniques that have been recently contours are more tightly packed for power made practical by the capabilities of DSP. than for gain. To obtain even modest bandwidths (e.g., 7a. LINEAR ARCHITECTURE above 5 percent), the use of quadrature bal- The conventional architecture for a linear anced stages is advisable (Figure 21). The microwave transmitter consists of a baseband main benefit of the quadrature balanced con- or IF modulator, an up-converter, and a power- figuration is that reflections from the transis- amplifier chain (Figure 20). The amplifier tors are cancelled by the action of the input chain consists of cascaded gain stages with and output couplers. An individual device can power gains in the range of 6 to 20 dB. If the therefore be deliberately mismatched (e.g., to transmitter must produce an amplitude-mod- achieve a power match on the output), yet the ulated or multi-carrier signal, each stage must quadrature-combined system appears to be have adequate linearity. This generally well-matched. This configuration also acts as requires class-A amplifiers with substantial an effective power combiner, so that a given power back-off for all of the driver stages. The power rating can be achieved using a pair of final amplifier (output stage) is always the devices having half of the required power per- most costly in terms of device size and current formance. For moderate-bandwidth designs, consumption, hence it is desirable to operate the lower-power stages are typically designed the output stage in class B. In applications using a simple single-ended cascade, which in requiring very high linearity, it is necessary to some cases is available as an RFIC. Designs use class A in spite of the lower efficiency. with bandwidths approaching an octave or

34 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

ture [60]. Even when larger devices are available, smaller devices often offer higher gain, a lower matching Q 0º factor (wider bandwidth), better phase linearity, and lower cost. Heat dissipation is more readily accomplished with a number of small

90º devices, and a soft-failure mode becomes possible. On the other hand, the increase in parts count, assembly time, and physical size are significant Figure 21 · Amplifier stages with quadrature combiners. disadvantages to the use of multiple, smaller devices. Direct connection of multiple PAs more require the use of quadrature- stage itself with k>1. This is another is generally impractical as the PAs balanced stages throughout the reason for using quadrature coupled interact, allowing changes in output entire chain. stages in the output of the chain. from one to cause the load impedance Simple linear-amplifier chains of Large RF power devices typically seen by the other to vary. A constant this kind have high linearity but only have very high transconductance, and load impedance, hence isolation of modest efficiency. Single-carrier this can produce low-frequency insta- one PA from the other, is provided by applications usually operate the final bility unless great care is taken to a hybrid combiner. A hybrid combiner amplifier to about the 1-dB compres- terminate both the input and output causes the difference between the sion point on amplitude modulation at low frequencies with impedances two PA outputs to be routed to and peaks. A thus-designed chain in for unconditional stability. Because of dissipated in a balancing or “dump” which only the output stage exhibits large separation from the RF band, resistor. In the event that one PA compression can still deliver an this is usually a simple matter requir- fails, the other continues to operate ACPR in the range of about –25 dBc ing a few resistors and capacitors. normally, with the transmitter out- with 50-percent efficiency at PEP. At X band and higher, the power put reduced to one fourth of nominal. Two practical problems are fre- gain of devices in the 10 W and above The most common power combin- quently encountered in the design of category can drop well below 10 dB. er is the quadrature-hybrid combiner. linear PA chains: stability and low To maintain linearity, it may be nec- A 90° phase shift is introduced at gain. Linear, class-A chains are actu- essary to use a similarly size device input of one PA and also at the out- ally more susceptible to oscillation as a driver. Such an architecture put of the other. The benefits of due to their high gain, and single- clearly has a major negative impact quadrature combining include con- path chains are especially prone to upon the cost and efficiency of the stant input impedance in spite of unstable behavior. Instability can be whole chain. In the more extreme variations of input impedances of the subdivided into the two distinct cate- cases, it may be advantageous to con- individual PAs, cancellation of odd gories: Low-frequency oscillation and sider a multi-way power combiner, harmonics, and cancellation of back- in-band instability. In-band instabili- where 4, 8, or an even greater num- ward-IMD (IMD resulting from a sig- ty is avoided by designing the indi- ber of smaller devices are combined. nal entering the output port). In vidual gain stages to meet the crite- Such an approach also has other addition, the effect of load impedance ria for unconditional stability; i.e., advantages, such as soft failure, bet- upon the system output is greatly the Rollet k factor [61] must be ter thermal management, and phase reduced (e.g., to 1.2 dB for a 3:1 greater than unity for both in-band linearity. However, it typically con- SWR). Maintenance of a nearly con- and out-of-band frequencies. Meeting sumes more board space. stant output occurs because the load this criterion usually requires sacri- impedance presented to one PA ficing some gain through the use of 7b. POWER COMBINERS decreases when that presented to the absorptive elements. Alternatively, The need frequently arises to other PA increases. As a result, how- the use of quadrature balanced combine the outputs of several indi- ever, device ratings increase and effi- stages provides much greater isola- vidual PAs to achieve the desired ciency decreases roughly in proportion between individual stages, and transmitter output. Whether to use a tion to the SWR [65]. Because the broadband response of the number of smaller PAs vs. a single quadrature combiners are inherently quadrature couplers can eliminate larger PA is one of the most basic two-terminal devices, they are used the need to design the transistor decisions in selection of an architec- in a corporate combining architecture

36 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 23 · Power consumption by Figure 22 · Multi-section Wilkinson combining architecture. PAs of different sizes.

(Figure 21). Unfortunately, the physical construction of pended in a machined cavity. Structures of this kind allow such couplers poses some problems in a PC-board envi- high-power 8-way combiners with octave bandwidths and ronment. The very tight coupling between the two quar- of 0.5 dB. ter-wave transmission lines requires either very fine gaps A wide variety of other approaches to power-combin- or a three-dimensional structure. This problem is circum- ing circuits are possible [62, 64]. Microwave power can vented by the use of a miniature co-axial cable having a also be combined during radiation from multiple anten- pair of precisely twisted wires to from the coupling sec- nas through “quasi-optical” techniques [66]. tion or ready-made, low-cost surface mount 3-dB couplers. The Wilkinson or in-phase power combiner [62] is 7c. STAGE SWITCHING AND BYPASSING often more easily fabricated than a quadrature combiner. The power amplifier in a portable transmitter gener- In the two-input form (as in each section in Figure 22), ally operates well below PEP output, as discussed in the outputs from two quarter-wavelength lines summed Section 4 (Part 1). The size of the transistor, quiescent into load R0 produce an apparent load impedance of 2R0, current, and supply voltage are, however, determined by which is transformed through the lines into at the load the peak output of the PA. Consequently, a PA with a impedances RPA seen by the individual PAs. The differ- lower peak output produces low-amplitude signals more ence between the two PA outputs is dissipated in a resis- efficiently than does a PA with a larger peak output, as tor connected across the two inputs. Proper choice of the illustrated in Figure 23 for class-B PAs with PEP effi- balancing resistor (2RPA) produces a hybrid combiner ciencies of 60 percent. Stage-bypassing and gate-switch- with good isolation between the two PAs. The Wilkinson ing techniques [67, 68] reduce power consumption and concept can be extended to include more than two inputs increase efficiency by switching between large and small [63]. amplifiers according to signal level. This process is analo- Greater bandwidth can be obtained by increasing the gous to selection of supply voltage in a class-G PA, and number of transforming sections in each signal path. A the average efficiency can be similarly computed [69]. single-section combiner can have a useful bandwidth of A typical stage-bypassing architecture is shown in about 20 percent, whereas a two-section version can have Figure 24. For low-power operation, switches SA and SB a bandwidth close to an octave. In practice, escalating cir- route the drive signal around the final amplifier. cuit losses generally preclude the use of more than two sections. All power-combining techniques all suffer from circuit losses as well as mismatch losses. The losses in a simple two-way combiner are typically about 0.5 dB or 10 percent. For a four-way corporate structure, the interconnects typically result in higher losses. Simple open microstrip lines are too lossy for use in combining structures. One technique that offers a good compromise among cost, produceability, and performance, uses sus- pended stripline. The conductors are etched onto double- sided PC board, interconnected by vias, and then sus- Figure 24 · Stage-bypassing architecture.

38 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 25 · Adaptive gate switching. Figure 26 · Kahn-technique transmitter.

Simultaneously, switch SDC turns-off cient but nonlinear RF power amplifi- linearity are the envelope bandwidth DC power to the final amplifier. The er (PA) with a highly efficient enve- and alignment of the envelope and reduction in power consumption can lope amplifier to implement a high- phase modulations. As a rule of improve the average efficiency signif- efficiency linear RF power amplifier. thumb, the envelope bandwidth must icantly (e.g., from 2.1 to 9.5 percent in In its classic form [73], a limiter elim- be at least twice the RF bandwidth [70]). The control signal is based upon inates the envelope, allowing the con- and the misalignment must not the signal envelope and power level stant-amplitude phase modulated exceed one tenth of the inverse of the (back-off). Avoiding hysteresis effects carrier to be amplified efficiently by RF bandwidth [76]. In practice, the and distortion due to switching tran- class-C, -D, -E, or -F RF PAs. drive is not hard-limited as in the sients are critical issues in imple- Amplitude modulation of the final RF classical implementation. Drive mentation. PA restores the envelope to the phase- power is conserved by allowing the A PA with adaptive gate switching modulated carrier creating an ampli- drive to follow the envelope except at is shown in Figure 25. The gate width fied replica of the input signal. low levels. The use of a minimum (hence current and power capability) EER is based upon the equiva- drive level ensures proper operation of the upper FET is typically ten to lence of any narrowband signal to of the RF PA at low signal levels twenty times that of the lower FET. simultaneous amplitude (envelope) where the gain is low [77]. At higher The gate bias for the high-power FET and phase modulations. In a modern microwave frequencies, the RF power keeps it turned off unless it is needed implementation, both the envelope devices exhibit softer saturation to support a high-power output. and the phase-modulated carrier are characteristics and larger amounts of Consequently, the quiescent drain generated by a DSP. In contrast to amplitude-to-phase conversion, current is reduced to low levels unless linear amplifiers, a Kahn-technique necessitating the use of predistortion actually needed. The advantages of transmitter operates with high effi- for good linearity [78]. this technique are the absence of loss ciency over a wide dynamic range in the switches required by stage and therefore produces a high aver- bypassing, and operation of the low- age efficiency for a wide range of sig- power FET in a more linear region nals and power (back-off) levels. (vs. varying the gate bias of a single Average efficiencies three to five large FET). The disadvantage is that times those of linear amplifiers have the source and load impedances been demonstrated (Figure 27) from change as the upper FET is switched HF [74] to L band [75]. on and off. Transmitters based upon the Kahn technique generally have excel- 7d. KAHN TECHNIQUE lent linearity because linearity The Kahn Envelope Elimination depends upon the modulator rather and Restoration (EER) technique than RF power transistors. The two Figure 27 · Efficiency of Kahn-tec- (Figure 26) combines a highly effi- most important factors affecting the nique transmitters.

40 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 28 · Class-S modulator. Figure 29 · Class-G modulator.

Class-S Modulator A class-S modulator (Figure 28) uses a transistor and (typically 90 percent) is achieved by a diplexing combiner. diode or a pair of transistors act as a two-pole switch to Obtaining a flat frequency response and resistive loads generate a rectangular waveform with a switching fre- for the two PAs is achieved by splitting the input signals quency several times that of the output signal. The width in a DSP that acts as a pair of negative-component filters of pulses is varied in proportion to the instantaneous (Figure 30) [79]. The split-band modulator should make amplitude of the desired output signal, which is recovered possible Kahn-technique transmitters with RF band- by a low-pass filter. Class S is ideally 100 percent efficient widths of tens or even hundreds of MHz. and in practice can have high efficiency over a wide dynamic range. Class-S modulators are typically used as 7e. ENVELOPE TRACKING parts of a Kahn-technique transmitter, while class-S The envelope-tracking architecture (Figure 31) is sim- amplifiers are becoming popular for the efficient produc- ilar to that of the Kahn technique. However, the final tion of audio power in portable equipment. A class-S mod- amplifier operates in a linear mode and the supply volt- ulator can be driven by a digital (on/off) signal supplied age is varied dynamically to conserve power [81, 82]. The directly from a DSP, eliminating the need for intermedi- RF drive contains both amplitude and phase information, ate conversion to an analog signal. and the burden of providing linear amplification lies Selection of the output filter is a compromise between entirely on the final RF PA. The role of the variable power passing the infinite-bandwidth envelope and rejecting supply is only to optimize efficiency. FM-like spurious components that are inherent in the Typically, the envelope is detected and used to control PWM process. Typically, the switching frequency must be a DC-DC converter. While both buck (step-down) or boost six to seven times the RF bandwidth. Modulators with (step-up) converters are used, the latter is more common switching frequencies of 500 kHz are readily implement- as it allows operation of the RF PA from a supply voltage ed using discrete MOSFETs and off-the-shelf ICs [74], higher than the DC-supply voltage. This configuration is while several MHz can be achieved using MOS ASICs or discrete GaAs devices [75].

Class-G Modulator A class-G modulator (Figure 29) is a combination of linear series-pass (class-B) amplifiers that operate from different supply voltages. Power is conserved by selecting the one with the lowest useable supply voltage [69] so that the voltage drop across the active device is minimized.

Split-Band Modulator Most of the power in the envelope resides at lower frequencies; typically 80 percent is in the DC component. The bandwidth of a class-S modulator can therefore be extended by combining it with a linear amplifier. While there are a number of approaches, the highest efficiency Figure 30 · Split-band modulator.

42 High Frequency Electronics Figure 32 · Efficiency of a GaAs FET envelope- Figure 31 · Envelope-tracking architecture. tracking transmitter. also more amenable to the use of npn or n-channel tran- 7f. OUTPHASING sistors for fast switching. The result is a minimum VDDRF Outphasing was invented during the 1930s as a corresponding to the DC-supply voltage and tracking of means of obtaining high-quality AM from vacuum tubes larger envelopes with a fixed “headroom” to ensure linear with poor linearity [86] and was used through about 1970 operation of the RF PA. If the RF PA is operated in class in RCA “Ampliphase” AM-broadcast transmitters. In the A, its quiescent current can also be varied. 1970s, it came into use at microwave frequencies under In general, excess power-supply voltage translates to the name “LINC” (Linear Amplification using Nonlinear reduced efficiency, rather than output distortion. In prin- Components) [87]. ciple, perfect tracking of the envelope by the supply volt- An outphasing transmitter (Figure 33) produces an age preserves the peak efficiency of the RF PA for all out- amplitude-modulated signal by combining the outputs of put amplitudes, as in the Kahn technique. In practice, two PAs driven with signals of different time-varying efficiency improvement is obtained over a limited range of phases. Basically, the phase modulation causes the output power. instantaneous vector sum of the two PA outputs to follow A high switching frequency in the DC-DC converter the desired signal amplitude (Figure 34). The inverse allows both a high modulation bandwidth and the use of sine of envelope E phase-modulates the driving signals smaller inductors and capacitors. The switching devices for the two PAs to produce a transmitter output that is in the converter can in fact be implemented using the proportional to E. In a modern implementation, a DSP same same transistor technology used in the RF PA. and synthesizer produce the inverse-sine modulations of Converters with switching frequencies of 10 to 20 MHz the driving signals. have recently been implemented using MOS ASICs [80], Hybrid combining (Figure 33) isolates the PAs from GaAs HBTs [83, 84] and RF power MOSFETs [85]. the reactive loads inherent in outphasing, allowing them Representative results for an envelope-tracking trans- to see resistive loads at all signal levels. However, both mitter based on a GaAs FET power amplifier are shown in PAs deliver full power all of the time. Consequently, the Figure 32. The efficiency is lower at high power than that efficiency of a hybrid-coupled outphasing transmitter of the conventional amplifier with constant supply voltage varies with the output power (Figure 35), resulting in an due to the inefficiency of the DC-DC converter. However, average efficiency that is inversely proportional to peak- the efficiency is much higher over a wide range of output to-average ratio (as for class A). Recovery of the power power, with the average efficiency approximately 40 per- from the dump port of the hybrid combiner offers some cent higher than that of the conventional linear amplifier. improvement in the efficiency [88]. Spurious outputs can be produced by supply-voltage The phase of the output current is that of the vector ripple at the switching frequency. The effects of the ripple sum of the two PA-output voltages. Direct summation of can be minimized by making the switching frequency suf- the out-of-phase signals in a nonhybrid combiner inher- ficiently high or by using an appropriate filter. Variation ently results in reactive load impedances for the power of the RF PA gain with supply voltage can introduce dis- amplifiers [89]. If the reactances are not partially can- tortion. Such distortion can, however, be countered by pre- celled as in the Chireix technique, the current drawn from distortion techniques [to be covered in Section 8 (Part 4)]. the PAs is proportional to the transmitter-output voltage.

September 2003 43 High Frequency Design RF POWER AMPLIFIERS

Figure 33 · Hybrid-combined outphasing transmitter. Figure 36 · Chireix-outphasing transmitter.

This results in an efficiency charac- ear amplification). length lines or networks. The “carri- teristic similar to that of a class-B PA. Virtually all microwave outphas- er” (main) PA is biased in class B The Chireix technique [86] uses ing systems in use today are of the while the “peaking” (auxiliary) PA is shunt reactances on the inputs to the hybrid-coupled type. Use of the biased in class C. Only the carrier PA combiner (Figure 36) to tune-out the Chireix technique at microwave fre- is active when the signal amplitude is drain reactances at a particular quencies is difficult because half or less of the PEP amplitude. amplitude, which in turn maximizes microwave PAs do not behave as ideal Both PAs contribute output power the efficiency in the vicinity of that voltage sources. Simulations suggest when the signal amplitude is larger amplitude. The efficiency at high and that direct (nonhybrid) combining than half of the PEP amplitude low amplitudes may be degraded. In increases both efficiency and distor- Operation of the Doherty system the classic Chireix implementation, tion [91]. Since outphasing offers a can be understood by dividing it into the shunt reactances maximize the wide bandwidth and the distortion low-power, medium-power (load-mod- efficiency at the level of the unmodu- can be mitigated by techniques such ulation), and peak-power regions lated carrier in an AM signal and pro- as predistortion, directly coupled and [96]. The current and voltage rela- duce good efficiency over the upper 6 Chireix techniques should be fruitful tionships are shown in Figure 38 for dB of the output range. With judi- areas for future investigation. ideal transistors and lossless match- cious choice of the shunt suscep- ing networks. In the low-power tances, the average efficiency can be 7g. DOHERTY TECHNIQUE region, the instantaneous amplitude maximized for any given signal [89, Development of the Doherty tech- of the input signal is insufficient to 90]. For example, a normalized sus- nique in 1936 [92] was motivated by overcome the class-C (negative) bias ceptance of 0.11 peaks the instanta- the observation that signals with sig- of the peaking PA, thus the peaking neous efficiency at a somewhat lower nificant amplitude modulation PA remains cut-off and appears as an amplitude, resulting in an average resulted in low average efficiency. open-circuit. With the example load efficiency of 52.1 percent for an ideal The classical Doherty architecture impedances shown in Figure 37, the class-B PA and a 10-dB Rayleigh- (Figure 37) combines two PAs of carrier PA sees a 100 ohm load and envelope signal (vs. 28 percent for lin- equal capacity through quarter-wave- operates as an ordinary class-B amplifier. The drain voltage increases linearly with output until reaching

supply voltage VDD. The instantaneous efficiency at this point (–6 dB from PEP) is therefore the 78.5 percent of the ideal class-B PA. As the signal amplitude increases into the medium-power region, the carrier PA saturates and the peaking PA becomes active. The additional

current I2 sent to the load by the Figure 34 · Signal vectors in out- Figure 35 · Efficiency of outphasing peaking PA causes the apparent load phasing. transmitters with ideal class-B PAs. impedance at VL to increase above

44 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

operates at peak efficiency peak to be shifted leftward efficiency, but so that the average efficiency is delivers an in- increased for signals with higher creasing amount peak-to-average ratios. For example, of power. At PEP α = 0.36 results in a 60 percent aver- output, both PAs age efficiency for a Rayleigh-envelope see 50-ohm loads signal with a 10-dB peak-to-average and each delivers ratio, which is a factor of 2.1 improve- half of the system ment over class B. Doherty transmit- output power. The ters with unequal power division can PEP efficiency is be implemented by using different that of the class-B PEP load impedances and different PAs. supply voltages in the two PAs [97]. The resulting Much recent effort has focused on Figure 37 · Doherty transmitter. instantaneous- accommodating non-ideal effects efficiency curve is (e.g., nonlinearity, loss, phase shift) shown in Figure into a Doherty architecture [93, 94, the 25 ohms of the low-power region. 39. The classical power division (α = 95]. The power consumed by the qui- Transformation through the quarter- 0.5) approximately maximizes the escent current of the peaking amplifi- wavelength line results in a decrease average efficiency for full-carrier AM er is also a concern. The measured in the load presented to the carrier signals, as well as modern single-car- ACPR characteristics of an S-band PA. The carrier PA remains in satu- rier digital signals. The use of other Doherty transmitter are compared to ration and acts as a voltage source. It power-division ratios allows the lower those of quadrature-combined class- B PAs in Figure 40. The signal is IS- 95 forward link with pilot channel, paging channel, and sync-channel. The PAs are based upon 50-W LDMOS transistors. Back-off is varied to trade-off linearity against output. For the specified ACPR of –45 dBc, the average PAE is nearly twice that of the quadrature-combined PAs. In a modern implementation, DSP can be used to control the drive and bias to the two PAs, for precise control and higher linearity. It is also Figure 38 · Ideal voltage and current relationships in Doherty transmitter. possible to use three or more stages to keep the instantaneous efficiency relatively high over a larger dynamic range [96, 98]. For ideal class-B PAs, the average efficiency of a three-stage Doherty can be as high as 70 percent for a Rayleigh-envelope signal with 10-dB peak-to-average ratio.

References 60. S. C. Cripps, RF Power Amplifiers for Wireless Communication, Norwood, MA: Artech, 1999. 61. J. M. Rollett, “Stability and power- gain invariants of linear twoports,” IRE Trans. Circuit Theory, pp, 29-32, March Figure 39 Instantaneous efficiency Figure 40 · Measured ACPR perfor- 1962. of the Doherty system with ideal mance of an S-band Doherty trans- 62. S. Cohn, “A class of broadband class-B PAs. mitter. three-port TEM modes hybrids,” IEEE Trans. Microwave Theory Tech., vol. MTT-

46 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

16, no. 2, pp. 110-116, 1968. IEEE Trans. Microwave Theory Tech., vol. 63. J. Goel, “K band GaAsFET amplifi- Acronyms Used in Part 3 47, no. 8, pp. 1467-1470 , Aug. 1999. er with 8.2 W output using a radial power EER Envelope Elimination and 89. F. H. Raab, “Efficiency of outphas- combiner,” IEEE Trans. Microwave Theory Restoration ing power-amplifier systems,” IEEE Tech., vol. MTT-32, no. 3, pp. 317-324, Trans. Commun., vol. COM-33, no. 10, pp. 1984. AM Amplitude Modulation 1094-1099, Oct. 1985. 64. A. Berr and D. Kaminsky, “The LINC Linear Amplification with 90. B. Stengel and W. R. Eisenstat, travelling wave power divider/combiner,” “LINC power amplifier combiner efficien- IEEE Trans. Microwave Theory Tech., vol. Nonlinear Components cy optimization,” IEEE Trans. Veh. MTT-28, no.12, pp. 1468-1473, 1980. Technol., vol. 49, no. 1, pp. 229- 234, Jan. 65. F. H. Raab, “Hybrid and quadra- 814, Anaheim, CA, June 1999. 2000. ture splitters and combiners,” Research 78. M. D. Weiss, F. H. Raab, and Z. B. 91. C. P. Conradi, R. H. Johnston, and Note RN97-22, Green Mountain Radio Popovic, “Linearity characteristics of X- J. G. McRory, “Evaluation of a lossless Research Company, Colchester, VT, Sept. band power amplifiers in high-efficiency combiner in a LINC transmitter,” Proc. 20, 1997. transmitters,” IEEE Trans. Microwave 1999 IEEE Canadian Conf. Elec. and 66. R. A. York and Z. B. Popovic, Active Theory Tech., vol. 47, no. 6, pp. 1174-1179, Comp. Engr., Edmonton, Alberta, Canada, and Quasi-Optical Arrays for Solid-State June 2001. pp. 105-109, May 1999. Power Combining, New York: Wiley, 1997. 79. F. H. Raab, “Technique for wide- 92. W. H. Doherty, “A new high effi- 67. S. Brozovich, “High efficiency mul- band operation of power amplifiers and ciency power amplifier for modulated tiple power level amplifier circuit,” U.S. combiners,” U. S. Patent 6,252,461, June waves,” Proc. IRE, vol. 24, no. 9, pp. 1163- Patent 5,661,434, Aug. 1997. 26, 2001. 1182, Sept. 1936. 68. J. Sevic, “Efficient parallel stage 80. J. Staudinger et al., “High efficien- 93. D. M. Upton and P. R. Maloney, “A amplifier,” U.S. Patent 5,872,481, Feb. cy CDMA RF power amplifier using new circuit topology to realize high effi- 1999. dynamic envelope tracking technique,” ciency, high linearity, and high power 69. F. H. Raab, “Average efficiency of Int. Microwave Symp. Digest, vol. 2, pp. microwave amplifiers,” Proc. Radio and Class-G power amplifiers,” IEEE Trans. 873-876, Boston, MA, June 13-15, 2000. Wireless Conf. (RAWCON), Colorado Consumer Electronics, vol. CE-32, no. 2, 81. A. A. M. Saleh and D. C. Cox, Springs, pp. 317-320, Aug. 9-12, 1998. pp. 145-150, May 1986. “Improving the power-added efficiency of 94. J. Schuss et al, “Linear amplifier 70. J. Staudinger, “Applying switched FET amplifiers operating with varying- for high efficiency multi-carrier perfor- gain stage concepts to improve efficiency envelope signals,” IEEE Trans. Microwave mance,” U.S. Patent 5,568,086, Oct. 1996. and linearity for mobile CDMA power Theory Tech., vol. 31, no. 1, pp. 51-55, Jan. 95. J. Long, “Apparatus and method amplification,” Microwave J., vol. 43, no. 9, 1983. for amplifying a signal,” U.S. Patent pp. 152-162, Sept. 2000. 82. B. D. Geller, F. T. Assal, R. K. 5,886,575, March 1999. 71. “30 way radial combiner for minia- Gupta, and P. K. Cline, “A technique for 96. F. H. Raab, “Efficiency of Doherty ture GaAsFET power applications” the maintenance of FET power amplifier RF-power amplifier systems,” IEEE 72. T. C. Choinski, “Unequal power efficiency under backoff,” IEEE 1989 Trans. Broadcasting, vol. BC-33, no. 3, pp. division using several couplers to split MTT-S Digest, Long Beach, CA, pp. 949- 77-83, Sept. 1987. and recombine the output,” IEEE Trans. 952, June 1989. 97. M. Iwamoto et al., “An extended Microwave Theory Tech., vol. MTT-32, 83. G. Hanington, P.-F. Chen, P. M. Doherty amplifier with high efficiency no.6, pp. 613-620, 1984. Asbeck, and L. E. Larson, “High-efficiency over a wide power range,” Int. Microwave 73. L. R. Kahn, “Single sideband power amplifier using dynamic power- Symp. Digest, Phoenix, AZ, pp. 931-934, transmission by envelope elimination and supply voltage for CDMA applications,” May 2001. restoration,” Proc. IRE, vol. 40, no. 7, pp. IEEE Trans. Micro-wave Theory Tech., vol. 98. B. E. Sigmon, “Multistage high 803-806, July 1952. 47, no. 8, pp. 1471-1476, Aug. 1999. efficiency amplifier,” U.S. Patent 74. F. H. Raab and D. J. Rupp, “High- 84. G. Hanington, A. Metzger, P. 5,786,938, July 28, 1998. efficiency single-sideband HF/ VHF trans- Asbeck, and H. Finlay, “Integrated dc-dc mitter based upon envelope elimination converter using GaAs HBT technology,” Author Information and restoration,” Proc. Sixth Int. Conf. HF Electronics Letters, vol. 35, no. 24, p.2110- Radio Systems and Techniques (HF ’94) 2112, 1999. The authors of this series of arti- (IEE CP 392), York, UK, pp. 21-25, July 4 85. D. R. Anderson and W. H. Cantrell, cles are: Frederick H. Raab (lead - 7, 1994. “High-efficiency inductor-coupled high- author), Green Mountain Radio 75. F. H. Raab, B. E. Sigmon, R. G. level modulator,” IMS ’01 Digest, Phoenix, Research, e-mail: [email protected]; Myers, and R. M. Jackson, “L-band trans- AZ, May 2001. Peter Asbeck, University of California mitter using Kahn EER technique,” IEEE 86. H. Chireix, “High power outphas- at San Diego; Steve Cripps, Hywave Trans. Microwave Theory Tech., pt. 2, vol. ing modulation,” Proc. IRE, vol. 23, no. 11, 46, no. 12, pp. 2220-2225, Dec. 1998. pp. 1370-1392, Nov. 1935. Associates; Peter B. Kenington, 76. F. H. Raab, “Intermodulation dis- 87. D. C. Cox and R. P. Leck, “A VHF Andrew Corporation; Zoya B. Popovic, tortion in Kahn-technique transmitters,” implementation of a LINC amplifier,” University of Colorado; Nick IEEE Trans. Microwave Theory Tech., vol. IEEE Trans. Commun., vol. COM-24, no. Pothecary, Consultant; John F. Sevic, 44, no. 12, part 1, pp. 2273-2278, Dec. 9, pp. 1018-1022, Sept. 1976. California Eastern Laboratories; and 1996. 88. R. Langridge, T. Thornton, P. M. 77. F. H. Raab, “Drive modulation in Asbeck, and L. E. Larson, “A power re-use Nathan O. Sokal, Design Automation. Kahn-technique transmitters,” Int. technique for improving efficiency of out- Readers desiring more information Microwave Symp. Digest, vol. 2, pp. 811- phasing microwave power amplifiers,” should contact the lead author.

RF and Microwave Power Amplifier and Transmitter Technologies — Part 4

By Frederick H. Raab, Peter Asbeck, Steve Cripps, Peter B. Kenington, Zoya B. Popovich, Nick Pothecary, John F. Sevic and Nathan O. Sokal

inearization tech- Linearization methods are niques are incorpo- the focus of Part 4 of our Lrated into power series on power amplifiers, amplifiers and transmit- which describes the basic ters for the dual purposes architecture and perfor- of improving linearity mance capabilities of feed- and for allowing opera- back, feedforward and tion with less back-off predistortion techniques and therefore higher effi- Fig 41 · Envelope feedback applied to a ciency. This article pro- complete transmitter. vides a summary of the three main families of techniques have been developed: Feedback, feedforward, and predistortion. stage. The gain in the feedback path reduces the power dissipated in the feedback compo- 8a. FEEDBACK nents. While such systems demonstrate IMD Feedback linearization can be applied reduction [105], they tend to work best at a either directly around the RF amplifier (RF specific signal level. feedback) or indirectly upon the modulation (envelope, phase, or I and Q components). Envelope Feedback The problem of delay in RF feedback is RF Feedback alleviated to a large extent by utilizing the The basis of this technique is similar to its signal envelope as the feedback parameter. audio-frequency counterpart. A portion of the This approach takes care of in-band distortion RF-output signal from the amplifier is fed products associated with amplitude nonlin- back to, and subtracted from, the RF-input earity. Harmonic distortion products, which signal without detection or down- conversion. are corrected by RF feedback, are generally Considerable care must be taken when using not an issue as they can easily be removed by feedback at RF as the delays involved must be filtering in most applications. Envelope feed- small to ensure stability. In addition, the loss back is therefore a popular and simple tech- of gain at RF is generally a more significant nique. sacrifice than it is at audio frequencies. For Envelope feedback can be applied to either these reasons, the use of RF feedback in dis- a complete transmitter (Figure 41) or a single crete circuits is usually restricted to HF and power amplifier (Figure 42). The principles of lower VHF frequencies [99]. It can be applied operation are similar and both are described within MMIC devices, however, well into the in detail in [100]. The RF input signal is sam- microwave region. pled by a coupler and the envelope of the input In an active RF feedback system, the volt- sample is detected. The resulting envelope is age divider of a conventional passive-feedback then fed to one input of a differential amplifi- system is replaced by an active (amplifier) er, which subtracts it from a similarly

38 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 42 · Envelope feedback applied to an RF power amplifier. obtained sample of the RF output. The difference signal, Figure 43 · Block diagram of a polar-loop transmitter. representing the error between the input and output envelopes, is used to drive a modulator in the main RF path. This modulator modifies the envelope of the RF sig- erally different bandwidths required for the amplitude nal which drives the RF PA. The envelope of the resulting and phase feedback paths. Thus, differing levels of output signal is therefore linearized to a degree deter- improvement of the AM-AM and AM-PM characteristics mined by the loop gain of the feedback process. Examples usually result, and this often leads to a poorer overall per- of this type of system are reported in [101] and [102]. formance than that achievable from an equivalent The degree of linearity improvement that can be Cartesian-loop transmitter. A good example of the differ- obtained when using this technique depends upon the rel- ence occurs with a standard two-tone test, which causes ative levels of the AM-AM and AM-PM conversion in the the phase-feedback path to cope with a discontinuity at amplifier. For a VHF BJT amplifier, AM-AM distortion is the envelope minima. In general, the phase bandwidth dominant and two-tone IMD is typically reduced by 10 must be five to ten times the envelope bandwidth, which dB. Since AM-PM distortion is not corrected by envelope limits available loop gain for a given delay. feedback, no linearity improvement is observed if phase distortion is the dominant form of nonlinearity. This is Cartesian Feedback often the case in, for example, class-C and LDMOS PAs. The Cartesian-feedback technique overcomes the The use of envelope feedback is therefore generally problems associated with the wide bandwidth of the sig- restricted to relatively linear class-A or AB amplifiers. nal phase by applying modulation feedback in I and Q (Cartesian) components [104]. Since the I and Q compo- Polar-Loop Feedback nents are the natural outputs of a modern DSP, the The polar-loop technique overcomes the fundamental Cartesian loop is widely used in PMR and SMR systems. inability of envelope feedback to correct for AM-PM distortion effects [103]. Essentially, a phase-locked loop is added to the envelope feedback system as shown in Figure 43. For a narrowband VHF PA, the improvement in two-tone IMD is typically around 30 dB. The envelope- and phase-feedback functions operate essentially independently. In this case, envelope detection occurs at the intermediate frequency (IF), as the input signal is assumed to be a modulated carrier at IF. Likewise, phase detection takes place at the IF, with limiting being used to minimize the effects of signal amplitude upon the detected phase. Alternatively, it is possible to supply the envelope and phase modulating signals sep- arately at baseband and to undertake the comparisons there. The key disadvantage of polar feedback lies in the gen- Figure 44 · Cartesian-loop transmitter configuration.

40 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 45 · Linearization of a class-C PA by Figure 46 · Block diagram of a feed-forward transmitter in its Cartesian feedback (courtesy WSI). basic form.

The basic Cartesian loop (Figure 44) consists of two ting/combining, and the delay lines ensure operation over identical feedback processes operating independently on a wide bandwidth. Loop-control networks, which consist the I and Q channels. The inputs are applied to differen- of amplitude- and phase-shifting networks, maintain sig- tial integrators (in the case of a first-order loop) with the nal and distortion cancellation within the various feed- resulting difference (error) signals being modulated onto forward loops. I and Q subcarriers and up-converted to drive the PA. A The input signal is first split into two paths, with one sample of the output from the PA is attenuated and path going to the high-power main amplifier while the quadrature-down-converted (synchronously with the up- other signal path goes to a delay element. The output sig- conversion process). The resulting quadrature feedback nal from the main amplifier contains both the desired sig- signals then form the second inputs to the input differen- nal and distortion. This signal is sampled and scaled tial integrators, completing the two feedback loops. The using attenuators before being combined with the delayed phase shifter shown in the up-converter local-oscillator portion of the input signal, which is regarded as distor- path is used to align the phases of the up- and down-con- tion-free. The resulting “error signal” ideally contains version processes, thereby ensuring that a negative feed- only the distortion components in the output of the main back system is created and that the phase margin of the amplifier. The error signal is then amplified by the low- system is optimized. power, high-linearity error amplifier, and then combined The effects of applying Cartesian feedback to a highly with a delayed version of the main amplifier output. This nonlinear (class-C) PA amplifying an IS-136 (DAMPS) second combination ideally cancels the distortion compo- signal are shown in Figure 45. The first ACPR is nents in the main-amplifier output while leaving the improved by 35 dB and the signal is produced within desired signal unaltered. specifications with an efficiency of 60 percent [100]. In practice, there is always some residual desired signal passing through the error amplifier. This is in gener- 8b. FEEDFORWARD al not a problem unless the additional power is sufficient The very wide bandwidths (10 to 100 MHz) required in in magnitude to degrade the linearity of the error ampli- multicarrier applications can render feedback and DSP fier and hence the linearity of the feedforward transmit- impractical. In such cases, the feedforward technique can ter. be used to achieve ultra-linear operation. In its basic configuration, feedforward typically gives improvements in Signal Cancellation distortion ranging from 20 to 40 dB. Successful isolation of an error signal and the removal of distortion components depend upon precise signal can- Operation cellation over a band of frequencies. In practice, cancella- In its basic form (Figure 46), a feedforward amplifier tion is achieved by the vector addition of signal voltages. consists of two amplifiers (the main and error amplifiers), The allowable amplitude and phase mismatches for dif- directional couplers, delay lines and loop control networks ferent cancellation levels are shown in Figure 47. For [110]. The directional couplers are used for power split- manufactured equipment, realistic values of distortion

42 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 48 · Feedforward performance with mixed- Figure 47 · Gain/phase matching requirements. mode modulation (TDMA and CDMA signals). cancellation are around 25 to 30. The limiting factor is changes in device characteristics over time, temperature, nearly always the bandwidth over which a given accura- voltage and signal level degrade the amplitude and phase cy can be obtained. matching and therefore increase distortion in the transmitter output. An automatic control scheme continuously Efficiency adjusts the gain and phase to achieve the best signal can- The outputs of the main and error amplifiers are typ- cellation and output linearity. The first step is to use FFT ically combined in a directional coupler that both isolates techniques, direct power measurement, or pilot signals to the PAs from each other and provides resistive input determine how well the loop is balanced. Both digital and impedances. For a typical 10 dB coupling ratio, 90 percent analog techniques can be used for loop control and adjust- of the power from the main PA reaches the output. For the ment. Signal processing can be used to reduce the peaks same coupling ratio, only 10 percent of the power from the in multi-carrier signals and to keep distortion products error amplifier reaches the load, thus the error amplifier out of the nearby receiving band [111]. must produce ten times the power of the distortion in the main amplifier. The peak-to-average ratio of the error sig- Performance nal is often much higher than that of the desired signal, An example of the use of feedforward to improve lin- making amplification of the error signal inherently much earity is shown in Figure 48. The signal consists of a mix less efficient than that of the main signal. As a result, the of TDMA and CDMA carriers. The power amplifiers are power consumed by the error amplifier can be a signifi- based upon LDMOS transistors and have two-tone IMD cant fraction (e.g., one third) of that of the main amplifi- levels in the range –30 to –35 dBc at nominal output er. In addition, it may be necessary to operate one or both power. The addition of feedforward reduces the level of amplifiers well into back-off to improve linearity. The distortion by approximately 30 dB to meet the required overall average efficiency of a feedforward transmitter levels of better than –60 dBc. The average efficiency is may therefore be only 10 to 15 percent for typical multi- typically about 10 percent. carrier signals. 8c. PREDISTORTION Automatic Loop Control The basic concept of a predistortion system (Figure 49) Since feedforward is inherently an open-loop process, involves the insertion of a nonlinear element prior to the

Figure 49 · Predistortion concept. Figure 50 · Amplitude correction by predistortion.

44 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 51 · An RF predistorter.

RF PA such that the combined transfer characteristic of both is linear (Figure 50). Predistortion can be accomplished at either RF or baseband. Figure 52 · Linearization by diode-based RF predistorter (courtesy WSI). RF Predistortion The block diagram of a simple RF predistorter is shown in Figure 51. A compressive characteristic, created Parametric Linearization (APL®), which is capable of by the nonlinearity in the lower path (e.g., a diode) is sub- multi-order correction [106]. Most RF-predistortion tech- tracted from a linear characteristic (the upper path) to niques are capable of broadband operation with practical generate an expansive characteristic. The output of the operational bandwidths similar to, or greater than, those linear path (typically just a time delay) is given by: of feedforward.

vl(vin) = a1vin (1) Digital Predistortion Digital predistortion techniques exploit the consider- and that of the compressive path is given by able processing power now available from DSP devices, which allows them both to form and to update the 3 vc(vin) = a2vin – bvin (2) required predistortion characteristic. They can operate with analog-baseband, digital-baseband, analog-IF, digi- Subtracting the above equations gives tal-IF, or analog-RF input signals. Digital-baseband and digital-IF processing are most common. 3 vpd(vin) = (a2 – a2)vin – bvin (3) The two most common types of digital predistorter are termed mapping predistorters [107] and constant-gain This is now an expansive characteristic with a linear gain of a1 – a2, and may be used to predistort a compressive amplifier characteristic (cubic in this example) by appropriate choice of a1, a2 and b. An example of the results from using a simple diode- based RF predistorter with a 120-W LDMOS PA amplifying an IS-95 CDMA signal is shown Figure 52. When applied to π/4-DQPSK modulation in a satellite application, the same predistorter roughly halves the EVM, improves the efficiency from 22 to 29 percent, and doubles the available output power. Predistortion bandwidths tend to be limited by similar factors to that of feedforward, namely gain and phase flatness of the predistorter itself and of the RF PA. In addition, memory effects in the PA and the predistorter limit the degree cancellation, and these tend to become poorer with increasing bandwidth. Better performance can be achieved with more complex forms of RF predistortion such as Adaptive Figure 53 · Mapping predistorter.

46 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

2.0

1.8

1.6

1.4 2 1.2

1.0

0.8

0.6 Output envelope 0.4

0.2

0 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 Input envelope2

Figure 55 · Linearization of the amplitude transfer characteristic using an RF input/output digital predistorter Figure 54 · Constant-gain predistorter. (courtesy WSI). predistorters [108]. A mapping predistorter utilizes two overall transfer characteristic is then linear: look-up tables, each of which is a function of two variables

(IIN and QIN), as shown in Figure 53. This type of predis- GPD(IIN(t),QIN(t))×GPA(IPD(t),QPD(t)) = k (4) torter is capable of excellent performance. However, it requires a significant storage and/or processing overhead An example of the improvement in the amplitude- for the look-up tables and their updating mechanism, and transfer characteristic by an RF-input/output digital pre- has a low speed of convergence. The low convergence distorter [109] is shown in Figure 55. The plot is based speed results from the need to address all points in the upon real-time using samples from a GSM-EDGE signal. I/Q complex plane before convergence can be completed. Both the gain expansion and compression are improved A constant-gain predistorter (Figure 54) requires only by the linearizer. EVM is reduced from around 4.5 to 0.7 a single-dimensional look-up table, indexed by the signal percent. The ACPR for IS-136 DAMPS modulation (π/4- envelope. It is therefore a much simpler implementation DQPSK) is reduced by nearly 20 dB (Figure 56). When and requires significantly less memory for a given level of generating mask-compliant EDGE modulation at full out- performance and adaptation time. It uses the look-up put power (850-900 MHz), the linearized PA has an effi- table to force the predistorter and associated PA to exhib- ciency of over 30 percent. it a constant gain and phase at all envelope levels. The An example of linearization of a PA with two 3G W-

Figure 56 · Linearization of DAMPS PA by RF input/out- Figure 57 · Linearization of 3G W-CDMA PA signal by put predistorter (courtesy WSI). digitial baseband input predistorter (courtesy WSI).

48 High Frequency Electronics 104. V. Petrovic, “Reduction of spurious emission from radio transmitters by means of modulation feedback,” Proc. IEE Conf. No. 224 on Radio Spectrum Conservation Techniques, UK., Sept. 6-8 1983. 105. E. Ballesteros, F. Perez, and J. Peres, “Analysis and design of microwave linearized amplifiers using active feedback,” IEEE Trans. Microwave Theory Tech., vol. 36, no. 3, pp. 499-504, March 1988. 106. P. B. Kenington, “Achieving high-efficiency in multi-carrier base-station power amplifiers,” Microwave Engr. Europe, pp. 83-90. Sept. 1999. 107. Y. Nagata, “Linear amplification techniques for digital mobile communications,” Proc. IEEE Veh. Tech. Conf. (VTC ’89), San Fransisco, pp. 159-164, May 1-3, 1989. 108. J. K. Cavers, “Amplifier linearisation using a digital predistorter with fast adaptation and low memory requirements,” IEEE Trans. Veh. Tech., vol. 39, no. 4, pp. 374-382, Nov. 1990. 109. P. B. Kenington, M. Cope, R. M. Bennett, and J. Bishop, “GSM-EDGE high power amplifier utilising digital linearisation,” IMS’01 Digest, Phoenix, AZ, May 20-25, 2001. 110. N. Pothecary, Feedforward Linear Power Amplifiers, Norwood, MA: Artech, 1999. Figure 58 · A multi-carrier S-band transmitter with digi- 111. J. Tellado, Multicarrier Modulation with Low tal predistorter (courtesy WSI). PAR, Boston: Kluwer, 2000.

Author Information CDMA signals by a digital baseband-input predistorter is The authors of this series of articles are: Frederick H. shown in Figure 57. The linearized amplifier meets the Raab (lead author), Green Mountain Radio Research, e- required spectral mask with a comfortable margin at all mail: [email protected]; Peter Asbeck, University of frequency offsets. The noise floor is set by the degree of California at San Diego; Steve Cripps, Hywave Associates; clipping employed on the waveform, which limits the Peter B. Kenington, Andrew Corporation; Zoya B. Popovic, ACPR improvement obtained. It clearly demonstrates, University of Colorado; Nick Pothecary, Consultant; John however, that digital predistortion can be used in broad- F. Sevic, California Eastern Laboratories; and Nathan O. band as well as narrowband applications. Figure 58 Sokal, Design Automation. Readers desiring more infor- shows a 3G transmitter that uses digital predistortion. mation should contact the lead author.

References Acronyms Used in Part 4 99. A. F. Mitchell, “A 135 MHz feedback amplifier,” IEE ACPR Adjacent Channel Power Ratio Colloq. Broadband High Frequency Amplifiers: Practice APL Adaptive Parametric Linearization and Theory, pp. 2/1-2/6, London, Nov. 22, 1979 BER Bit Error Rate 100. P. B. Kenington, High Linearity RF Amplifier DAMPS Digital American Mobile Phone System Design, Norwood, MA: Artech, 2000. EDGE Enhanced Data for GSM Evolution 101. W. B. Bruene, “Distortion reducing means for sin- EVM Error Vector Magnitude gle-sidedband transmitters,” Proc. IRE, vol. 44, no. 12, pp. IF Intermediate Frequency 1760-1765, Dec. 1956. LDMOS Laterally Diffused Metal Oxide Semiconductor 102. T. Arthanayake and H. B. Wood, “Linear amplifi- PA Power Amplifier cation using envelope feedback,” Electronics Letters, vol. PDF Probability-Density Function 7, no. 7, pp. 145-146, April 8, 1971. PMR Private Mobile Radio 103. V. Petrovic and W. Gosling, “Polar-loop transmit- SMR Specialized Mobile Radio ter,” Electronics Letters, vol. 15, no. 10, pp. 286-287, May W-CDMA Wideband Code-Division Multiple Access 10, 1979.

RF and Microwave Power Amplifier and Transmitter Technologies — Part 5

By Frederick H. Raab, Peter Asbeck, Steve Cripps, Peter B. Kenington, Zoya B. Popovich, Nick Pothecary, John F. Sevic and Nathan O. Sokal

The ever-increasing Emerging techniques are demands for more band- examined in this final width, coupled with installment of our series on requirements for both power amplifier technolo- high linearity and high gies, providing notes on efficiency create ever- new modulation methods increasing challenges in and improvements in the design of power linearity and efficiency amplifiers and transmitters. A single W-CDMA signal, for example, taxes the capabilities of a Kahn-technique transmitter with a conventional class-S modulator. More acute are the problems in base-station and satellite transmitters, where multiple carriers must be amplified simultaneously, resulting in peak- to-average ratios of 10 to 13 dB and band- Figure 59 · RF pulse-width modulation. widths of 30 to 100 MHz. A number of the previously discussed techniques can be applied to this problem, including conveys signal phase information as in the the Kahn EER with class-G modulator or split- Kahn and other techniques. band modulator, Chireix outphasing, and Radio-frequency pulse-width modulation Doherty. This section presents some emerging (RF PWM) eliminates the series-pass losses technologies that may be applied to wideband, associated with the class-S modulator in a high efficiency amplification in the near future. Kahn-technique transmitter. More important- ly, the spurious products associated with RF Pulse-Width Modulation PWM are located in the vicinity of the har- Variation of the duty ratio (pulse width) of monics of the carrier [113] and therefore easi- a class-D RF PA [112] produces an amplitude- ly removed. Consequently, RF PWM can modulated carrier (Figure 59). The output accommodate a significant RF bandwidth envelope is proportional to the sine of the with only a simple, low-loss output filter. pulse width, hence the pulse width is varied in Ideally, the efficiency is 100 percent. In proportion to the inverse sine of the desired practice, switching losses are increased over envelope. This can be accomplished in DSP, or those in a class-D PA with a 50:50 duty ratio by comparison of the desired envelope to a because drain current is nonzero during full-wave rectified sinusoid. The pulse timing switching.

46 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 60 · Current-switching PA for 1 GHz (courtesy Figure 61 · Prototype class-D PA for delta-sigma mod- UCSD). ulation (courtesy UCSD).

Previous applications of RF PWM have been limited to average of the cycles in the PA. Phase is again conveyed LF and MF transmitters (e.g., GWEN [114]). However, the in pulse timing. recent development of class-D PAs for UHF and The delta-sigma modulator employs an algorithm microwave frequencies (Figure 60) offers some interesting such as that shown in Figure 63. The signal is digitized by possibilities. a quantizer (typically a single-bit comparator) whose output is subtracted from the input signal through a digital Delta-Sigma Modulation feedback loop, which acts as a band-pass filter. Basically, Delta-sigma modulation is an alternative technique the output signal in the pass band is forced to track the for directly modulating the carrier produced by a class-D desired input signal. The quantizing noise (associated RF PA (Figure 61) [PA8],[PA9]. In contrast to the basical- with the averaging process necessary to obtain the ly analog operation of RF PWM, delta-sigma modulation desired instantaneous output amplitude) is forced outside drives the class-D PA at a fixed clock rate (hence fixed of the pass band. pulse width) that is generally higher than the carrier fre- The degree of suppression of the quantization noise quency (Figure 62). The polarity of the drive is toggled as depends on the oversampling ratio; i.e., the ratio of the necessary to create the desired output envelope from the digital clock frequency to the RF bandwidth and is relatively independent of the RF center frequency. An example of the resultant spectrum for a single 900-MHz carrier and 3.6-GHz clock is shown in Figure 64. The quantization noise is reduced over a bandwidth of 50 MHz, which is sufficient for the entire cellular band. Out-of- band noise increases gradually and must be removed by a band-pass filter with sufficiently steep skirts. As with RF PWM, the efficiency of a practical delta- sigma modulated class-D PA is reduced by switching losses associated with nonzero current at the times of switching. The narrow-band output filter may also introduce significant loss.

Carrier Pulse-Width Modulation Carrier pulse-width modulation was first used in a UHF rescue radio at Cincinnati Electronics in the early 1970s. Basically, pulse-width modulation as in a class-S modulator gates the RF drive (hence RF drain current) on and off in bursts, as shown in Figure 65. The width of each Figure 62 · Delta-sigma modulation. burst is proportional to the instantaneous envelope of the

48 High Frequency Electronics Figure 63 · Delta-sigma modulator.

–30

–40

–50 Figure 65 · Carrier pulse-width modulation.

–60 Power Recovery –70 A number of RF processes result in significant RF power dissipated in “dump” resistors. Examples include –80 power reflected from a mismatched load and dumped by a

Power Spectral Density (dBm/Hz) circulator and the difference between two inputs of hybrid –90 combiner dumped to the balancing resistor. The notion of recovering and reusing wasted RF power was originally –100 applied to the harmonics (18 percent of the output power) 600 800 1000 1200 Frequency (MHz) of an untuned LF class-D PA [117]. More recently, power recovery has been applied to out- Figure 64 · Spectrum of delta-sigma modulation. phasing PAs with hybrid combiners [118, 119]. The instantaneous efficiency of such a system depends upon both the efficiency of the PA and that of the recovery sys- desired output. The amplitude-modulated output signal is tem. Since the two PAs operate at full power regardless of recovered by a band-pass filter that removes the side- the system output, inefficiency in the PA has a significant bands associated with the PWM switching frequency. The impact upon the system efficiency at the lower outputs. PWM signal can be generated by a comparator as in a Nonetheless, a significant improvement over convention- class-S modulator or by delta-sigma techniques. al hybrid-coupled outphasing is possible, and the PAs are As with RF PWM and delta-sigma modulation, the presented with resistive loads that allow them to operate series-pass losses and bandwidth limitations of the high- optimally. Typically, 50 percent of the dumped power can level modulator are eliminated. The switching frequency be recovered. in carrier PWM is not limited by capabilities of power- The power-recovery technology can also be used to switching devices and can therefore easily be tens of implement miniature DC-DC converters. Basically, a MHz, allowing large RF bandwidths. A second advantage high-efficiency RF-power amplifier (e.g., class-E) converts is that carrier PWM can be applied to almost any type of DC to RF and a high-efficiency rectifier circuit converts RF PA. A disadvantage is that a narrow-band output fil- the RF to DC at the desired voltage. Implementation at ter with steep skirts is required to remove the switching- microwave frequencies reduces the size of the tuning and frequency sidebands, and such filters tend to have losses filtering components, resulting in a very small physical of 1 to 2 dB at microwave frequencies. Nonetheless, the size and high power density. In a prototype that operates losses in the filter may be more than offset by the at C band [120], the class-E PA uses a single MESFET to improvement in efficiency for signals with high peak-to- produce 120 mW with a PAE of 86 percent. The diode rec- average ratios. tifier consists of a directional coupler with two Schottky

January 2004 49 High Frequency Design RF POWER AMPLIFIERS

Figure 66 · Switched PAs with quarter-wavelength transmission line combiner.

Figure 67 · Instantaneous efficiency of switched PAs. diodes connected at the coupled and through ports and has a 98-percent conversion efficiency and an overall efficiency (including mismatch loss) of 83 percent. For a typ- are active is the minimum needed to produce the current ical DC output of 3 V, the DC-DC conversion efficiency is output power. The peak power is thus kept relatively close 64 percent. to the saturated output, eliminating most of the effects of operating in back-off. The efficiency can therefore reach Switched PAs with Transmission-Line Combiners PEP efficiency at a number of different output levels, as RF-power amplifiers cannot simply be connected in shown in Figure 67. series or parallel and switched on and off to make a trans- The advantage of this technique is the ease in design mitter module that adapts to variable peak envelope associated with relying on short circuits rather than open power. Attempting to do so generally produces either lit- circuits to isolate the off-state PAs. A possible disadvan- tle effect or erratic variations in load impedance, some- tage is operating individual PAs from multiple load times leading to unstable operation and destruction of the impedances without retuning and a limited number of transistors. Systems of microwave PAs that are toggled power steps available (e.g., 9/9, 4/9, 1/9 for a three-PA sys- on and off are therefore connected through networks of tem). quarter-wavelength transmission lines. The Doherty transmitter (discussed in part 4 of this series) is a classic Electronic Tuning example of this sort of technique. The performance of virtually all power amplifiers is An alternative topology (Figure 66) uses shorting degraded by load- impedance mismatch. Mismatched switches and quarter-wavelength lines to to decouple off- loads not only reduce efficiency, but also create higher state PAs [121, 122]. The inactive PA is powered-down (by stresses on the transistors. Because high-efficiency PAs switching off its supply voltage), after which its output is generally require a specific set of harmonic impedances, shorted to ground. The quarter-wavelength line produces their use is often restricted to narrow-band applications an open circuit at the opposite end where the outputs with well-defined loads. from multiple PAs are connected together to the load. Electronic tuning allows frequency agility, matching of This technique may be more easy to implement (especial- unknown and variable loads, and amplitude modulation. ly for multiple PAs) than Doherty because a short is more Components for electronic tuning include pin-diode readily realized than an open. switches, MEMS switches, MEMS capacitors, semicon-

If PA #1 is the only PA active, its load is simply R0.If ductor capacitors, ceramic capacitors (e.g., BST), and bias- both #1 and #2 are active, the combination produces an controlled inductors. To date, electronic tuning has been effective load impedance of 2Ro at the load ends of the applied mainly to small-signal circuits such as voltage- lines. Inversion of this impedance through the lines controlled oscillators. Recently demonstrated, however, places loads of R0/2 on the RF PAs. Consequently, the are two electronically tuned power amplifiers. One oper- peak power output for two active PAs is four times that ates in class E, produces 20 W with an efficiency of 60 to with a single PA. As in discrete envelope tracking, the RF 70 percent, and can be tuned from 19 to 32 MHz (1.7:1 PAs operate as linear amplifiers. The number of PAs that range) through the use of voltage-variable capacitors

50 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

Figure 68 · Electronically tunable class-D PA (courtesy GMRR). Figure 70 · Load-modulated class-E PA (courtesy GMRR).

[123, 124]. The second (Figure 68) operates in class D, produces 100 W with an efficiency of 60 to 70 percent, and mum locus is found by examination of load-pull contours. can be tuned from 5 to 21 MHz (4.25:1 range) through the The simple T filter has a single electronically variable use of electronically tunable inductors and capacitors element, but provides an approximately optimum locus [125]. for class E over the top 12 dB of the dynamic range. The experimental 20-W, 30-MHz [124, 126] shown in Figure Load Modulation 70 achieves a 41-dB range of amplitude variation. The The output of a power amplifier can be controlled by measured instantaneous-efficiency curve (Figure 71) cor- varying the drive, gate bias, DC supply voltage, or load responds to a factor of 2.1 improvement in the average impedance. “Load modulation” uses an electronically efficiency for a Rayleigh-envelope signal with a 10-dB tuned output filter (Figure 69) to vary load impedance peak-to-average ratio. and thereby the instantaneous amplitude of the output A load-modulated PA for communications follows the signal. The modulation bandwidth can be quite wide, as it electronically tuned filter with a passive filter to remove is limited only by the bias feeds to the tuning components. the harmonics associated with the nonlinear elements. A key aspect of load modulation is a diligent choice of Predistortion compensates for the incidental phase mod- the impedance locus so that it provides both good dynam- ulation inherent in dynamic tuning of the filter. Variation ic range and good efficiency. For ideal saturated PAs of of the drive level can be used to conserve drive power and classes A, B, C, and F, the optimum locus is the pure resis- to extend the dynamic range. tance line on the Smith chart that runs from the nominal load to an infinite load. For ideal class-E PAs with series inductance and shunt susceptance for optimum operation with the nominal load, the optimum locus is the unity- efficiency line running from the nominal load upward and rightward at an angle of 65° [126]. For real PAs, the opti-

Figure 71 · Instantaneous efficiency of load modula- Figure 69 · Load modulation by electronic tuning. tion compared to class-B linear amplification.

52 High Frequency Electronics High Frequency Design RF POWER AMPLIFIERS

References ing scheme with optimum efficiency 112. Figures 8 and 9 in Part 2 of under variable outputs,” 2002 Int. tis series, High Frequency Microwave Symp. Digest, vol. 2, pp. Electronics, July 2003. 913-916, Seattle, WA, June 2-7, 2002. 113. F. H. Raab, “Radio frequency 123. F. H. Raab, “Electronically pulsewidth modulation,” IEEE Trans. tunable class-E power amplifier,” Int. Commun., vol. COM-21, no. 8, pp. Microwave Symp. Digest, Phoenix, 958-966, Aug. 1973. AZ, pp. 1513-1516, May 20-25, 2001. 114. F. G. Tinta, “Direct single 124. F. H. Raab, “Electronically sideband modulation of transmitter tuned power amplifier,” Patent pend- output switcher stages,” Proc. RF ing. Expo East ’86, Boston, MA, pp. 313- 125. F. H. Raab and D. Ruppe, 398, Nov. 10-12, 1986. “Frequency-agile class-D power 115. A. Jayaraman, P. F. Chen, G. amplifier,” Ninth Int. Conf. on HF Hanington, L. Larson, and P. Asbeck, Radio Systems and Techniques,pp. “Linear high-efficiency microwave 81-85, University of Bath, UK, June power amplifiers using bandpass 23-26, 2003. delta-sigma modulators,” IEEE 126. F. H. Raab, “High-efficiency Microwave Guided Wave Lett., vol. 8, linear amplification by dynamic load no. 3, pp. 121-123, March 1998. modulation,” Int. Microwave Symp. 116. J. Keyzer et al., “Generation Digest, vol. 3, pp. 1717-1720, of RF pulsewidth modulated Philadelphia, PA, June 8-13, 2003. microwave signals using delta-sigma modulation,” 2002 Int. Microwave Correction Symp. Digest, vol. 1, pp. 397-400, In Part 4 of this series (Novermber Seattle, WA, June 2-7, 2002. 2003 issue), Figures 45, 52, 55, 56, 57 117. J. D. Rogers and J. J. Wormser, and 58 should have been credited as “Solid-state high-power low frequency “Courtesy Andrew Corporation” telemetry transmitters,” Proc. NEC, instead of “Courtesy WSI.” vol. 22, pp. 171-176, Oct. 1966. 118. R. E. Stengel and S. A. Olson, Author Information “Method and apparatus for efficient The authors of this series of arti- signal power amplification,” U.S. cles are: Frederick H. Raab (lead Patent 5,892,395, Apr. 6, 1999. author), Green Mountain Radio 119. R. Langridge, T. Thornton, P. Research, e-mail: [email protected]; M. Asbeck, and L. E. Larson, “A power Peter Asbeck, University of California re-use technique for improving effi- at San Diego; Steve Cripps, Hywave ciency of outphasing microwave Associates; Peter B. Kenington, power amplifiers,” IEEE Trans. Andrew Corporation; Zoya B. Popovic, Microwave Theory Tech., vol. 47, no. 8, University of Colorado; Nick pp. 1467-1470 , Aug. 1999. Pothecary, Consultant; John F. Sevic, 120. S. Djukic, D. Maksimovic, and California Eastern Laboratories; and Z. Popovic, “A planar 4.5-GHz dc-dc Nathan O. Sokal, Design Automation. power converter,” IEEE Trans. Readers desiring more information Microwave Theory Tech., vol. 47, no. 8, should contact the lead author. pp. 1457-1460, Aug. 1999. 121. A. Shirvani, D. K. Su, and B. Acronyms Used in Part 5 A. Wooley, “A CMOS RF power amplifier with parallel amplification for BST Barium Strontium efficient power control,” IEEE J. Titanate Solid- State Circuits, vol. 37, no. 6, GWEN Ground Wave pp. 684-693, June 2002. Emergency Network 122. C. Y. Hang, Y. Wang, and T. PWM Pulse-Width Modulation Itoh, “A new amplifier power combin-

54 High Frequency Electronics