Wireless Transmitter Capabilities Through Supply Modulation

Pedro Miguel Cabral,1 Lorena Cabria,2 Filipe Rodrigues,1 Jose Angel Garcia,2 Jose Carlos Pedro1
1 2

Instituto de Telecomunicacoes, Universidade de Aveiro, Aveiro, Portugal Departamento de Ingenieria de Comunicaciones, Universidad de Cantabria, Santander, Spain

Received 29 June 2009; accepted 2 November 2009

ABSTRACT: In this overview article, the effective capabilities of power ampliers and wireless transmitters operating under supply modulation will be addressed. Special attention will be paid to: (i) identication of the circuit and system level limiting factors, (ii) possible ways of overcoming them, and (iii) analysis and implementation of one CDMA2000 polar transmitter. This is believed to offer the reader a comprehensive understanding of the overall transmitter behavior and of the underlying constraints that must be accounted C for. V 2010 Wiley Periodicals, Inc. Int J RF and Microwave CAE 20: 241251, 2010. Keywords: efficiency; envelope elimination and restoration; envelope tracking; linearity; polar transmitters; supply modulation

I. INTRODUCTION

The deployment of modern digital telecommunication systems, with continuously increasing capacity and using more and more complex modulation schemes, has demanded a steady improvement of the wireless communications transmitter (Tx) links. The classic architecture, presented in Figure 1, uses at least one power amplier (PA) conceived to convert the energy of a dc power supply into RF/microwave output power, [1]. This direct quadrature conversion is very exible supporting different modulation formats including nonconstant envelope modulation schemes. The overall Tx performance must be a carefully selected compromise between output power, gain, efciency, bandwidth, or even linearity. PAs dominate the power budget of RF Txs. Therefore, improving their efciency, by driving the active device into saturation, can signicantly enhance the overall Tx performance in terms of battery size and heat dissipation. On the other hand, amplitude and phase modulated signals, used for improved spectral efciency, have very high peak-to-average power ratios, imposing the use of very linear PAs. These are typically operated in backed-off class A or AB modes (to ensure that the envelope peaks

are within the PA linear operation regime), providing average efciencies below 30%. The compromise between power efciency and linearity has been one of the major design constrains of wireless communication systems. This is due to the fact that, while power efciency determines the talk time of wireless handsets or the network grid consumption of base-stations (in this latter case, not only it is important the waste of power as heat, but also the power needed to cool down the equipment), nonlinearity determines the signal quality specications, and thus the spectral usage efciency. In fact, nonlinear impairments of wireless circuits play a key role on both co-channel and adjacent-channel distortion. Co-channel distortion, along with the additive noise, impose a lower bound on the transmission channel bit-error-rate (BER), while adjacent-channel distortion sets a threshold on the attainable bit-error-rate of any potential adjacent-channel [1]. Two of the most important linearity specications of wireless systems are the adjacentchannel power ratio (ACPR) and the normalized mean square error (NMSE), also known as raw-error vector magnitude (EVM). These can be evaluated by: ACPRup
1 Bw 1 Bw

PkBw=2 Pk3Bw=2
kBw=2 kBw=2

jYkj2 jYkj2

(1a)

Correspondence to: P. M. Cabral; e-mail: pcabral@ua.pt DOI 10.1002/mmce.20427 Published online 6 January 2010 in Wiley InterScience (www. interscience.wiley.com).
C V 2010 Wiley Periodicals, Inc.

ACPRlow

2 kBw=2 jYkj PkBw=2 2 1 k3Bw=2 jYkj Bw 1 Bw

PkBw=2

(1b)

241

242

Cabral et al. we immediately conclude that theoretical 100% PAE and simultaneous perfect linearity (and thus ideal spectral confinement) can only be obtained if the supplied power, Pdc, is made proportional to the input power, Pin. This is the underlying theory behind the experiments that have recently being conducted with several types of supplied modulated PAs. In what the efciency is concerned, one should keep in mind that a 100% power efcient system can only be achieved with a 100% efcient PA and a 100% efcient supply modulator, (SM). Indeed, assuming PDC is the actual power delivered by the power supply to the SM, and Pdc is the power delivered from the SM to the PA, it can be easily shown that the overall system efciency, given by g Pout Pdc gPA gSM Pdc PDC (7)

Figure 1 Block diagram of a typical direct-conversion wireless communication transmitter link.

1

PN

NMSE N

^ n1 jyn ynj PN 2 1 y n1 j^nj N

(2)

^ where y(n) is the actual output, yn stands for the desired output, i.e., a linear scaled version of the input, |Y(k)|2 is the output power spectral density. This is integrated within the Bw ; Bw interval to give the signal output 2 2 power, or within the Bw ; 3Bw or the 3Bw ; Bw inter2 2 2 2 vals to provide the upper and lower adjacent channel output powers, respectively [1]. A simple argument, but surprisingly general, supporting the widely known knowledge that power efciency and spectral efciency (linearity) are two conicting requirements can be derived from the very basic energy conservation relationships of any electronic transducer as a PA. According to Figure 2, we immediately conclude that the sum of the powers delivered to the amplier (the signal input power, Pin, and the dc supplied power, Pdc) must equal the sum of the output powers (signal power delivered to the load, Pout, and dissipated power, Pdis): Pdc Pin Pout Pdis (3)

So, as the amplier is supposed to be active, it must deliver a signal gain given by: G Pout Pdc Pdis 1 Pin Pin (4)

which leads to the conclusion that, as the dissipated power must be non-negative, any PA dependent on a real power supply (i.e., one whose available power is finite) must evidence a compressing gain, and thus nonlinearity, after a certain level of increasing input signal power. Fortunately, recent advances in power amplier design have shown that this apparent physical limitation can be overcome as long as the available supply power is modulated according to the input signal amplitude envelope. Indeed, in [2] and [3], the authors have shown that if one takes into account the power added efciency (PAE) denition and the power relations of the PA in eq. (5): PAE which can be rewritten as: PAE G 1 Pin Pdc (6) Pout Pin Pdc (5)

is the product of the individual efficiencies of the PA and the SM. Just to have a concrete idea of what this means, we need to have both a PA and a SM with (not easy to get) efficiency values slightly higher than 70% to reach the nowadays wireless infrastructure base-stations goal of 50% overall efficiency. Under ideally linear current mode Class A or Class B, a power transducer (either the SM or the RF PA) can provide a theoretical maximum of gMax 50% or gMax 78.5%, respectively. However, using in the RF PA new wide-bandgap power HEMTs, operating at 28 V but suffering from a non-negligible knee-voltage of VK 6 V, this maximum RF PA efciency value becomes reduced to gMax 61.7%. Furthermore, when applying a real communications signal with non-constant envelope, the average efciency will be even more affected. As an example, considering a Class AB PA excited with a CDMA2000FW signal, the average efciency will be around 25.6%, validating the assumptions previously presented. Figure 3 shows the simulation results obtained for a Class AB PA: efciency vs. output voltage characteristic and output voltage probability density function (pdf), under CDMA2000FW signal.

Figure 2 Power relationships in a general electronic amplier. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

International Journal of RF and Microwave Computer-Aided Engineering/Vol. 20, No. 2, March 2010

Wireless Transmitter Capabilities

243

For example, if Sigma-Delta modulation is adopted for the one-bit digital coding, this digital representation accuracy, evaluated as the signal to quantization noise ratio, (SNR), is dependent on the ratio between the Nyquist sampling rate (two times the signal bandwidth) and the actual sampling rate, usually known as over-sampling ratio, (OSR), and the Mth-order quantization noise shaping (Mth-order loop-lter) according to [7]: SNR 6 2M 1 OSR2M1 p2M (8)

Figure 3 Class AB PA simulation results: efciency vs. output voltage characteristic and output voltage pdf function, under CDMA2000FW excitation. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley. com.]

So, a SNR on the order of 45 dB would require an unbearable OSR 5270, if no quantization noise shaping would be provided (i.e., M 0), an OSR 26 for M 1 and a much more manageable OSR 10 for M 2.
II. EMERGING WIRELESS TRANSMITTER ARCHITECTURES

Therefore, in order to reach the above desired efciency gures, both the RF PA and the SM, must operate in switched-mode. However, their design requirements are quite different in nature. So, it is worth spending time looking to the specicities of each one. In what the RF PA is concerned, and as it deals with phase modulated signals whose useful information is totally contained in the signals zero crossings, not in the actual waveform, traditional linear class AB current mode ampliers can be replaced by saturated class F, class D or class E PAs, [4, 5] whose efciencies are capable of fullling the desired values. Conversely, the SM handles a base-band amplitude signal, and so it can only operate in switched-mode if the desired amplitude waveform is perfectly coded as a high accuracy digital signal. This is something similar to the highly efcient switched-mode power supplies with variable reference (according to the input signal envelope) found in the so-called digital audio ampliers [6]. Unfortunately, since switched-mode operation can be thought as one-bit digital coding, one must conclude that the required amplitude accuracy, and thus amplitude resolution (high number of bits) determined by the quantization noise, can only be met at the expense of very high over-sampling ratios, imposing a tight compromise between efciency, distortion and switching speed. Indeed, a closer look into eqs. (1) and (2) indicates that the stringent linearity requirements of modern wireless systems (with NMSE and ACPR on the order of 30 to 40 dB and 40 to 45 dBc, respectively) can only be achieved with almost ideally transparent (i.e., linear and noise free) supply modulators. As a matter of fact, although easier to get than in a RF PA because of the much lower frequency involved (please remember that we are no longer processing a modulated RF carrier of several hundreds of MHz or GHz, but an amplitude envelope signal of several MHz), this still presents a tremendous challenge that has recently drawing the attention of an increasing number of researchers.

Nowadays, the RF power amplier, and the wireless transmitter in which it is applied, are both facing a change of paradigm. Not only the need to handle different communication standards has set a quest for recongurability, but also the desire to support higher and higher data rates has determined an increasing demand to overcome the always present spectral vs. power efciency compromise, already discussed in the previous section. Emerging wireless transmitter architectures have recently been receiving a growing attention by both industry and academia, [8, 9]. In comparison to the still more common I/Q, or Cartesian, transmitter, a polar topology, [10], derived from the envelope elimination and restoration (EER), or Kahn technique [11], can achieve a much better trade-off between nonlinear distortion and supply power efciency. In fact, as shown in the discussion regarding eqs. (5) and (6), the use of a highly efcient switched-mode RF power amplier, whose supply voltage is modulated by the input envelope amplitude, provides the polar transmitter the capability of achieving, at the same time, high linearity and efciency. As depicted in Figure 4, in the polar Tx architecture, the envelope signal is amplied through a power efcient bias modulator and then combined, at the RF transistor drain, with the phase-modulated RF carrier, this way

Figure 4 Envelope elimination and restoration, EER, system architecture.

International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce

244

Cabral et al. Proting from the evolution in digital techniques, both the amplitude and phase signals can be generated digitally. The EER and this digital polar Tx topology (see Figure 6) are very similar, except that the former operates the Cartesian-to-polar conversion in the analogue domain (via an envelope detector and an amplitude limiter, following the Cartesian I/Q modulator), while the second does that in the digital domain. Thus, it allows the independent generation of the amplitude and phase signals in the digital domain offering, therefore, a more exible way to digitally pre-distort these amplitude and phase signals. Moreover, a switched-mode PA can be used in both architectures, theoretically allowing a higher efciency when compared with the ET scheme [17]. As the aforementioned architecture suffers from some of the efciency problems mentioned for the EER system, there is also the possibility of allowing some amplitude variation in the PM branch, this way solving the RF power dissipation problem when the envelope goes to zero. Nowadays, there are already several practical implementations of some of the aforementioned transmitter architectures not only on the handset level but also in base stations both with ET and EER systems and with some hybrid approaches, proving that the theoretical predictions are correct and that these topologies can achieve high output powers (tenths of watts), delivered with very good efciency (>50%) and in a very linear way (ACPR > 45 dB), [1821].
III. SYSTEM LEVEL CONSTRAINTS IN POLAR ARCHITECTURES

Figure 5 Envelope tracking, ET, system architecture.

generating the desired amplitude (AM) and phase (PM) modulated signal at the Tx output. Unfortunately, as with any other engineering solution, this architecture does not follow an ideal distortion free operation, presenting a series of nonlinear impairment mechanisms at the device, circuit and system levels [1215]. In a pure EER system, the switched-mode PA should handle the bandwidth of a constant amplitude phasemodulated RF carrier which is much higher than the one of the original signal, modulated in both amplitude and phase. This problem can be solved by changing the PA from saturated switched-mode to linear current-mode and driving it with the original AM and PM modulated signal. This hybrid architecture, sometimes known as envelope tracking (ET) [16], is based on a conventional currentmode PA, whose biasing conditions (mainly the drain supply voltage) are dynamically adapted to the input driving envelope signal, operating the PA close to its maximum efciency condition, as illustrated Figure 5. Note, however, that, contrary to the EER in which the output amplitude directly results from the supply modulation, in an envelope tracking arrangement the output amplitude is already dened by the input signal. So, in this latter case, the PA should always be kept in its Pout/Pin linear mode, regardless of the supply voltage. Otherwise, a signicant even-order amplitude distortion will be generated. Clearly, the resulting average efciency of the PA stage will certainly be much higher when compared to a traditional pure current-mode operation (i.e., without supply voltage modulation). Furthermore, and despite a comparison between the ideal theoretical operation of the ET and EER systems would indicate that the latter would always be much more efcient than the former, the truth is that, in practice, they may be not that different. To understand why, we need to rst realize that keeping a very high power added efciency for a wide bandwidth signal, as the one driving the EER PA, is certainly more difcult than for the narrow band signal driving the ET PA. Second, a closer look into the practical operation of the EER PA reveals an inherent loss of PAE that does not affect the ET PA. Because the EER PA is driven by a constant amplitude signal, it dissipates input RF power even when the envelope is zero, and so, there is no output power.

Now, we will focus our attention in the polar architecture and evolve to the more detailed system level transmitter description shown in Figure 7. As almost everything in engineering, the theoretical ideal operation, predicted from these polar architectures, cannot be obtained in practice. The rst and most obvious issue that must be accounted for is the differential delay between the amplitude signal and the phase modulated carrier. As the AM and PM signals travel through two different paths, there will be an unavoidable differential delay, s, between

Figure 6 Polar system architecture. AM(t) and PM(t) stand for discrete-time digital representations of the desired AM and PM modulations, respectively, and LPF is a low-pass digital to analog reconstruction lter.

International Journal of RF and Microwave Computer-Aided Engineering/Vol. 20, No. 2, March 2010

Wireless Transmitter Capabilities

245

Figure 7

System model of a wireless Tx with polar architecture.

them. Thus, the analog amplitude signal, i.e., the supply voltage of the RF PA, will be of the form vDD(t) ax(ts), which leads to a reconstructed AM and PM modulated carrier whose modulating signal is y(t) ax(ts).ej/xt . Therefore, to obviate this form of nonlinear distortion (see Ref. 12 and 14 for the mathematical details), an equal delay must be introduced in the phase modulation path, so that, the AM and PM modulated signal can be a delayed, but undistorted, version of the original modulating signal: y(t) ax(ts).ej/xt Unfortunately, the SM will not be the only responsible for nonlinear distortion. The RF switched-mode PA is proved to present a nonlinear transfer function [vDD(t)-toay(t) conversion, [22], and carrier feed-through] that will also be responsible for a linearity degradation at the output [23]. An illustrative example of these is depicted in Figure 8. In fact, this gure presents the measured ay(vDD) function of a pHEMT based Class-E PA. Note the nonzero ay for vanishingly small vDD and the subsequent nonlinear ay(vDD) relation encountered for higher vDD values. The observed phenomenon, for the lowest end of vDD operation, is due to the carrier feed-through (consequence of the RF carrier leakage via the inputoutput capacitance). The ay(vDD) compression observed for the highest vDD values appears whenever the vDD is so large that the PA can no longer be kept in switched-mode and begins to

enter the current-mode operation. There, the output current (and thus the output RF signal amplitude) is no longer uniquely determined by the output voltage (typical of the triode or saturation regions of FETs and bipolar transistors, respectively) but becomes also controlled by the input voltage, a constant amplitude signal in polar Tx and EER PAs (as happens in the active region of both FETs and bipolars). Furthermore, as the envelope input power, Pin, results from processing the original complex (amplitude and phase) envelope, x(t) ax(t).ej/xt , through a strong evenorder nonlinearity, ax(t) abs[x(t)], this amplitude signal will have an innite bandwidth, even if x(t) has a conned spectrum. Indeed, the sharp transitions provide a much richer frequency content to ax(t), in comparison to the smooth x(t). As a matter of fact, as illustrated in Figure 9, for two time-domain CDMA2000 amplitude envelope signals (9a, forward and 9c, reverse), in spite of having the same average power, the sharp transitions are much more evident in the forward signal case as it has a higher peak-tominimum variation than the one obtained for the reverse case, as it is possible to see in Figures 9b and 9d. Fortunately, the amplitude signal spectrum depicted in Figure 10a shows that, not only there is a large amount of power concentrated at dc (consequence of the non-null average of ax(t), as |x(t)| is assumed to be always positive), but also the power spectral density of ax(t) falls very rapidly with the increasing frequency. Therefore, it is important to identify the minimum speed requirements that the SM circuitry must obey, not only in terms of linear bandwidth but also in terms of nonlinear slew rate (SR), so that the supply modulated RF PA, or the polar transmitter, can meet the NMSE and ACPR specs. Also, in Figure 10b, it is possible to see the spectra of two signals with constant amplitude and with the CDMA2000 forward and reverse correspondent phases, re-enforcing the already mentioned problem (see Section II) related with the modulated RF carrier large bandwidth, much higher than the one of the original signal, modulated in amplitude and phase. A. Linear Bandwidth Requirements To illustrate the bandwidth requirements, we conceived an ideal linear supply modulator whose bandwidth limitation

Figure 8 vDD(t)-to-ay(t) characteristic of a pHEMT based Class-E PA.

International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce

246

Cabral et al.

Figure 9 Example of time-domain amplitude envelope of two CDMA2000 signals: (a) forward and (c) reverse and correspondent pdf functions, (b) and (d), respectively. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

is imposed by a low-pass reconstruction lter as the ones normally used in digital audio ampliers, switching mode power supplies and in some of previously reported supply modulators [24]. This lter is responsible for eliminating all the modulation replicas, centered at the harmonics of xs, the Class-S modulator clock sampling frequency (the LPF reconstruction lter of Fig. 7), [21]. We started by extracting the amplitude envelope of a CDMA2000FW signal with an ideal envelope detector, ax(t) abs[x(t)], and then forced it to pass through the lter to obtain ay(t) H[ax(t)], where H[.] stands for the

linear dynamic lter operator. After this, we considered that the RF PA was ideal, so that the resulting complex envelope signal was rebuilt by: y(t) ay(t).ej/xt . Finally, the signal quality was quantied via the co-channel and adjacent-channel distortion with the so-called raw-EVM, or NMSE, and ACPR metrics dened in eqs. (1) and (2). The obtained results are shown in Figure 11a and 11b for NMSE and ACPR, respectively, as a function of the SM bandwidth, normalized to the original complex envelope bandwidth, Bw 1.25 MHz, for a 2nd order LC Butterworth lter.

Figure 10 Example of two CDMA2000 signals complex envelope, amplitude and phase spectra, (forward and reverse). [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

International Journal of RF and Microwave Computer-Aided Engineering/Vol. 20, No. 2, March 2010

Wireless Transmitter Capabilities

247

Figure 11 NMSE (a) and ACPR (b) as a function of supply modulator bandwidth, normalized to the complex envelope bandwidth, Bw 1.25 MHz. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

As seen, a lter bandwidth of about 2Bw is enough to guarantee the required values of 45 dBc of ACPR and 35 dB of NMSE. Additional tests conducted with other lter types (namely Chebyshev with a ripple of 1dB) and orders (3, 4, and 5) showed that, to achieve the mentioned gures of ACPR and NMSE, bandwidths ranging from 2Bw to 5Bw are enough. B. Nonlinear Slew Rate Requirements Now, let us direct our attention to another speed requirement, often overlooked: the slew rate. As is known from basic electronics [25], the slew rate of a voltage amplier evaluates the maximum derivative with respect to time with which its output can vary. Above that speed limit, the output will not follow the amplied input, but will increase (or decrease) at a constant rate of SR (V/s), converting any waveform into a positive or negative slope ramp. Slew rate constitutes, therefore, a fundamental nonlinear process that, contrary to the linear bandwidth limitation, only manifests itself when very high excitation amplitudes and frequencies are simultaneously involved. In high power devices, as the ones necessary for driving the Si LDMOS or GaN HEMT based drain-modulated PAs that are being designed for wireless base-stations, slew rates of tens or hundreds of V/ls can be the limiting factor in selecting appropriate SM ampliers. For studying the effect of SM slew rate, we again extracted the input amplitude envelope, ax(t), and processed it by the following algorithm accounting for the slew rate.  8   < Av ax n ay n 1 SR Ts ) ay n Av ax n Av ax n ay n 1 > SR Ts ) ay n ay n 1 : signAv ax n ay n 1 SR Ts (9) where, Av is the amplifier voltage gain, ax(n) and ay(n) are the input and output amplitudes, respectively, sampled at the time instant n and with a sampling period

of Ts. Then, we again used the resulting output amplitude to build the output complex envelope as y(t) ay(t).ej/xt . Figure 12 depicts a sample of the input, ax(t), and output, ay(t), time-domain waveforms when the SM is intended to provide a supply voltage of about ay_Max 30 V peak and is subject to a slew rate of 45 V/ls. Note that it is in the region of lowest values of ay(t) where the limited SR manifests its impact. In fact, it is there that the input amplitude has its highest time derivative, something on the order of ayMax pBw. So, we may anticipate that a signal format like EDGE, that is known for presenting a hole in the center of the constellation diagram, and thus having smoother transitions of ay(t) in the low voltage end, will be more immune to limited SM slew rate, [26, 27]. Again, to illustrate this speed-limit requirement, we tested several SR values and calculated the corresponding NMSE and ACPR. Figures 13a and 13b depict these results.

Figure 12 Input and output time-domain samples of a supply modulated voltage of about 30 V peak and subject to a slew rate of 45 V/ls. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce

248

Cabral et al.

Figure 13 NMSE (a) and ACPR (b) as a function of supply modulator slew rate for a CDMA2000 envelope amplitude signal with 30 V peak voltage. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

It is clear that, for the adopted CDMA2000FW signal format, a slew rate of about 45 V/ls, i.e., of nearly one half the maximum output signal slope, is needed to guarantee the desired 45 dBc of ACPR and 35 dB of NMSE. It is obvious that there are two ways of overcoming these very stringent gures. The rst one will be to construct the ideal hardware setup capable of fullling all the requirements. Unfortunately, with the demand for higher communication rates, we have been witnessing a growth in new and more complex signals with higher bandwidths and with drastic amplitude variations, putting the hardware developers in big stress to accomplish something that is changing and evolving every day. This being the case, the other possibility is to somehow try to alleviate the requirements employing digital predistortion (DPD) techniques. Although this is still subject to signicant research, there are already some practical implementations in both EER and ET systems [28, 29]. These rely on some form of reshaping the amplitude signal (this way relaxing the SM bandwidth requirements) then followed by some DPD linearization procedure (to fulll the linearity and efciency requirements). This can either be implemented applying a DPD algorithm to the input signal of an ET system [30] or, alternatively, separately applying independent predistorted signals to the AM and PM branches, as is shown in the architecture of Figure 14, [10, 31]. In

the DPD example described by the authors in [17], rst, an amplitude static linearization function, vg(vin), is designed to compensate for the carrier feed-through and to obviate the loss of PAE associated to the low end of envelope values. Then, the transmitter AM/AM nonlinearity is linearized using the vDD(vin) function. Finally, the residual AM/PM conversion is compensated via the /g(vin) function.

IV. ILLUSTRATIVE POLAR TRANSMITTER IMPLEMENTATION

Figure 14 Functional diagram of a possible digital predistorter to be used with the polar transmitter.

As a practical illustration of these polar transmitter linearity impairments, we assembled a complete polar transmitter in the RF laboratory, according to the polar system architecture discussed in Section II and presented in Figure 6. Figure 15 shows a schematic diagram of the implemented test setup. In this test bench, two Agilent E4438C vector signal generators are used to obtain the desired estimated vDD(t) (stored in the look-up-table of one of them), and amplitude and phase signals (in the other). The output, either in frequency or in time domain, was obtained using Agilents 89600 vector signal analyzer. In what PA switching mode classes of operation are concerned, any of inverse Class-D, Class-E, inverse ClassE, Class-F, or inverse Class-F is possible [5]. Among these, Class-E conguration, invented in 1975 by Sokal and Sokal [32], has become one of the most popular and effective ones, in the microwave range, because of its high efciency [5, 33, 34]. In terms of transistor technology, Gallium Nitride (GaN) HEMTs have already conquered their space because of their outstanding RF power characteristics (high breakdown voltage, high power per unit gate periphery, high fT, low input capacitance, manageable output capacitance and low ON-state resistance, RON) converting them into ideal candidates for switchedmode PAs. So, on the basis of Ref. 35, we designed a transmission line Class-E amplier using a GaN HEMT

International Journal of RF and Microwave Computer-Aided Engineering/Vol. 20, No. 2, March 2010

Wireless Transmitter Capabilities

249

Figure 15 Implemented test setup of the polar transmitter architecture. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

from Cree, (model CGH35015F), which has a maximum 73% PAE, for an output power of 15 W, under CW operation. Figure 16 presents a photograph of the PA. This PA imposes quite stringent requirements to the envelope amplier, namely simultaneous high drain current ($1 A) and drain voltage (30 V). Moreover, as in Section III, a wide bandwidth between 2Bw and 5Bw is required in the polar transmitter AM modulator. Besides, considering the maximum output voltage available from the employed arbitrary waveform generator (nearly 1V), the envelope amplier has to provide a high voltage gain. Considering all these requirements, two cascade noninverse ampliers were implemented for the envelope amplier. This circuit is based in a high gain-bandwidth product small-signal general purpose amplier followed by a high slew-rate (900 V/ls), high current (4 A), and high voltage (640 V) power operational amplier from Apex Microtechnology (PA119). The complete envelope amplier presents a 30 dB voltage gain, 10 MHz bandwidth and is capable of giving 30 V and 1.2 A. Unfortunately, being a continuous wave amplier, its efciency is not aligned with the other outstanding characteristics.

The test was conducted with a CDMA2000FW excitation signal. A special care was put on the compensation of the delay mismatch between the amplitude and phase branches, thus minimizing the distortion arising from this source. Using a reasonably high oversampling ratio, we managed to adjust the delay compensation digitally, on a sample by sample basis. With the setup described earlier, we obtained the illustrative polar transmitter input and output power spectral densities shown in Figure 17. The measured values for ACPR and NMSE were 42 dBc and 35 dB, respectively, which are in line with the previously presented studies. The average efciency (under the tested CDMA2000FW signal) of the RF PA was found to be 70%.
V. CONCLUSION

This overview article introduced the supply modulated PA as an ideal candidate to overcome the traditional

Figure 16 Photograph of the class E PA based on Cree CGH35015F GaN HEMT. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 17 Polar transmitter input and output power spectral densities. [Color gure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce

250

Cabral et al.
13. D. Milosevic, J. Van Der Tang, and A. Van Roermund, Intermodulation products in the EER technique, Int Symp Circuits Syst Dig vol. I, (2004) 637640. 14. J.C. Pedro, J.A. Garcia, and P.M. Cabral, Nonlinear distortion analysis of polar transmitters, IEEE Trans Microwave Theory Tech 55 (2007), 27572765. 15. X. Narisi Wang, V. Peng, D. Yousefzadeh, S. Maksimovic, and Z. Pajic, Linearity of X-band class-E power ampliers in EER operation, IEEE Trans Microwave Theory Tech 53 (2005), 10961102. 16. F. Wang, D. Kimball, and P. Asbeck, Envelope tracking power amplier with pre-distortion linearization for WLAN 802.11g, IEEE MTT-S Int Microwave Symp Dig 3 (2004), 15431546. 17. P.M. Cabral, J.C. Pedro, J.A. Garcia, and L. Cabria, A linearized polar transmitter for wireless applications, IEEE MTT-S Int Microwave Symp Dig (2008), 935938. 18. J. Choi, D. Kim, and D. Kang, A polar transmitter with CMOS programmable hysteretic-controlled hybrid switching supply modulator for multistandard applications, IEEE Trans Microwave Theory Tech 57 (2009), 16751686. 19. P. Warr, K. Morris, G. Watkins, T. Horseman, K. Takasuka, Y. Ueda, Y. Kobayashi, and S. Miya, A 60% PAE WCDMA handset transmitter amplier, IEEE Trans Microwave Theory Tech 57 (Part 1) (2009), 23682377. 20. D. Kimball, J. Jeong, C. Hsia, P. Draxler, S. Lanfranco, W. Nagy, K. Lithicum, L. Larson, and P. Asbeck, High-efciency envelope-tracking W-CDMA base-station amplier using GaN HFETs, IEEE Trans Microwave Theory Tech 54 (2006), 38483856. 21. I. Kim, Y. Woo, J. Kim, J. Moon, and J. Kim, High-efciency hybrid EER transmitter using optimized power amplier, IEEE Trans Microwave Theory Tech 56 (Part 2) (2008), 25822593. 22. M. Kazimierczuk, Collector amplitude modulation of the class E tuned power amplier, IEEE Trans CAS 31 (1984), 543549. 23. J.A. Garca, J.C. Pedro, P.M. Cabral, C. Fager, H.M. Nemati, A. Zhu, and P. Colantonio, Nonideality sources and implementation considerations in polar transmitters, Int J Microwave Wireless Technol 1 (2009), 109116. 24. F. Wang, D. Kimball, J. Popp, A. Yang, D. Lie, P. Asbeck, and L. Larson, Wideband envelope elimination and restoration power amplier with high efciency wideband envelope amplier for WLAN 802.11g applications, IEEE MTT-S Int Microwave Symp Dig (2005), 645648. 25. A.S. Sedra and K.C. Smith, Microelectronic circuits, New York, Oxford University Press, 2004. 26. D. Rudolph, Kahn EER technique with single-carrier digital modulations, IEEE Trans Microwave Theory Tech 51 (2003), 548552. 27. N. Lopez, X. Jiang, D. Maksimovic, and Z. Popovic, A highefciency linear polar transmitter for EDGE, IEEE Radio Wireless Symp Dig (2008), 199202. 28. I. Kim, J. Kim, J. Moon, and B. Kim, Optimized envelope shaping for hybrid EER transmitter of mobile WiMAX Optimized ET operation, IEEE Microwave Wireless Compon Lett 19 (2009), 335337. 29. J. Jeong, D. Kimball, M. Kwak, C. Hsia, and P. Draxler, Wideband envelope tracking power amplier with reduced bandwidth power supply waveform, IEEE MTT-S Int Microwave Symp Dig (2009), 13821384. 30. A. Zhu, P. Draxler, C. Hsia, T. Brazil, D. Kimball, and P. Asbeck, Digital predistortion for envelope-tracking power ampliers using decomposed piecewise volterra series, IEEE Trans Microwave Theory Tech 56 (2008), 22372247. 31. L. Cabria, P.M. Cabral, J.A. Garcia, and J.C. Pedro, Linearization of a polar transmitter for EDGE applications, in

efciency linearity compromise. Concentrating on the polar Tx architecture, because of its theoretical 100% efciency, it then discussed the major practical nonlinearity sources of this promising arrangement. Linearity impairments caused by amplitude and phase signal delay mismatch, nite AM modulator bandwidth, slew rate and nonlinear transfer function of the switched-mode PA were analyzed, when the polar transmitter is subject to a typical mobile communications signal. Finally, an illustrative experimental setup was assembled and tested. Its measured results sustained the system level study constraints theoretically advanced on the attainable linearity and efciency of polar transmitters.
ACKNOWLEDGMENT

The second and fourth authors would like to thank the support provided by MICINN and CICE-JA through the projects TEC2008-06684-C03-01 and P07-TIC-2649, respectively. The rst, third, and fth authors would like to thank to Portuguese Science Foundation, F.C.T., and Instituto de Telecomunicacoes for the nancial support provided under Projects: PTDC/EEA-TEL/65988/2006 Digital_PAs and SWIPA Ref. Nr.: P423, respectively.
REFERENCES 1. J.C. Pedro and N.B. Carvalho, Intermodulation distortion in microwave and wireless circuits, Norwood, MA, Artech House, 2003. 2. J.C. Pedro, Design strategies for linearity and efciency, short-course on device level linearization techniques, presented at European Microwave Week, Munich, October 2007. 3. J.C. Pedro, Theoretical results on the power efciency versus spectral efciency compromise of wireless transmitters, presented at Eighth WSEAS International Conference on Simulation Modeling and Optimization, SMO08 CDROM Digest, Santander, September 2008. 4. S. Cripps, RF power ampliers for wireless communications, 2nd ed., Norwood, MA, Artech House, 2006. 5. A. Grebennikov and N. Sokal, Switchmode RF power ampliers, New York, Newnes, 2007. 6. B. Putzeys, Digital audios nal frontier, IEEE Spectrum 40 (2003), 3441. 7. J. Candy and G. Temes, Oversampling delta-sigma data converters: Theory, design and simulation, IEEE Press, New York, 2001. 8. F. Raab, P. Asbeck, S. Cripps, P. Kenington, Z. Popovic, N. Pothecary, J. Sevic, and N. Sokal, Power ampliers and transmitters for RF and microwave, IEEE Trans Microwave Theory Tech 50 (2002), 814826. 9. P. Kenington, High linearity RF amplier design, Norwood, MA, Artech House, 2001. 10. R. Staszewski, J. Wallberg, S. Rezeq, C. Hung, O. Eliezer, S. Vemulapalli, C. Fernando, K. Maggio, R. Staszewski, N. Barton, M. Lee, P. Cruise, M. Entezari, K. Muhammad, and D. Leipold, All-digital PLL and transmitter for mobile phones, IEEE J Solid-State Circuits 40 (2005), 24692482. 11. L.R. Kahn, Single-sideband transmission by envelope elimination and restoration, Proc IRE 40 (1952), 803806. 12. F.H. Raab, Intermodulation distortion in Kahn-technique transmitters, IEEE Trans Microwave Theory Tech 44 (1996), 22732278.

International Journal of RF and Microwave Computer-Aided Engineering/Vol. 20, No. 2, March 2010

Wireless Transmitter Capabilities

Integrated non-linear microwave and millimetre-wave circuits conference Digest, Malaga, Spain, 2008, pp 115 118. 32. N. Sokal, and A. Sokal, Class EA new class of high-efciency tuned single-ended switching power ampliers, IEEE J Solid-State Circuits 10 (1975), 168176. 33. N. Kumar, C. Prakash, A. Grebennikov, and A. Mediano, High-efciency broadband parallel-circuit class E RF power

251

amplier with reactance-compensation technique, IEEE Trans Microwave Theory Tech 56 (2008), 604612. 34. A. Grebennikov and H. Jaeger, Class E with parallel circuitA new challenge for high-efciency RF and microwave power ampliers, IEEE MTT-S Int Microwave Symp Dig, (2008), 16271630. 35. T. Mader and Z. Popovic, The transmission-line high-efciency class-E amplier, IEEE Microwave Guided Wave Lett 5 (1995), 290292.

BIOGRAPHIES

Pedro M. Cabral was born in Oli veira de Azemeis, Portugal on October 1979. He earned Electrical Engineering and Ph.D. degrees from University of Aveiro, Portugal, in 2002 and 2006, respectively. Currently, he is a Pos-Doc Researcher at the Telecommunications Institute in Portugal and an Invited Assistant Professor at University of Aveiro. His main research interests include active device nonlinear modeling, design of microwave circuits, high efciency power ampliers, and wireless transmitter architectures. Dr. Pedro has been a reviewer for several journals: IEEE Transactions on Microwave Theory and Techniques, IEEE Transactions on CAD of Integrated Circuits and Systems, IEEE Transactions on Instrumentation & Measurement and IEEE Transactions on Circuits and Systems. In 2002, he received the best Electrical Engineering student prize at the University of Aveiro, and in 2004, he was nalist of the Student Paper Competition presented at the IEEE International Microwave Symposium. Lorena Cabria was born in Santander, Spain. She received Telecommunications Engineering and Ph.D. degrees from the University of Cantabria, Santander, in 2001 and 2007, respectively. Currently, she is working as a Research Associate in the RF and Microwave Group at the University of Cantabria. Her main research interests include microwave integrated circuits, intermodulation distortion control on RF and microwave applications, planar active antennas and retrodirective arrays, high efciency power ampliers, and emerging wireless transmitter architectures. Filipe Rodrigues was born in Anadia, Portugal, on August 1985. He received his M.Sc. degree in Electronics and Telecommunications Engineering, in 2009 from the University of Aveiro, where he is working toward his Ph.D. His main scientic interests include switch-mode audio power ampliers, dcdc converters,

high-efciency microwave/RF PAs, and new RF front end architectures. Jose A. Garcia was born in Havana, Cuba, on January 1966. He received his Telecommunication Engineering degree from the Instituto Superior Politecnico Jose A. Echeverra (ISPJAE) in 1988 and his Ph.D. from the University of Cantabria, Santander, Spain, in 2000. From 1988 to 1991, he was a Radio System Engineer at a High Frequency Communication Center, and he was a Instructor Professor in the Telecommunication Engineering Department, ISPJAE, in 1991. From 1999 to 2001, he was an RF/Microwave Design Engineer in Thaumat Global Technology Systems and TTI Norte. He has been with the University of Cantabria since 2002, rst as a Senior Research Scientist and currently as an Associate Professor. His main research interests include nonlinear characterization and modeling of active devices, intermodulation distortion control on RF, and microwave applications, as well as active antennas and arrays. Jose C. Pedro received his diploma, doctoral, and habilitation degrees in electrical engineering from University of Aveiro, Portugal, where he is a Professor currently. His scientic interests include active device modeling and the analysis and design of various nonlinear microwave circuits. He is the leading author of the book Intermodulation Distortion in Microwave and Wireless Circuits and has authored or co-authored more than 150 papers in international journals and symposia. He is an Associate Editor for the IEEE MTT Transactions and reviewer for the MTT-IMS and the EuMC. Prof. Pedro has served his university department as the Coordinator of the Scientic Council and as the Department Head. Dr. Pedro received the Marconi Young Scientist Award in 1993 and the 2000 Institution of Electrical Engineers (IEE) Measurement Prize. In 2007, he was elected as Fellow of the IEEE for his contributions to the nonlinear distortion analysis of microwave devices and circuits.

International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce