High-Efficiency GaN Doherty Power Amplifier for 100 ... - IEEE Xplore

IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 61, NO. 8, AUGUST 2013

2911

High-Efficiency GaN Doherty Power Amplifier for 100-MHz LTE-Advanced Application Based on Modified Load Modulation Network Jing Xia, Student Member, IEEE, Xiaowei Zhu, Member, IEEE, Lei Zhang, Jianfeng Zhai, and Yinjin Sun

Abstract—This paper presents a high-efficiency GaN Doherty power amplifier (PA) with 100-MHz instantaneous bandwidth for 3.5-GHz long-term-evolution (LTE)-advanced application. A modified load modulation network, employing an enlarged peaking amplifier to carrier amplifier power ratio and moderately increased load impedance of the carrier amplifier, is proposed for enhancing efficiency and achieving improved load modulation. To increase the power ratio and alleviate the influence of slight impedance mismatch, a proposed load impedance strategy and corresponding stepped-impedance matching network are adopted for high-efficiency and wideband operation. By tuning the carrier offset line, the inconsistency of efficiency, gain, and output power in the operation band can be alleviated. Measurement results show that the Doherty PA has a drain efficiency of approximately 40% with gain fluctuation less than 0.5 dB at 9-dB back-off power, and maximum efficiency of about 60% at saturation in the signal band of 3.4–3.5 GHz. By using the digital pre-distortion (DPD) technique, the Doherty PA achieves adjacent channel leakage ratio of about 48 dBc at an average output power of 40.4 dBm with efficiency of 42.5%, when driven by 100-MHz LTE-advanced signal. To the best of the authors’ knowledge, this is the first high-performance result of linearization using conventional DPD technique with 100-MHz bandwidth signals for the GaN Doherty PA at 3.5-GHz frequency band thus far. Index Terms—Digital linearization, Doherty power amplifiers (PAs), GaN, high-efficiency amplifiers, long-term-evolution (LTE) advanced, wideband microwave amplifiers.

I. INTRODUCTION

L

ONG-TERM evolution (LTE)-advanced is one of the most promising technologies for future mobile communication systems because of its high spectrum efficiency and transmission data rate [1]. In order to achieve high data rate and large system capacity, wide signal bandwidth (up to 100 MHz) will be utilized, resulting in high peak to average power ratio (PAPR) due to large envelope fluctuation of the signal in the time domain. To avoid clipping the peak signals and prevent

Manuscript received February 03, 2013; revised May 17, 2013; accepted May 21, 2013. Date of publication July 02, 2013; date of current version August 02, 2013. This work was supported in part by the National Key Technologies Research and Development Program of China under Grant 2010ZX03007-003-02. The authors are with the State Key Laboratory of Millimeter Waves, Southeast University, Nanjing 210096, China (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TMTT.2013.2269052

in-band distortion and out-of-band emission, the power amplifiers (PAs) are usually operated at the large back-off region, which leads to poor power efficiency. To effectively amplify 100-MHz bandwidth signals, wide instantaneous bandwidth PA, which can simultaneously meet efficiency requirements and linearity specifications, are needed in practical LTE-advanced systems. For efficiency enhancement, various techniques have been proposed to increase the average efficiency of the PA when excited with modulated signals. Among these, the Doherty PA is one of the most common solutions for high PAPR applications due to its significant efficiency enhancement at back-off operation and ease of implementation [2]–[14]. Recently, several asymmetrical Doherty PAs have been investigated to achieve higher efficiency at 6-dB or more back-off power [8]–[11]. To achieve better load modulation, the asymmetrical Doherty PA usually employs an uneven power divider to deliver more input power to the peaking amplifier, which results in gain degradation [10], [11]. In [3], a load modulation network (LMN) was proposed to overcome incomplete load modulation, but it may lead to efficiency degradation at same output power when compared with the conventional LMN. Referring to the linearity requirement, a wide instantaneous bandwidth PA and wideband linearization technique are required to amplify wideband modulated signals. That means the efficiency, output power, and gain of the PA should have high consistency over the whole signal bandwidth. In addition, the PA should also achieve smooth AM–AM response to show appropriate linearity. Recently, several design approaches have been used to increase the bandwidth of the Doherty PA [12]–[14], which makes it a good candidate for wide signal bandwidth applications. With the help of the digital pre-distortion (DPD) technique, many works have been done on high-linearity Doherty PAs [15]–[17]. However, due to the restriction of instantaneous bandwidth of the PAs, as well as the processing bandwidth of the DPD systems, most of the reported implementations were designed to work with narrowband signals, with instantaneous bandwidth of or lower than 20 MHz [15], [17]. Although a 2.14-GHz Doherty PA has been linearized to achieve good linearity for 100-MHz signals, a band-limited DPD rather than a conventional one was utilized due to its limited analysis bandwidth [18]. In addition, some wideband Doherty PAs do not have gradually compressed gain or large back-off region to work with the modulated signals having PAPR of about 9 dB [12], [14], which might lead to degradation of linearity and efficiency. Therefore, they can

0018-9480/$31.00 © 2013 IEEE

2912


hardly satisfy the requirements of future wireless communication systems. In [19], we have presented a GaN Doherty PA for 3.5-GHz LTE-advanced application. Due to the limitation of the analysis bandwidth of the DPD test bench, the linearization was only accomplished with a 50-MHz modulated signal. In this paper, a wide instantaneous bandwidth Doherty PA for 100-MHz LTE-advanced application and its linearization results by using a wideband DPD platform are reported. The carrier and peaking amplifiers are fabricated based on a 60-W Cree CGH35060 GaN HEMT and implemented with unequal saturation power. A modified LMN is proposed to enhance efficiency and achieve improved load modulation. With a proposed load impedance strategy and corresponding matching network, up to 100-MHz instantaneous bandwidth operation can be achieved. In addition, the consistency of efficiency, output power, and gain over the band of interest can be enhanced by tuning the carrier offset line. This paper is organized as follows. Section II analyses and derives the proposed LMN for efficiency enhancement. With this modified LMN, the design approaches for wide instantaneous bandwidth Doherty PA are discussed in Section III. The implementation and experimental results are presented in Section IV. Section V demonstrates linearity improvements of the proposed Doherty PA when excited with a 100-MHz LTE-advanced signal. Conclusions are presented in Section VI. II. ANALYSIS OF THE LMN FOR EFFICIENCY ENHANCEMENT To efficiently amplify modulated signals with high PAPR, the power efficiency should be enhanced in the back-off region. This section analyzes the LMN in the Doherty PA with unequal carrier and peaking amplifiers and its influence on the overall efficiency. The proposed design approach for high efficiency will be introduced accordingly. A. Analysis of the LMN in Doherty PA With Unequal Carrier and Peaking Amplifiers

Fig. 1. Operation diagram of the Doherty PA.

(2)

The operating principle of the LMN can be analyzed according to the different input powers using (1) and (2). When the input power is low, the peaking amplifier in the off-state is open circuited seen from the combining point. Thus, the effective load impedances in the low-power region can be given by (3) (4) When the input power is high, the output powers of the carrier and peaking amplifiers reach saturation and can be expressed in terms of maximum fundamental currents and load impedances of the two amplifiers as follows: (5) (6)

The classic Doherty PA is usually realized by employing class-AB and class-C operating conditions for the carrier and peaking amplifiers, respectively. The peaking amplifier is turned on at suitable input power level by tuning the bias point appropriately. In general, the output current of the peaking amplifier cannot attain the same level as the carrier amplifier due to lower biasing, when two identical devices are adopted. As a result, the optimum load modulation cannot be fully achieved, leading to output power and efficiency degradation. To solve this problem, the operating principle of the Doherty PA with unequal carrier and peaking amplifiers will be analyzed as follows. Fig. 1 shows the operation diagram of the Doherty PA. If each transistor is considered to be an ideal current source, according to the analysis of an ideal Doherty PA, the effective load impedances seen by each amplifiers can be expressed in terms of and , as well as the fundamental currents of the two amplifiers as follows [3]:

For further use, two new parameters and are defined as the power ratio and current ratio, respectively. According to (2), the modulated load impedance seen by the peaking amplifier at the high-power region becomes

(1)

(9)

(7) By means of (5)–(7), taking the power ratio and current ratio into consideration, the modulated load impedance of the carrier amplifier can be derived as follows: (8) Substituting (8) into (1), the impedance seen at the combining point is given by

XIA et al.: HIGH-EFFICIENCY GaN DOHERTY PA FOR 100-MHz LTE-ADVANCED APPLICATION BASED ON MODIFIED LMN

According to the impedance transformation of the , the characteristic impedance can be calculated as

2913

line

(10) is 50 . Therefore, (7)–(10) illustrate In the general case, that the modulated load impedances and , the impedance seen at the combining point , and the characteristic impedance are functions of current ratio, power ratio, and characteristic impedance . Once these design parameters are determined, a configuration of the LMN, which can achieve complete load modulation even if the two amplifiers have unequal saturation powers, can be obtained. B. Approach to Enhance Efficiency in the Back-Off Region The influence of the effective load impedance seen by the carrier amplifier on the efficiency of the Doherty PA has been discussed in [4]. It can be found that the load impedance of the carrier amplifier larger than 2 can achieve higher efficiency in the back-off region, while maintaining output matching network of the carrier matched to . For the general case, the saturation power of the peaking amplifier is usually lower than that of the carrier amplifier due to lower biasing. Thus, the power ratio is always less than 1. For instance, by assuming the power ratio and the corresponding current ratio when both amplifiers have equal load impedances, the normalized (with reference to ) load impedances , , and can be calculated according to (3), (7), and (8), as illustrated in Fig. 2(a). For convenience, the influences of improper load modulation on the carrier and peaking amplifiers are assumed to be identical so the power ratio is assumed to be invariable to calculate these impedances. To ensure modulated load impedances of both amplifiers at the high-power region equal to , can be determined to be 0.84 . Meanwhile, the effective load impedance drops to as low as 1.7 , resulting in efficiency degradation compared with conventional Doherty PA at same output power, according to the analysis of the ideal Doherty. This LMN is just the one proposed in [3], and similar efficiency degradation can also be observed. To enhance efficiency, higher effective load impedance should be achieved. Actually, this requirement can be met by using a modified LMN. Substituting (9) into (3), the following relationship can be finally inferred in consideration of (8): (11)

Fig. 2. Normalized load impedances at low- and high-power region under dif. (a) and . (b) and . (c) ferent and .

From (11), it is can be seen that increased power ratio can enlarge the effective load impedance seen by the carrier amplifier in the low-power region, and hence, enhance efficiency in the back-off region, as an asymmetrical Doherty PA does [8]. For example, by enlarging the power ratio to 0.84 and setting to 0.92 , the effective load impedance can be increased to 1.84 , while maintaining modulated load impedances of both amplifiers at the high-power region equal to , as depicted in Fig. 2(b).

Moreover, higher modulated load impedance can also lead to further improvement. Taking into account (8), both and increased lead to higher modulated load impedance , when the power ratio is determined. As shown in Fig. 2(b), when increases to and the corresponding modulated load impedance and is equal to 1.1 , the effective load impedance becomes 2 , which is just the case using a conventional LMN. However,

2914


TABLE I COMPARISONS OF THE LMNs

it should also be mentioned that the modulated load impedances of both amplifiers at saturation have deviated from , resulting in impedance mismatch. The impedance mismatch of amplifiers, especially the peaking amplifier, will lead to output power degradation and incomplete load modulation, and finally influence the overall efficiency and output power in practice. To alleviate the performance deterioration induced by impedance mismatch, the modulated load impedance can be kept to by enlarging to 1 and letting to , the effective load impedance can be increased to 2.2 , as depicted in Fig. 2(c). The power ratio is also assumed to be 0.84 for convenience. By adopting this modified LMN, the modulated load impedance is from 2.2 to 1.2 for the carrier amplifier, and from to for the peaking one. The is larger than 2 , resulting in expected efficiency enhancement. Table I compares the load impedances, characteristic impedances, efficiency, and load modulation of the LMNs for the above cases. The expected efficiency at back-off is compared with the ideal Doherty. It shows that the proposed design exhibits larger effective load impedance , which means higher efficiency can be expected in the back-off region. Although the modulated load impedance differs from the ideal case and may lead to impedance mismatch, its influence can be moderately alleviated by selecting appropriate impedance to present to the device package plane and suitable matching network, as discussed in Section III. III. DESIGN APPROACH OF WIDE INSTANTANEOUS BANDWIDTH DOHERTY PA The design strategy for efficiency enhancement, mentioned in Section II, becomes insufficient when the Doherty PA is excited with modulated signals having high PAPR, as well as wide instantaneous bandwidth. Combined with the proposed LMN, a design approach is devised for the Doherty PA to achieve approximately consistent efficiency, gain, and output power on the signal band for 100-MHz LTE-advanced application, while maintaining enhanced efficiency in the back-off region.

Fig. 3. Simulated output power and efficiency contours normalized to 10 for 3.4–3.5 GHz. (Case I represents 47.8-dBm output power and 68% efficiency and Case II means 47.4-dBm output power and 73% efficiency.)

A. Proposed Load Impedance Strategy and Corresponding Matching Network Design To obtain the optimal load impedances to present to the device package plane of the carrier and peaking amplifiers, the load–pull simulations based on the Cree CGH35060F GaN HEMT are adopted. The model was biased at conventional class-AB mode in Agilent Technologies’ Advanced Design System (ADS) simulations. To realize consistent output power, gain, and efficiency, the influence of different fundamental load impedances was analyzed across the band of 3.4–3.5 GHz by producing efficiency and power contours. The results shown in Fig. 3 were generated by sweeping the resistance and reactance of fundamental load impedances while setting second and third harmonic terminations to nearly short circuit. For the sake of clarity, the normalized impedance in Fig. 3 is reduced to 10 . In general, the design criterion of optimal load impedances of the carrier and peaking amplifiers gives consideration to both output power and efficiency, as shown in Case I. The inside region of the contour represents the area on the Smith chart where the desired design criterion


2915

Fig. 5. Schematic diagram of the proposed LMN.

Fig. 4. Simplified matching networks for the carrier and peaking amplifiers and the matching traces on the Smith chart.

of greater than 47.8-dBm output power and 68% efficiency are satisfied. According to (11) and analysis in Section II-B, an enlarged power ratio between the peaking and carrier amplifiers leads to efficiency enhancement. Meanwhile, the Doherty PA is usually excited with modulated signals and should be optimized for high efficiency in the back-off region associated with the PAPR of the modulated signal, and the efficiency at the peak power region is not as important [4]. Thus, a new design criterion of optimal load impedances is proposed. The carrier amplifier can be implemented with higher efficiency and corresponding slightly lower output power than the peaking amplifier. This leads to larger power ratio, and hence, improved load modulation. Moreover, with the load impedances of the carrier amplifier located in the higher efficiency region, further enhanced efficiency can be achieved. Thus, this new design criterion for the carrier amplifier cannot only enhance the back-off efficiency, but also improve the load modulation, with the cost of slight degradation of the output power. After making a tradeoff between the efficiency, load modulation, and output power, 73% efficiency and 47.4-dBm output power were chosen as the criteria for the carrier amplifier, as illustrated in Case II of Fig. 3. To achieve constant performance over the whole band, the optimum load impedances should be chosen close to the center of these contours in Fig. 3. This is suitable for the peaking amplifier, but not always for the carrier amplifier when using the proposed LMN. As illustrated in Section II-B, the modulated load impedance will differ from the ideal case, which leads to impedance mismatch. To alleviate output power and efficiency degradation, the optimal load impedances of the carrier amplifier should be chosen carefully to ensure that the load impedance transformed to the package plane can still be accepted when the modulated load impedance slightly differs from (50 ). Meanwhile, an appropriate matching network should be used for this purpose. The stepped-impedance matching network is one of the most widely used techniques for impedance matching in wideband PA design [22]. By selecting the desired impedances inside the contours with a compact impedance trajectory, a stepped-impedance matching network with appropriate low

Fig. 6. Simulated load modulation results for the conventional and proposed carrier amplifier with different load impedances at frequency of 3.45 GHz.

Fig. 7. Simulated gain and efficiency variations with different electrical lengths of the carrier offset line.

cannot only provide wideband matching over the band of interest, but also alleviate the influence of slight impedance mismatch. For instance, assuming the modulated impedance in the proposed LMN to be 60 , Fig. 4 shows the matching network schematics of the carrier and peaking amplifiers and compares the matching traces on the Smith chart. The matching network I and II are designed for the carrier and peaking amplifiers, respectively. It is obvious that the matching network I with the modulated load impedances of 50 and 60 can achieve similar performances because the

2916


Fig. 8. Complete schematic of the Doherty PA. (Electrical lengths refer to center frequency of 3.45 GHz.)

destinations of the two matching traces are locating on the same efficiency and power contours even though the starting points are different. Regarding the peaking amplifier, using the matching network II, the optimum impedances are located on the center of the contours considering both efficiency and power. This results in high output power and enlarged power ratio between the two amplifiers. After the output matching networks are specified, the input matching networks can be designed by using a similar topology to cover the required bandwidth. B. Modified LMN for Higher Efficiency and Wideband Operation In Section III-A, the optimum load impedances of the carrier and peaking amplifiers have been decided. Thus, the power ratio between the two amplifiers can be evaluated by simulation for the proposed Doherty PA. When the carrier and peaking amplifiers were biased at class AB and a predetermined class-C mode, which is associated with 9-dB back-off, as discussed in Section III-C, the power ratio was calculated to be 0.84. For comparison, the power ratio was about 0.7 when both carrier and peaking amplifiers employing matching network II, which leads to improper load modulation using the conventional LMN, as discussed in Section II. Consequently, for the proposed Doherty, the characteristic impedances and were evaluated to be 50 and 33.7 , when assuming the current ratio . Fig. 5 illustrates the schematic diagram of the proposed LMN. It can be noted that the effective load impedance in this case becomes 110 rather than 100 . Moreover, as mentioned in the literature [3], [6], two additional offset lines should be inserted behind the carrier and peaking amplifiers, respectively. With the phase adjustment of the peaking offset line, the peaking amplifier can achieve a quasi-open circuit at the junction point to prevent power leakage in the low-power region. Meanwhile, the carrier amplifier can achieve the expected load impedance through the quarter-wave impedance transformer with the appropriate carrier offset line. Fig. 6 illustrates simulated load modulation results for the carrier amplifiers employing load impedances of 100 and 110 at a

frequency of 3.45 GHz. For comparison, the proposed Doherty PA employs the proposed LMN and matching network I and II for the carrier and peaking amplifiers, while the conventional Doherty PA uses the conventional LMN and matching network II for both carrier and peaking amplifiers. For the three cases in the results, the electrical lengths of the carrier offset lines are 175 , 175 , and 135 , respectively. It is shown that the three cases achieve efficiency of 41%, 40%, and 38% at an output power of 40 dBm, respectively. As expected, the proposed carrier amplifier with a 110- load delivers higher efficiency than the one with a 100- load, and the conventional carrier amplifier with a 100- load presents lowest efficiency among the three cases. To achieve wide instantaneous band operation, the influence of the carrier offset line on the efficiency and gain performances in the whole band was also analyzed. Fig. 7 shows simulated gain and efficiency variations of the proposed carrier amplifier with a 110- load with different electrical lengths of the carrier offset line. An appropriate electrical length of the carrier offset line (175 ) can achieve gain and efficiency fluctuation less than 0.3 dB and 3%, respectively, while increasing or decreasing the electrical length may cause inconsistent gain and efficiency in the band of interest. C. Schematic and Simulations The circuit schematic of the proposed Doherty PA has been implemented and simulated in ADS. Fig. 8 shows the schematic represented by equivalent ideal lines, whose electrical lengths refer to a center frequency of 3.45 GHz. The carrier and peaking amplifiers are implemented by using two equal-sized devices achieving unequal efficiency and output power due to different output matching networks I and II. Using the matching networks illustrated in Fig. 4, the output matching networks of the carrier and peaking amplifiers can be designed using Smith chart utility in ADS to show appropriate load impedances associated with the proposed design criterion. To achieve higher efficiency in the back-off region and improved load modulation, the modified LMN is adopted. After determining the source impedances by using source–pull simulation, the input matching networks


Fig. 9. Simulated load impedance and efficiency of the conventional and proposed Doherty PAs as a function of output power.

can be designed by using a similar approach to cover the required bandwidth. Moreover, the input matching network of the peaking amplifier was also optimized for predetermined class C mode to achieve consistent gain and output power performances. In addition, appropriate tuning of the carrier offset line was made for performance enhancement due to the influence of the quasi-open circuit of the peaking amplifier at the junction point. For further comparison, a conventional Doherty PA with both amplifiers employing an output matching network II and a conventional LMN was also simulated. As mentioned in Section II-A, higher biasing of the peaking amplifier leads to enlarged output power and improved load modulation. Considering the purpose for high PAPR applications, it might be an appropriate tradeoff between the efficiency and load modulation to turn the peaking amplifier on at about the 9-dB back-off region when the same devices are used. In addition, higher biasing can also enhance the consistency of the saturation power of the peaking amplifier in class-C mode. Fig. 9 shows simulated load impedance and efficiency of the conventional and proposed Doherty PAs as a function of output power with V and V for the carrier amplifiers. For appropriate comparison, the peaking amplifiers in both Doherty PAs turn on at about 9-dB back-off power. The onset of the soft turn-on region was used to determine the back-off power region. Therefore, V and V were chosen for the peaking amplifiers in both Doherty PAs. It is shown that the fundamental load impedances of the carrier and peaking amplifiers in the proposed Doherty PA are modulated from 110 to 60 and from high impedance to 50 , respectively. For the conventional one, both modulated load impedances are larger than 50 in the high-power region. These results confirm that expected load modulation can be achieved using the proposed LMN while improper load modulation is observed when using the conventional LMN. The results also verify the analysis in Section II. Moreover, as can be noted, with similar saturation power, the proposed Doherty PA can achieve higher efficiency than the conventional one in the back-off power region, leading to overall efficiency enhancement for modulated signals with high PAPR.

2917

Fig. 10. Simulated efficiency and gain of the proposed Doherty PA versus frequency at saturation and 6-dB back-off power.

Fig. 11. Photograph of the implemented Doherty PA.

In Fig. 10, the simulated efficiency and gain of the proposed Doherty PA at 6-dB back-off power and in saturation versus frequency are depicted. It is shown that the variation of efficiency is less than 5% and the gain fluctuation is less than 0.5 dB. Appropriately consistent performance of efficiency, gain, and output power can be observed, which ensured wide instantaneous bandwidth operation. IV. REALIZATION AND EXPERIMENTAL CHARACTERIZATION In order to experimentally validate the proposed design strategy, two Doherty PAs corresponding to the proposed and conventional designs were implemented using Cree CGH35060F GaN HEMTs and on a Taconic RF35 substrate with and a thickness of 30 mil at the frequency band of 3.4–3.5-GHz, as shown in Fig. 11. The carrier amplifiers were biased in the class-AB condition, with 0.25-A quiescent current (about 8.5% of the maximum current), while the peaking amplifiers were biased in class-C condition to turn on at about 9-dB back-off power. The input and output matching networks for both amplifiers were optimized separately to conform to the proposed and conventional designs. Post-tuning of the offset lines, especially the carrier offset line, is recommended for achieving high efficiency and maintaining good consistency of gain and efficiency.

2918


Fig. 12. Measured efficiencies of the conventional and proposed Doherty PAs at frequency of 3.45 GHz. The PDF of a 100-MHz LTE-advanced modulated signal with a PAPR of about 9 dB was also depicted.

Fig. 15. Measured efficiencies and ACLRs of the conventional and proposed Doherty PAs for a 100-MHz LTE-advanced signal.

Fig. 16. Schematic of the linearization experimental test bench.

Fig. 13. Measured efficiency and gain of the proposed Doherty PA versus output power at the frequency of 3.4, 3.45, and 3.5 GHz.

Fig. 17. Measured efficiency and ACLR characteristic before and after linearization for the conventional and proposed Doherty PA versus average output powers at carrier frequency of 3.45 GHz.

Fig. 14. Measured efficiency and gain of the proposed Doherty PA versus frequency at saturation and at 6- and 9-dB back-off powers, respectively.

Fig. 12 shows the measured efficiency characteristics of the conventional and proposed Doherty PAs as a function of output power under continuous wave (CW) signal measurements at the frequency of 3.45 GHz, and a probability density function (PDF) of a 100-MHz LTE-advanced modulated signal with a

PAPR of about 9 dB was also depicted. For appropriate comparison, the peaking amplifiers in both Doherty PAs turn on at about 9-dB back-off power. It is shown that about 3% efficiency improvement can be observed for the proposed design. Although the excepted Doherty region may not be visible, the influence on the average efficiency might be slight when the PA is excited with modulated signals at about 9-dB back-off power in terms of statistical analysis according to the PDF. In addition, similar saturation powers (higher than 49.5 dBm) can be achieved by both Doherty PAs. The measured efficiency and gain of the proposed Doherty PA versus output power at the frequency of 3.4, 3.45, and 3.5 GHz


2919

TABLE II COMPARISON WITH 3.5-GHz DOHERTY PAs AND SOME RESULTS OF 2.14- AND 2.5-GHz DOHERTY PAs

are shown in Fig. 13. It can be noticed that the efficiency, gain, and output power maintain appropriate consistency at different frequencies across a large back-off range. In addition, due to the effect of the soft turn-on region, the knee points at about 9-dB back-off power cannot be observed obviously in the results and slight performance degradation might be inferred. Fig. 14 shows the measured efficiency and gain of the proposed Doherty PA versus frequency at different back-off powers. The maximum efficiency is between 59%–63% at saturation, and is between 50%–46% and 40%–38% at 6- and 9-dB back-off power, respectively. Regarding the gain, its fluctuation is less than 0.5 dB, which implies appropriate consistency. To evaluate the performance of the proposed Doherty PA for LTE-advanced application, a 100-MHz LTE-advanced signal with a contiguous five-carrier configuration and PAPR of about 8.5 dB at a carrier frequency of 3.45 GHz was used to assess the efficiency and linearity performances, as depicted in Fig. 15. The results are compared with that of the conventional Doherty PA. The proposed Doherty PA can achieve higher (2%–3%) efficiency than the conventional one, while maintaining similar linearity performances with it.

TABLE III COMPARISONS OF THE PERFORMANCES

V. DPD LINEARIZATION RESULTS For validation of the linearity improvement of the proposed Doherty PA, the linearization was implemented using a wideband DPD platform with maximum system bandwidth of 400 MHz for LTE-advanced applications [20]. The measurement results are compared with the conventional Doherty PA. The 100-MHz five-carrier LTE-advanced signal was used to excite the Doherty PAs. The block diagram of the experimental test beach is shown in Fig. 16. A baseband in-phase/quadrature (I/Q) complex signal is generated in MATLAB and is then downloaded into the baseband board. After converted into the analog domain, the signal is sent into the RF board to be modulated and up-converted to the desired RF frequency, and then fed into the Doherty PA after being pre-amplified by a wideband driver amplifier. Meanwhile, in order to extract the model parameters of the PA, the output signal should be coupled and transferred to baseband by a down-converter with wide bandwidth and high

Fig. 18. Measured spectrum before and after linearization for 100-MHz fivecarrier LTE-advanced signal.

linearity, and then converted to baseband signals. The DPD model employed was the piecewise second-order dynamic deviation reduction (DDR) model described in [21]. The magnitude threshold was set as {0.4 0.7} for the normalized data, the corresponding nonlinearity order was selected as {7, 7, 7}, and the memory length was set to {5, 5, 5}. The comparison of the linearization experimental results for the conventional and proposed Doherty PAs at different output powers are shown in Fig. 17. Although similar or slightly better linearity can be achieved by the conventional Doherty PA, its average efficiency is lower than the proposed one. It is shown

2920


that the proposed Doherty PA can achieve an efficiency of 42.5% and an ACLR of 48 dBc at about 9-dB back-off power (40.4 dBm) when using DPD linearization. Comparison with similar 3.5-GHz Doherty PAs is outlined in Table II, including some results of the 2.14- and 2.5-GHz Doherty PA for more comparison in modulated signal applications. For CW measurements, the proposed Doherty PA outperforms most 3.5-GHz Doherty PAs in terms of output power and 9-dB back-off efficiency. Although the efficiency in [25] is higher than this study, a harmonic control technique was employed for efficiency enhancement. Considering the operating frequency, this study might also be comparable with those Doherty PAs operating at lower frequency. For modulated signal measurements, the proposed Doherty PA was linearized over a 100-MHz bandwidth, while the widest signal bandwidth in other research is only 28 MHz. Good efficiency performance was also achieved at about 9-dB back-off power, especially compared with 3.5-GHz Doherty PAs. With DPD, the proposed Doherty PA can achieve good linearity when excited with a 100-MHz modulated signal. In addition, a well-linearized an adjacent channel leakage power ratio (ACLR) of about 50 dBc can be obtained by the proposed Doherty PA in the back-off region. For example, the measured linearization performances at an average output power of about 40 dBm are summarized in Table III. About 18-dB ACLR reduction has been achieved for the proposed Doherty PA, which results in an ACLR of 49.2 dBc ( 20 MHz) and 49.6 dBc ( 20 MHz). Fig. 18 shows the measured frequency-domain spectrum with and without linearization of the conventional and proposed Doherty PA. It is shown that the distortion caused by the PA nonlinearities and memory effects can be effectively removed. VI. CONCLUSION In this paper, a high-efficiency GaN Doherty PA with 100-MHz instantaneous bandwidth for 3.5-GHz LTE-advanced application and its related linearization results have been reported. The influence of the LMN for efficiency enhancement of the Doherty PA was analyzed, thus a modified LMN was proposed to enhance efficiency in the back-off region and achieve improved load modulation. With a new load impedance strategy and corresponding wideband matching network, wide instantaneous bandwidth operation can be achieved. Experimental results show that the proposed Doherty PA has good efficiency and high-performance linearity when using a DPD experimental test bench. ACKNOWLEDGMENT The authors wish to acknowledge Dr. A. Zhu, C. Yu, and the RF and Microwave Research Group, University College Dublin, Dublin, Ireland, for the cooperation on PA modeling. The authors also want to thank the ZTE Corporation for the technical supports on the LTE-advanced signal. REFERENCES [1] A. Ghosh, R. Ratasuk, B. Mondal, N. Mangalvedhe, and T. Thomas, “LTE-advanced: Next-generation wireless broadband technology,” IEEE Wireless Commun., vol. 17, pp. 10–22, Jun. 2010.

[2] S. Kawai, Y. Takayama, R. Ishikawa, and K. Honjo, “A high-efficiency low-distortion GaN HEMT Doherty power amplifier with a series-connected load,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 2, pp. 352–360, Feb. 2012. [3] S. Chen and Q. Xue, “Optimized load modulation network for Doherty power amplifier performance enhancement,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 11, pp. 3474–3481, Nov. 2012. [4] J. Moon, J. Kim, J. Kim, I. Kim, and B. Kim, “Efficiency enhancement of Doherty amplifier through mitigation of the knee voltage effect,” IEEE Trans. Microw. Theory Techn., vol. 59, no. 1, pp. 143–152, Jan. 2011. [5] P. Colantonio, F. Ciannini, R. Giofre, and L. Piazzon, “Increasing Doherty amplifier average efficiency exploiting device knee voltage behavior,” IEEE Trans. Microw. Theory Techn., vol. 59, no. 9, pp. 2295–2305, Sep. 2011. [6] D. Kang, J. Choi, D. Kim, and B. Kim, “Design of Doherty power amplifiers for handset applications,” IEEE Trans. Microw. Theory Techn., vol. 58, no. 8, pp. 2134–2142, Aug. 2010. [7] J. Kim, J. Moon, Y. Y. Woo, S. Hong, I. Kim, J. Kim, and B. Kim, “Analysis of a fully matched saturated Doherty amplifier with excellent efficiency,” IEEE Trans. Microw. Theory Techn., vol. 56, no. 2, pp. 328–338, Feb. 2008. [8] K. Jangheon et al., “Power efficiency and linearity enhancement using optimized asymmetrical Doherty power amplifiers,” IEEE Trans. Microw. Theory Techn., vol. 59, no. 2, pp. 425–434, Feb. 2011. [9] T. Kitahara, T. Yamamoto, and S. Hiura, “Doherty power amplifier with asymmetrical drain voltages for enhanced efficiency at 8 dB backed-off output power,” in IEEE MTT-S Int. Microw. Symp. Dig., Baltimore, MD, USA, Jun. 2011, pp. 1–1. [10] M. Nick and A. Mortazawi, “Adaptive input-power distribution in Doherty power amplifiers for linearity and efficiency enhancement,” IEEE Trans. Microw. Theory Techn., vol. 58, no. 11, pp. 2764–2771, Nov. 2010. [11] J. Kim, J. Cha, I. Kim, and B. Kim, “Optimum operation of asymmetrical-cells-based linear Doherty power amplifiers-uneven power drive and power matching,” IEEE Trans. Microw. Theory Techn., vol. 53, no. 5, pp. 1802–1809, May 2005. [12] J. M. Rubio, J. Fang, V. Camarchia, R. Quaglia, M. Pirola, and G. Ghione, “3–3.6-GHz wideband GaN Doherty power amplifier exploiting output compensation stages,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 8, pp. 2543–2548, Aug. 2012. [13] G. Sun and R. H. Jansen, “Broadband Doherty power amplifier via real frequency technique,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 1, pp. 99–111, Jan. 2012. [14] D. Y. Wu and S. Boumaiza, “A modified Doherty configuration for broadband amplification using symmetrical devices,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 10, pp. 3201–3213, Oct. 2012. [15] S. Jung, O. Hammi, and F. M. Ghannouchi, “Design optimization and DPD linearization of GaN-based unsymmetrical Doherty power amplifiers for 3G multicarrier applications,” IEEE Trans. Microw. Theory Techn., vol. 57, no. 9, pp. 2105–2113, Sep. 2009. [16] R. N. Braithwaite and S. Carichner, “An improved Doherty amplifier using cascaded digital predistortion and digital gate voltage enhancement,” IEEE Trans. Microw. Theory Techn., vol. 57, no. 12, pp. 3118–3126, Dec. 2009. [17] O. Hammi, S. Carichner, B. Vassilakis, and F. M. Ghannouchi, “Synergetic crest factor reduction and baseband digital predistortion for adaptive 3G Doherty power amplifier linearizer design,” IEEE Trans. Microw. Theory Techn., vol. 56, no. 11, pp. 2602–2608, Nov. 2008. [18] C. Yu, L. Guan, E. Zhu, and A. Zhu, “Band-limited volterra seriesbased digital predistortion for wideband RF power amplifiers,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 12, pp. 4198–4208, Dec. 2012. [19] J. Xia, X. Zhu, L. Zhang, J. Zhai, J. Wang, M. Yang, and Y. Sun, “Highly efficient GaN Doherty power amplifier with 100 MHz signal bandwidth for 3.5 GHz LTE-advanced application,” in IEEE MTT-S Int. Microw. Symp. Dig., Montreal, QC, Canada, Jun. 2012, pp. 1–3. [20] H. Wu, J. Xia, J. Zhai, L. Tian, M. Yang, L. Zhang, and X. Zhu, “A wideband digital pre-distortion platform with 100 MHz instantaneous bandwidth for LTE-advanced applications,” in Proc. Int. Integr. Nonlinear Microw. Millimeter-Wave Circuits Workshop, Sep. 2012, pp. 1–3. [21] A. Zhu, P. J. Draxler, J. J. Yan, T. J. Brazil, D. F. Kinball, and P. M. Asbeck, “Open-loop digital predistorter for RF power amplifiers using dynamic deviation reduction-based Volterra series,” IEEE Trans. Microw. Theory Techn., vol. 56, no. 7, pp. 1524–1534, Jul. 2008.


2921

[22] J. Moon, J. Kim, I. Kim, Y. Woo, S. Hong, H. Kim, J. Lee, and B. Kim, “GaN HEMT based Doherty amplifier for 3.5-GHz WiMAX applications,” in Proc. 37th Eur. Microw. Conf., Sep. 2007, pp. 1193–1196. [23] J. Park, D. Kim, C. Yoo, W. Lee, J. Yook, and C. Hahn, “Efficiency enhancement of the Doherty amplifier for 3.5 GHz WiMAX application using class-F circuitry,” Microw. Opt. Tech. Lett., vol. 52, no. 3, pp. 570–573, Mar. 2010. [24] J. Fang, J. Moreno, R. Quaglia, V. Camarchia, M. Pirola, S. Guerrieri, C. Ramella, and G. Ghione, “3.5 GHz WiMAX GaN Doherty power amplifier with second harmonic tuning,” Microw. Opt. Technol. Lett., vol. 54, no. 11, pp. 2601–2605, Nov. 2012. [25] C. Musolff, M. Kamper, Z. Abou-Chahine, and G. Fischer, “A linear and efficient Doherty PA at 3.5 GHz,” IEEE Microw. Mag., vol. 14, no. 1, pp. 95–101, Jan. 2013. [26] A. Z. Markos, K. Bathich, and G. Boeck, “Design of GaN HEMT based Doherty amplifiers,” in IEEE Wireless Microw. Technol. Conf., Apr. , pp. 1–5. [27] S. Wood, R. Pengelly, and J. Crescenzi, “A high efficiency Doherty amplifier with digital predistortion forWiMAX,” High Frequency Electron., vol. 7, no. 12, pp. 18–28, Dec. 2008. [28] Y. Lee, M. Lee, S. Kam, and Y. Jeong, “Highly linear and efficient asymmetrical Doherty power amplifiers with adaptively bias-controlled predistortion drivers,” in IEEE MTT-S Int. Microw. Symp. Dig., Boston, MA, USA, Jun. 2009, pp. 1393–1396.

technologies for wireless communications, as well as microwave and millimeter-wave theory and technology. He is also interested in PA nonlinear character and its linearization research with a particular emphasis on wideband and high-efficiency GaN PAs. Dr. Zhu was the recipient of the 1994 First-Class Science and Technology Progress Prize presented by the Ministry of Education of China and the 2003 Second-Class Science and Technology Progress Prize of Jiangsu Province, China.

Jing Xia (S’12) received the B.E. degree in communication engineering and M.E. degree in computer science and technology from Jiangsu University, Zhenjiang, China, in 2003 and 2007, respectively, and is currently working toward the Ph.D. degree at Southeast University, Nanjing, China. His current research interests include highly efficient RF PA design and digital predistortion techniques.

Jianfeng Zhai received the B.S. degree in radio engineering and Ph.D. degree in electromagnetic field and microwave technology from Southeast University, Nanjing, China, in 2004 and 2009, respectively. He is currently with the School of Information Science and Engineering, Southeast University. His current research interests include digital signal processing, neural networks, nonlinear modeling, microwave circuits design, PA linearization, and embedded systems.

Xiaowei Zhu (S’88–M’95) was born in Nanjing, Jiangsu Province, China, in 1963. He received the B.E., M.E., and Ph.D. degrees in radio engineering from Southeast University, Nanjing, China, in 1984, 1996, and 2000, respectively. Since 1984, he has been with Southeast University, Nanjing, China, where he is currently a Professor with the School of Information Science and Engineering. He has authored or coauthored over 80 technical publications. He holds ten patents. His research interests include RF and antenna

Yinjin Sun received the B.S. degree in electronic and information engineering from Xidian University, Xi’an, China, in 2009, and is currently working toward the Ph.D. degree in information science and engineering at Southeast University, Nanjing, China. His current research interests include high-efficiency and dual-band Doherty PA design and linearization techniques.

Lei Zhang received the M.S. degree in signal and information processing from Southeast University, Nanjing, China, in 1999, and the Ph.D. degree in electromagnetic field and microwave technology from Southeast University, Nanjing, China, in 2009. He is currently with the School of Information Science and Engineering, Southeast University. His current researches include highly linear and efficient RF/microwave PA design, nonlinear modeling, and linearization techniques.