Design of Compact Wideband and Efficient Power Amplifier for Future Cellular Base Station

Wipro Tech Blogs
11 min readJan 23


By Dr. Dushyant Sharma, Esnagari Abhinaya kumari

Cellular base stations must support the advances of today’s networks by addressing the need for increased capacity and throughput at lower system cost. High-speed data transmission, support for many connected devices, low latency, low power consumption and extremely high reliability are essential for future applications. A crucial aspect of the evolution to a future cellular base station is solving complex hardware challenges. Existing towers need to provide higher performance to carry more channels at higher data rates. One way to successfully meeting expectations is through the introduction of massive multiple input, multiple output (MIMO) technology. Another approach is a large-scale integration of components, along with higher performance and greater power savings. The transition and deployment of future cellular base-station has an inherent requirement for adoption of smart power management in the underlying hardware.

A cellular base station comprises of multiple transceivers (TRX). Each TRX comprises of a power amplifier (PA), an RF small-signal section including a transmitter receiver section and Antenna as shown in Fig.1. A power Amplifier (PA) is one of the very critical components of any cellular base-station (BS). The efficiency of BS is solely decided by PA efficiency. In addition PA should be reliable to obtain a desired level of signal integrity while transmitting high power electromagnetic (EM) signal through an antenna.

Fig.1.Typical Base Station block diagram

The immense cost of the licensed wireless frequency spectra has pushed next-generation cellular standards to deploy spectrally efficient modulation schemes that enhance network capacity and data throughput [1]. Unfortunately, endeavors to escalate the transmitted bits per signal, landed in a kind of signals having a very high peak-to-average power ratio (PAPR). These signals produce significant challenges in the design of a RF power amplifier (PA), and it drastically reduces the efficiency of PA at the average (i.e., back-off) power [2]. In addition, in the cellular base stations, PA’s are dealing with a large amount of power. So, cellular network efficiency is highly affected by PA efficiency. Enhancing PA efficiency at extended back-off levels has attracted significant research focus [3].

Various PA configuration have been proposed by various researchers to enhance the back-off efficiency, such as Doherty PA, out-phasing, envelope tracking (ET) etc. Of the various configurations, the Doherty PA is one of the most favorable candidates for attaining the efficiency enhancement in the extended back-off range and is highly adopted in cellular base-stations due to its simpler implementation and reliability [4]. However, the conventional Doherty PA is intrinsically narrow in bandwidth in nature. Meanwhile, with the fast growth in wireless standards, more and wider frequency bands have been joined in wireless communication systems to provide higher data throughput. This rising number of wider bands emerge the urgent need of bandwidth expansion techniques for PAs. In addition to that with the evolution of the Future cellular base station era, this need is even more compelling since the Future cellular base station wireless standards adopted more and more new frequency spectrums and the transmit signal bandwidth is much wider. In reply to this demand, a few approaches have been proposed for extending the Doherty PA’s bandwidth. In [1] a through frequency analysis of Doherty PA is presented for wider bandwidth design. A modified impedance configuration for load modulation network (LMN) is displayed in [2] for wideband design. However, the device uses uneven bias voltages for carrier and peaking PAs. The input and output parasitic reactance of carrier and peaking PAs transistor is also a limiting factor for broadband matching and result of that, Doherty PA is narrow in bandwidth. For wideband design this reactance is compensated by adding inductance, stub, or tank circuit [3], [4], [5], [6], [7] in the design. In [8] the authors adopted mixed design approach which include proper biasing, harmonics injection technique, phase compensation network and achieved 40% fractional bandwidth. However, the proposed Doherty PA architecture is very complex and difficult to implement. To avoid the adverse effect of peaking transistor parasitic capacitance on the load. In [9] Doherty PA equipped with a half-wave transmission line (TL) or two quarter-wave transmission line (QWTL) is added in between load and peaking PA. Recently, for extending the bandwidth on [10] authors introduced a QWTL between peaking PA and the load, and in addition to that, a TL is added before a load as post matching TL. In this article, we present a modified and compact Doherty PA architecture. To achieve this an improved impedance configuration is proposed for LMN which extends the bandwidth of PA and eliminate the need of output IMN.

Doherty Power Amplifier

The theoretical investigation of Doherty PA is extensively covered in [1] and [2]. Here, we shortly review the Doherty concepts to assist the subsequent analysis. The basic architecture of conventional Doherty PA is shown in Fig.2. It has two current sources. Both are connected to a transmission line which having an electrical length (θ) of 90˚ and a characteristic impedance (ZT ) of 50 ohms. The main and auxiliary PA input currents IM and IA are controlled by respective voltages VM and VA. Let’s consider VM and VA are equal in magnitude (Vin). As equal power division is considered in this analysis, the auxiliary PA is connected in parallel to the load RL.

Fig. 2. Basic architecture of conventional Doherty power amplifier

For main device the load modulation is seen from impedance ZM and expressed as:

where VM is expressed as

and IT is expressed as

Let IZ and IA are same in magnitude and difference in phase by 90˚. Then ZM can be written as:

From Eq. 4, the load modulation of ZM is effectively described by IA, IM, RL, and ZT. By properly choosing these parameters, ZM can be follow any impedance profile. As the main device in the Doherty PA is biased in class-B mode. At an applied input voltage (Vin) to maximize the efficiency the ZM must follow the optimum load Ro:

Here Vdcm is the drain bias voltage of the main device and gm is the class B device transconductance. At the maximum input drive voltage Vin_max, corresponding current Imax/2, the optimum impedance Ropt is expressed as:

Here Imax is device saturation current.

In order to find Ro, ZM must be equal to 2Ropt and Ropt at the 6-dB output back-off (OBO) and peak power levels, respectively.

Challenges of the Doherty Power Amplifier

Bandwidth of Doherty PA was analyzed with ZM plotted along with the frequency. From that, it is observed only at the center frequency fc the real part of ZM has a peak value of 2Ropt, and it deteriorates with variation from fc. This impedance deterioration is due to presence of narrow bandwidth behavior of QWTL (TL1 in Fig.2) which is part of LMN. It will clearly degrade the drain efficiency of the device with deviation from center frequency fc as shown in Fig.3. Efficiency is highly dispersive at the OBO level and less non-dispersive at the peak power level.

Fig. 3. Effect of frequency deviation on drain efficiency of conventional Doherty PA.
Fig. 4. Proposed modified Doherty PA architecture.

Result of that conventional Doherty PA is narrow in bandwidth and LMN represents the major bottleneck in limiting the bandwidth. The fractional bandwidth of QWTL can be expressed by:

Here Γm is tolerable magnitude of reflection coefficient, ZL and ZO represent the input and output impedance of QWTL.

Fig. 5. Effect of frequency deviation on drain efficiency for proposed modified Doherty PA.

Our Solution

To enhance the bandwidth of QWTL, it is required that the difference between input and output impedances of QWTL should be low [2]. In our case, the ZM and RL are the impedances. Here we have proposed a modified impedance configuration for LMN where ZM, ZT and RL should be 50 ohm in the OBO range. Theoretically it extends the bandwidth of QWTL to ∞ and extend the BW of Doherty PA. This configuration eliminates the need of output IMN and result of that Doherty PA is compact. In addition to that fabrication of 50 ohm TL is much easier as compared to higher impedance TL [2]. The modified architecture of Doherty PA is shown in Fig.4. For the proposed device, drain efficiency is plotted in Fig.5 for different variation of frequency. From that it is observed that the proposed PA exhibits a wider bandwidth as compared to the conventional case and efficiency is non-dispersive at the OBO level and deviates a little bit at the peak power level.

The proposed design was performed on a Rogers RO4350B substrate which has dielectric constant (εr) of 3.66, loss tangent of 0.0037, and a substrate height of 0.508 mm. For main and peaking PA, two symmetric Cree devices of 10 W GaN HEMTs CGH40010F model were utilized. A hybrid coupler was used to divide the input power equally and supply 90° phase shift at the input port of the main PA. The Class-AB and C mode was used for main and peaking PA, respectively. The gate bias voltage (Vgs) of main and peaking PA were set to -2.8 V and -4.3 V, respectively. The drain bias voltage (Vds) of both the PAs were set at 28 V. The architecture of the LMN is based on Fig. 4. Here R’L, ZT, and ZM were 50 ohm during carrier power mode (or OBO power levels). In peaking power mode due to load-modulation R’L and ZM become 100 ohm and 25 ohm, respectively. The integrated layout of the Doherty PA is shown in Fig.6. To avoid EMI/EMC issues, a ground-backed co-planar waveguide (GCPW) transmission technology was used in the amplifier.

Fig. 6. Picture of integrated layout of the designed wideband Doherty PA.

The Doherty PA was characterized by a small-signal and continuous wave (CW) single-tone stimulus from 3.0 to 3.8 GHz. The S-parameter results in the frequency band from 3.1 to 3.7 GHz is shown in Fig.7.

Fig. 7. Small signal transmission (S21) and reflection (S21) S-parameters versus frequency.

In the designated band (3.07 GHz to 3.68 GHz) the PA exhibits small-signal gain (S21) which is higher than 12 dB and a return loss (S11) which is lower than -12 dB. The 6-dB OBO power and its drain efficiency versus frequency is illustrated in Fig. 8.

Fig. 8. Single-tone CW characterization of PA for 6-dB OBO power and its efficiency for different frequencies excitation.

From the graph, it is observed that in the band from 3.07 GHz to 3.68 GHz the efficiency and output power is fluctuating between 38.0% to 46.0% and 37.4 dBm to 38 dBm, respectively. A similar graph is plotted for the peak output power and its drain efficiency in Fig. 9.

Fig. 9. Single-tone CW characterization of PA for peak power and its efficiency for different frequencies excitation.

In the same band, the amplifier exhibits efficiency and peak output power is varying between 55% to 68% and 43.3 dBm to 44.6 dBm, respectively. In Fig. 10 the gain at 6-dB OBO and peak output power is plotted versus frequency.

Fig. 10. Gain at the peak and OBO power versus different frequencies excitation.

The PA exhibit at 6-dB OBO and peak output power, the gain is fluctuating between 7 dB to 9 dB and 4 dB to 5 dB, respectively. To assess the linearity of designed Doherty PA. A two-tone signal of 1 MHz frequency spacing was utilized as a stimuli for estimation of inter-modulation distortion. The plot of input and output the third order-intermodulation (TOI) versus frequency graph is shown in Fig. 11.

Fig. 11. Third-order intercept (TOI) versus frequency.

The attained value of output and input third order intercept (TOI) is varying between 48.5 to 51dB and 38 to 39.85, respectively. In order to calculate adjacent channel leakage ration (ACLR) a 64 QAM single carrier signal of 500 KHz spectrum was used. On an average power the achieved value of ACLR is -35 dBc as shown in Fig.12.

Fig. 12. ACLR (dBc) versus modulated signal power.

Future Scope

The design of Doherty PA architecture by utilizing a modified LMN has been proposed in this work. The compact and wideband design has been achieved by exploiting an improved impedance configuration at LMN. The analytical solution of the proposed LMN is presented and the load modulation process for the proposed Doherty PA is also described. The proposed PA was realized by using a 10 W GaN HEMT Cree transistor within 3.07 GHz to 3.68 GHz band. In the band, the drain efficiency at 6-dB OBO and peak output power is 38.0% to 46.0% and 55% to 68%, respectively. The device exhibits 18% fractional bandwidth. The proposed configuration is a promising solution for a wideband and high-efficiency PAs implementation for Future cellular base-stations.


[1] K. Bathich, A. Z. Markos, and G. Boeck, “Frequency Response Analysis and Bandwidth Extension of the Doherty Amplifier,” IEEE Tran. On Microwave Theory and Tech., Vol. 59, pp. 934–944, April 2011.

[2] D. Y. Wu and S. Boumaiza, “A Modified Doherty Configuration for Broadband Amplification Using Symmetrical Devices,” IEEE Tran. On Microwave Theory and Tech., Vol. 60, pp. 3201–3206, Oct 2012.

[3] J. Rubio, J. Fang, V. Camarchia, R. Quaglia, M. Pirola and G. Ghione, “3.6-GHz Wideband GaN Doherty Power Amplifier Exploiting Output Compensation Stages,” IEEE Tran. on Microwave Theory and Tech., Vol. 60, pp. 2543–2548, Aug 2011.

[4] J. H. Qureshi, N. Li, W.e. E. Neo, F. van Rijs, I. Blednov and L.e.N. de Vreede, “A Wide-Band 20W LMOS Doherty Power Amplifier, in IEEE MTT-S International Microwave Symposium, 2010.

[5] M. N. A. Abadi, H. Golestaneh, H. Sarbishaei, and S. Boumaiza, “An Extended Bandwidth Doherty Power Amplifier Using a Novel Output Combiner, in IEEE MTT-S International Microwave Symposium, 2014.

[6] N. Giovanne, A. Cidronali, P. Singer, S. Maddiol, C. Schuberth, A. D. Chiaro], G. Manes, “A 250W LDMOS Doherty PA with 31% of Fractional Bandwidth for DVB-T Applications, in IEEE MTT-S International Microwave Symposium, 2014.

[7] J. Xia, M. Yang, Y. Guo and A. Zhu, “A Broadband High-Efficiency Doherty Power Amplifier With Integrated Compensating Reactance,” IEEE Tran. on Microwave Theory and Tech., Vol. 64, pp. 2014–2024, June 2016.

[8] X. Y. Zhou, W. S. Chan, S. Y. Zheng, W. Feng, H. Liu, K. M. Cheng, and D. Ho, “A Mixed Topology for Broadband High-Efficiency Doherty Power,” IEEE Tran. on Microwave Theory and Tech., Vol. 67, pp. 1050–1063, March 2019.

[9] A. Barakat, M. Thian, V. Fusco, S. Bulja, and L. Guan, “Toward a More Generalized Doherty Power Amplifier Design for Broadband Operation,” IEEE Tran. on Microwave Theory and Tech., Vol. 65, pp. 846–859, March 2017.

[10] M. Li, J. Pang, Y. Li, and A. Zhu, “Bandwidth Enhancement of Doherty Power Amplifier Using Modified Load Modulation Network,” IEEE Tran. on Microwave Theory and Tech., Vol. 67, pp. 1824–1834, Feb 2020.

[11] D. K. Sharma and R. T. Bura, “A Novel and Compact Wideband Doherty Power Amplifier Architecture for 5G Cellular Infrastructure,” 2021 IEEE 4th 5G World Forum (5GWF), 2021, pp. 323–327, doi: 10.1109/5GWF52925.2021.00063.