www.ietdl.org Published in IET Microwaves, Antennas & Propagation Received on 12th May 2013 Revised on 25th September 2013 Accepted on 2nd October 2013 doi: 10.1049/iet-map.2013.0241 ISSN 1751-8725 Complementary metal-oxide semiconductor Doherty power amplifier based on voltage combining method Chenxi Zhao, Byungjoon Park, Bumman Kim Department of Electrical Engineering, Pohang University of Science and Technology (POSTECH), Pohang, Gyeongbuk 790-784, Republic of Korea E-mail: cxzhao@postech.ac.kr Abstract: A 1.75 GHz Doherty power amplifier (PA) is designed and implemented in a 0.18-µm complementary metal-oxide semiconductor (CMOS) process. This Doherty PA uses a voltage combining transformer to combine the output power and realise the load modulation which is different from conventional current combining Doherty amplifiers. The prototype has a power-added efficiency (PAE) of 31.6% at a maximum output power of 28.6 dBm from 3.4 V supply voltage. The PAE at 6 dB back-off is still high, about 25%. It shows clearly the efficiency enhancement at the power back-off point because of the Doherty operation. This is the first use of voltage combining techniques in CMOS Doherty PA design. 1 Introduction The power amplifier (PA) is the main energy consumption component in wireless portable device. When the signals have a high peak to average power ratio, the PA should work at a back-off point from the peak power to ensure linear operation. However, power-added efficiency (PAE) at the low power level is poor because of the fixed supply voltage and the load impedance optimised at the maximum power level. Consequently, a large portion of the battery power is wasted at the low power amplification. Therefore, improving the efficiency of a PA at the power back-off condition can save energy and increase battery lifetime. The Doherty architecture is a good choice to enhance the back-off efficiency because only half of the power cell works in the low power region [1, 2]. Recently, the increasing number of RF components are being integrated using the complementary metal-oxide semiconductor (CMOS) process to meet the requirements of cost reduction and small size in wireless consumer markets. However, the CMOS PA is difficult to design because of the low breakdown voltage, conductive Si substrate and lack of ground via. Here, a 1.75 GHz Doherty PA is designed and implemented in a 0.18-µm CMOS process. This Doherty PA uses a voltage combining transformer to combine the output power and realise the load modulation which is different from the conventional current combining Doherty amplifier. The prototype has a power-added efficiency (PAE) of 31.6% at a maximum output power of 28.6 dBm from a 3.4 V supply voltage. The PAE at 6 dB back-off is still high, at about 25%. It shows clearly the efficiency enhancement at the power back-off point due to the Doherty operation. This is the first reported use of voltage combining techniques in CMOS Doherty PA design. IET Microw. Antennas Propag., 2014, Vol. 8, Iss. 3, pp. 131–136 doi: 10.1049/iet-map.2013.0241 2 Voltage combining method of Doherty amplifier The Doherty technique is based on load modulation at the output. The output load is modulated by the current ratio between the carrier and peaking amplifiers [3]. The carrier amplifier operates in class AB or class B, and the peaking amplifier operates in class C. For a low power level, only the carrier amplifier is turned on. The peaking amplifier turns on at the power level of 6 dB back-off from the peak power. A quarter wavelength transmission-line (λ/4 T-line) combines the power from both amplifiers and realises the load modulation. The conventional current combining structure places the λ/4 T-line behind the carrier amplifier as shown in Fig. 1. It acts as an impedance inverter, which causes the resistive impedance seen by the carrier amplifier to decrease as the peaking amplifier current increases. In CMOS PA design, the voltage combining method is utilised and the current combining method can cause problems. In order to obtain high output power, the optimum resistance (Ropt) of CMOS amplifier is small, and the impedance after the current combining is Ropt/2 leading to a large loss when it is matched to 50 Ω, thus limiting the bandwidth. Second, the standard RF CMOS process does not have a ground via that can connect transistor’s source to ground perfectly. Therefore, the differential structure is widely used to supply a virtual ground. Last, the transformer provides a low loss matching circuit on conductive Si substrate [4]. The voltage combining transformer is a good solution to solve those problems. Owing to the voltage combining output matching of the CMOS PA, it is advantageous to design a Doherty amplifier based on the voltage combining method. Fig. 2 depicts an idealised schematic of the CMOS Doherty PA based on the transformer. To simplify the analysis, the 131 & The Institution of Engineering and Technology 2014 www.ietdl.org Fig. 1 Block diagram of conventional current combining Doherty PA ideal voltage combining transformer is used. Theoretically, at the maximum output power, the peaking and the carrier amplifiers should have the same fundamental output voltages (Vp = Vc = V ) and currents (Ip = Ic = I ) so that the two amplifiers have the same load impedance (Zp = Zc = Ropt) and generate the same power. In this structure, the AC voltages on the secondary loops are added, while the primary loops are driven by the voltages of the peaking and the carrier amplifiers. However, the current of the secondary loops are identical to that of the primary loops (Iout = I ) because the transformer is assumed as ideal. The Zout can then be calculated as Vout I out +VP − −VP + (+VC ) − (−VC ) = I VP VP VC VC + + + = I I I I = Zc+ + Zc− + Zp+ + Zp− 4V = 4Ropt = I Zout = (1). This series combining is the main difference from the current combining in parallel. Compared to current combining, voltage combining has a smaller impedance transformation ratio when matched to 50 Ω and a wider bandwidth. At a low power level (0 ≤ Vin ≤ 0.5Vin,max), the peaking amplifier is turned off and sees infinite impedance Zpeaking+ = Zpeaking− = 1 The λ/4 T-line which is placed after the peaking amplifier has a characteristic impedance of Ropt. It functions as an impedance inverter. Consequently, the impedances seen at the transformer’s input are Zp+ = Zp− = 0 (3) Therefore, the carrier amplifier sees the load impedance of Zc+ = Zc− = (1) Therefore, the output impedance of the voltage combining transformer Zout is fixed to four times of Ropt of the power cell, which is four-way series combined, as expressed in (4) Operation of proposed CMOS Doherty PA 3.1 Fig. 2 Block diagram of an ideal voltage combining Doherty PA 4Ropt − 2 × 0 = 2 × Ropt 2 Hence, the carrier amplifier sees twice the larger impedance of the optimum resistance (2Ropt). This high impedance leads to premature saturation of the carrier amplifier while the current Ic is only half of its maximum value. At a high power level (0.5Vin,max ≤ Vin < Vin,max), the carrier amplifier reaches its maximum output swing and the peaking amplifier turns on the generating current Ip. Increasing of Ip decreases the effective impedance of (Zpeaking+) and (Zpeaking−) from infinity. Owing to the λ/4 T-line, (Zp+) and (Zp−) increase from 0 to Ropt and (Zc+) and (Zc−) decrease from 2Ropt to Ropt, allowing the carrier amplifier to supply more current and keep the output voltage swing at its peak value. At the peak power (Vin = Vin,max), (Zp+) and (Zp−) change from 0 to Ropt and this results in (Zc+) and (Zc−) decreasing from 2Ropt to Ropt. Both amplifiers can work with the peak efficiency and achieve the second peak in the overall efficiency. The impedance variation curves are shown in Fig. 3, which is the ideal load modulation for Doherty operation. 3 132 & The Institution of Engineering and Technology 2014 (2) Circuit design methodology Based on the basic concept of the voltage combining technique, a CMOS Doherty amplifier has been designed. The circuit topology is shown in Fig. 4. The attached IET Microw. Antennas Propag., 2014, Vol. 8, Iss. 3, pp. 131–136 doi: 10.1049/iet-map.2013.0241 www.ietdl.org T-lines are employed. The second harmonic short circuits are attached at the drain of the each power cell to improve the linearity. 3.2 Fig. 3 Load impedance variation of an ideal voltage combining Doherty PA capacitors at the transformer tune the output impedance for 50 Ω matching. To reduce the chip size and facilitate high integration, lumped element pi networks instead of λ/4 Voltage combining transformer design The voltage combining transformer plays a key role in this design. It not only combines the output power from both amplifiers but also realises the output impedance matching. More importantly, it works with the λ/4 T-line as an inverter and reduces the load impedance of the carrier amplifier when the input power increases. The transformer used in this work is shown in Fig. 5. It has an 8-shape composed of two loops and the loops are twisted to have an opposite current flow, immune to the common mode oscillation [5–7]. The secondary loop is placed between the double primary loops to enhance the coupling factor k and reduce the primary inductance. Both the loops utilise the top layer (4.0 µm thickness copper metal) and dual secondary loops are connected with lower layers. The Fig. 4 Schematic of the proposed CMOS Doherty PA Fig. 5 Eight-shape transformer IET Microw. Antennas Propag., 2014, Vol. 8, Iss. 3, pp. 131–136 doi: 10.1049/iet-map.2013.0241 133 & The Institution of Engineering and Technology 2014 www.ietdl.org Table 1 Transformer characteristics at 1.75 GHz Primary inductance Lp, nH 0.66 Secondary inductance Ls, nH Primary resistance Rp, Ω Secondary inductance Rs, Ω Primary quality factor Qp Secondary quality factor Qp Coupling factor k Turn ratio 3.4 0.6 5.6 12.04 6.67 0.73 1:2.27 transformer characteristics at 1.75 GHz are listed in Table 1. The test pattern of the transformer is shown in Fig. 6. To measure the three-port transformer using a two-port network analyser, the same two transformers are laid out in a back-to-back configuration. The differential ports of one transformer are directly connected to the other transformer’s Fig. 6 Test pattern of transformer Fig. 7 Simulation results of input impedance 134 & The Institution of Engineering and Technology 2014 differential ports. The insertion loss of a single transformer is half the total loss in dB. The insertion loss of the single transformer is 1.23 dB (efficiency of 76%) at 1.75 GHz. 3.3 Input driven unevenly Besides the output transformer design, the input power dividing ratio is very important for proper Doherty operation [8]. Owing to the different bias conditions of the two amplifiers, the gain of the peaking amplifier is lower than that of the carrier amplifier. For this reason, an uneven drive is adopted to improve the output power and it can be realised through matching the fundamental input impedance of the carrier and the peaking amplifiers to different points as shown in Fig. 7. At a low power operation, the fundamental input impedance of the peaking amplifier is highly mismatched and the carrier amplifier is slightly mismatched. Hence, more input power is driven to the carrier amplifier, improving the gain and preventing the peaking amplifier turning on early. At a high power level, the fundamental input impedance of the carrier amplifier remains almost constant with the input power variation, while that of the peaking amplifier changes significantly because of the class-C bias. Therefore, the fundamental input impedance of the carrier amplifier is still slightly mismatched and that of the peaking amplifier is matched to the port impedance at the maximum output power condition. The differential structure offers a cancellation effect of even-order harmonics, and second harmonic short circuits are attached at the output. Moreover, the second harmonic input impedance affects efficiency very slightly. The third harmonic input impedance of the carrier and the peaking amplifiers are located around the open area as shown in Fig. 7. Consequently, the peaking amplifier obtains more input power, compensating for the smaller currents and gain due to the low bias. Simulation results are shown in Fig. 8. The carrier input power is reduced by 1.5 dB, and 1.5 dB more power is delivered to the peaking Fig. 8 Simulation results of uneven drive IET Microw. Antennas Propag., 2014, Vol. 8, Iss. 3, pp. 131–136 doi: 10.1049/iet-map.2013.0241 www.ietdl.org Fig. 10 Fabricated chip photograph of CMOS Doherty power amplifier Fig. 9 Simulation results for input power-dependent variations a Device voltages b Device currents Fig. 11 Measured PAE, power gain and DC consumptions against output power at 1.75 GHz amplifier at the maximum output power, realising the uneven driven for optimum operation of the Doherty PA. Thus, at the maximum output power point, both amplifiers can have the same fundamental voltages and currents, generating the same output power. Simulation has been carried out to check the load modulation behaviour. Fig. 9 shows the current and voltage amplitudes of the carrier and peaking devices over the whole input power range. The peaking amplifier does not consume any current at a low power region until 18 dBm input power. The voltage and current of it are close to 0. For the upper 6 dB region, the current and voltage of the peaking amplifier increase very quickly while the voltage of carrier amplifier remains almost constant at 3.0 V. At the maximum power point, the carrier and peaking amplifiers achieve the same amplitudes of current and voltage. 4 Implementation and measurement results of the CMOS Doherty PA The proposed CMOS Doherty PA has been fabricated using a 0.18-µm CMOS process. A chip micrograph is shown in Fig. 10. The prototype was assembled on a four-layer FR-4 test board. The PA output pads and 3.4-V power supply pads are double-wire bonded to minimise the bonding wire inductance. The loss of the bonding wire, the input balun and the output transformer are included in the measurement results. IET Microw. Antennas Propag., 2014, Vol. 8, Iss. 3, pp. 131–136 doi: 10.1049/iet-map.2013.0241 Fig. 12 Two-tone measurement results at 1.75 GHz with a tone spacing 10 MHz The test results for a 1.75 GHz one-tone signal are given in Fig. 11. The PA achieves 28.6 dBm maximum output power with a peak PAE of 31.6%. The PAE is kept above 25% over a 6 dB range of output power. This prototype shows a significant improvement of PAE at the power back-off 135 & The Institution of Engineering and Technology 2014 www.ietdl.org 6 Table 2 Comparison of CMOS performance Design Technology Pout, max, dBm PAE (%) at 6 dB back-off/ peak Technique transformer combining current combining current combining voltage combining [9] 45 nm 31.5 20/27 [10] 0.13-µm 31.5 20/33 [11] 90 nm 26.5 12.2/39 this work 0.18-µm 28.6 25.1/31.6 point. The DC current consumption of the carrier and the peaking amplifiers is also shown in Fig. 11. At the maximum output power point, both amplifiers consume almost the same current and generate the same output power. This result demonstrates the successful realisation of the uneven drive. Fig. 12 shows the measured IMDs from the two-tone test. Table 2 compares this work with other reported CMOS PA designs. This work achieves the highest 6 dB back-off efficiency performance using the voltage combining method. 5 Conclusion A CMOS Doherty PA using a voltage combining method has been proposed. It is believed that this is the first reported use of the voltage combining technique in CMOS Doherty PA design. The λ/4 T-line and voltage combining transformer work together as an impedance inverter and reduce the load impedance of the carrier amplifier when the peaking amplifier turns on, realising the load modulation characteristic of Doherty operation. With the Doherty technique, the amplifier’s PAE is kept above 25% over a 6 dB range of output power. The proposed method of voltage combining is an attractive solution in the design of CMOS Doherty PA. 136 & The Institution of Engineering and Technology 2014 Acknowledgment This research was supported by the World Class University program funded by the Ministry of Education, Science and Technology through the National Research Foundation of Korea (R31-10100), and by the Brain Korea 21 Project in 2012. 7 References 1 Kang, D., Kim, D., Moon, J., Kim, B.: ‘Broadband HBT Doherty power amplifiers for handset applications’, IEEE Trans. Microw. Theory Tech., 2010, 12, (58), pp. 4031–4039 2 Cho, K., Kim, J., Stapleton, S.P.: ‘A highly efficient Doherty feedforward linear power amplifier for W-CDMA base-station applications’, IEEE Trans. Microw. Theory Tech., 2005, 1, (53), pp. 292–300 3 Kang, J., Yu, D., Min, K., Kim, B.: ‘A ultra-high PAE Doherty amplifier based on 0.13-µm CMOS process’, IEEE Microw. Wirel. Compon. Lett., 2006, 9, (16), pp. 505–507 4 Kang, J., Yoon, J., Min, K., et al.: ‘A highly linear and efficient differential CMOS power amplifier with harmonic control’, IEEE J. Solid-State Circuits, 2006, 6, (41), pp. 1314–1322 5 Jin, B., Moon, J., Zhao, C., Kim, B.: ‘A 30.8-dBm wideband CMOS power amplifier with minimized supply fluctuation’, IEEE Trans. Microw. Theory Tech., 2012, 6, (60), pp. 1658–1666 6 Liu, G., Haldi, P., Liu, T.K., Niknejad, A.M.: ‘Fully integrated CMOS power amplifier with efficiency enhancement at power back-off’, IEEE J. Solid- State Circuits, 2008, 3, (43), pp. 600–609 7 Haldi, P., Debopriyo, D., Min, P., Reynaert, D., Liu, G., Niknejad, A.M.: ‘A 5.8 GHz 1 V linear power amplifier using a novel on-chip transformer power combiner in standard 90 nm CMOS’, IEEE J. Solid- State Circuits, 2008, 5, (43), pp. 1054–1064 8 Kang, D., Choi, J., Kim, D., Kim, B.: ‘Design of Doherty power amplifiers for handset applications’, IEEE Trans. Microw. Theory Tech., 2010, 8, (58), pp. 2134–2142 9 Tai, W., Xu, H., Ravi, A., et al.: ‘A 31.5 dBm outphasing class-D power amplifier in 45 nm CMOS with back-off efficiency enhancement by dynamic power control’. Proc. ESSCIRC, September 2011, pp. 131–141 10 Wongkomet, N., Tee, L., Gray, P.: ‘A 31.5 dBm CMOS RF Doherty power amplifier for wireless communications’, IEEE J. Solid-State Circuits, 2006, 12, (41), pp. 2852–2859 11 Elmala, M., Paramesh, J., Soumyanath, K.: ‘A 90-nm CMOS Doherty power amplifier with minimum AM-PM distortion’, IEEE J. Solid-State Circuits, 2006, 6, (41), pp. 1323–1332 IET Microw. Antennas Propag., 2014, Vol. 8, Iss. 3, pp. 131–136 doi: 10.1049/iet-map.2013.0241