FR2A-2 Proceedings of Asia-Pacific Microwave Conference 2010 A 57-66 GHz Medium Power Amplifier in 65-nm CMOS Technology Chia-Yu Hsieh 1, Jhe-Jia Kuo 2, Zuo-Min Tsai 3, Kun-You Lin 4 Department of Electrical Engineering and Graduated Institute of Communication Engineering, National Taiwan University No. 1, Sec. 4, Roosevlt Road, Taipei, 10617 Taiwan 1 charlenefish@gmail.com 2 f94942020@ntu.edu.tw 3 zuomintsai@gmail.com 4 kunyou@ntu.edu.tw Abstract — This paper presents the design and measurement results of a 57-66 GHz medium power amplifier in 65-nm LP CMOS process. This amplifier is designed with broadband matching concern, which can achieve a measured gain more than 21 dB from 57-66 GHz and have a 3-dB bandwidth more than 14 GHz while consuming 54 mW from a 1.2 V supply. The measured results exhibit Psat of 10.3 dBm, P1dB of 6.2 dBm, and the peak PAE is 16 % at 58 GHz. The chip size is only 0.3 mm2. Index Terms — Broadband, CMOS, 60 GHz, MMIC, power amplifiers. In this paper, the design flow and strategy of a CMOS medium power amplifier is presented. Especially the wideband design techniques for power amplifier are discussed. With the proper device selection and wideband matching networks, the proposed medium power amplifier achieves a Psat of 10.3 dBm and 16% PAE. The design methods described in this paper can be used in CMOS wideband power amplifier designs. II. CIRCUIT DESIGN I. INTRODUCTION To meet the requirement of high data rate, high speed, and high quality wireless communication, high frequency circuit design becomes more and more necessary. Since Federal Communications Commission (FCC) released unlicensed band from 57-66 GHz to support Wireless Personal Area Networks (WPANs) in IEEE 802.15.3C, the researches of high frequency integrated circuit grow rapidly [1]. Recently, CMOS millimeter-wave circuits become more popular because of the high level integration with digital circuits. Compared to III-V compound technologies for high frequency circuits, the characteristic of rather low power consumption makes CMOS technology more suitable in hand-held devices. Power amplifier plays an important role in the RF transceiver circuits. It must deliver sufficient output power to drive the following power amplifier or antenna. Recently, numerous 60 GHz medium power amplifiers realized by CMOS process were reported. Multi-stage common-source medium power amplifiers are presented to achieve a linear gain of 13-dB and a saturated output power of 7.9-dBm at 60 GHz [2]-[3]. Transformer coupled technique is used to realize a three-stage pseudo-differential power amplifier to achieve a 16-dB linear gain, and a 11.5-dBm saturated output power with a peak PAE of 15.2 % at 60 GHz [4]. In addition, combined common-source topology at output stage [5] and cascode topology at prior stages [6] are also applied in 60 GHz medium power amplifiers. However, some previous works, cannot achieve high gain and high output power at the same time, and others cannot achieve high gain and high output power within wide bandwidth. Copyright 2010 IEICE This circuit was fabricated in 65-nm CMOS LP process in 1.2 V applications, which provides one poly layer for the gates of CMOS transistors and nine metal layers with ultrathick top metal (metal 9) of 3.4 μm while featuring deep Nwell (DNW). The RF NMOS transistors can achieve maximum oscillation frequency (fmax) of around 140 GHz with 1.2 V supply. In addition, metal-insulator-metal (MIM) capacitors are available between metal 8 and metal 7. The circuit schematic of proposed medium power amplifier is shown in Fig. 1. This amplifier consists of four stages common-source devices with input, output and interstage matching networks. In order to achieve high output power, the size of output stage transistor (M4) needs to be chosen carefully. In general, the device with larger size can deliver higher output power while dissipating more dc power. Therefore, making a compromise to achieve design goal can decide proper size of M4. The final chosen M4 with width of 63 μm and length of 65 nm can deliver maximum power of 9.9 dBm and maximum available gain of about 9 dB from 5766 GHz under class A operation. To achieve high gain performance without high power consumption, other stages device can be estimated from prior stage. Since the total width of device is proportional to the output power it can deliver, the device size of M3 can be calculated as 1617 OP3,1dB IP4,1dB OP4,1dB OP3,1dB w4 w3 (1) (2) Fig. 1 Schematic of proposed medium power amplifier. , where IP4,1dB, OP4,1dB, and w4 are decided according to the simulation already. Therefore, after the determination of the OP3,1dB under the condition of (1), w3 can be computed by (2). In this circuit, we choose OP3,1dB IP4,1dB 3 (dB) (3) to guarantee that the power won’t saturate at M3. Finally, the device sizes of other stages can also be calculated by the same formulas. (a) Output Matching Network 5 Term 1 50 Ω 3 4 1 2 Term 2 50 Ω (b) Fig. 2 (a) Simulated load-pull points and S11 of output matching from 57-66 GHz with matching path at 60 GHz. (b) Output matching network. The matching networks of this circuit are realized by thinfilm microstrip (TFMS) lines with metal 9 as the signal lines and metal 1 as the ground. In addition, the microstrip lines with 60 Ω of characteristic impedance are used in matching networks. In general, output matching network is the most essential part in power amplifier since it can affect the delivered power of amplifier directly. Therefore, the matching network of the power amplifier should be designed in sequence from output matching network, inter-stage matching network, to input matching network. The output matching circuit design is plotted in Fig. 2 (a) and (b). The numbers in Fig. 2 (a) represent the sequence of 60 GHz matching path and the numbers in Fig. 2 (b) represent the corresponding matching stubs. Notice that number 1 and 5 are absent on matching path since these two series line are used to connect terminal with little matching effect. The output matching is designed to make the input impedance close to simulated load-pull positions from 57-66 GHz which can result in broadband power matching, as shown in Fig. 2 (a). Notice that the load-pull positions from 57-66 GHz are very close. The matching network which consists of one open stub, one series line and one short stub with two series line connecting to two terminals can achieve broadband matching. Open stubs, series lines and short stubs affect S-parameters differently from high frequency to low frequency. Open stubs and series line manifest themselves on higher frequency while short stubs manifest themselves on lower frequency. Using these frequency-dependent properties, the loci of S11 in the interested frequency band can finally circle the load-pull position on the Smith chart. Each inter-stage matching network is matched to output conjugate impedance of prior stage device. The devices of prior stages are smaller with high gain and low current consumption benefits. However, stability will be a challenge at inter-stage matching. In addition to fine tune the matching network, source degeneration can make conjugate matching easily with a little sacrifice on linear gain. The input matching circuit design is plotted in Fig. 3 (a) and (b). The numbers in Fig. 3 (a) represent the sequence of 60 GHz matching path and the numbers in Fig. 3 (b) represent the corresponding matching stubs. The input matching network is designed to match to 50 Ω from 57-66 GHz with the same broadband matching technique in the output matching network. It can be observed that S11 circles 50 Ω 1618 peak gain of 23.7 dB at 64 GHz while return losses are larger than 7.8 dB from 57-66 GHz. Power measurement was done by adding signal generator Agilent E8257D at input while a power meter Agilent E4419 and a power sensor Agilent V8486A was set to measure the output power at output. Fig.6 shows the large signal behavior 50j 25j 100j 57 GHz 10j 250j 4 3 60 GHz 10 25 50 5 100 RFout VDD 250 2 Vg 66 GHz -10j -250j Vg 1 -25j GND -100j RFin Input Matching from 0-100 GHz Matching Path at 60 GHz -50j Fig. 4 The chip micro-photograph of proposed medium power amplifier with chip size of 0.36 × 0.8 mm2 including all the testing pads. (a) Interstage Matching Network Input Matching Network 5 3 1 M1 Term 1 50 Ω 4 30 2 20 Output Matching Network M4 M3 S-Parameters (dB) Interstage Matching Network Interstage Matching Network Term 2 50 Ω M2 (b) Fig. 3 (a) S11 of input matching from 0-100 GHz with matching path at 60 GHz. (b) Input matching network. 10 0 -10 -20 Measured |S21| Measured |S11| Measured |S22| -30 -40 40 50 60 70 80 Frequency (GHz) Fig. 5 Measured small signal gain and I/O return loss of the proposed medium power amplifier under VDD of 1.2 V. 1619 20 30 15 25 10 20 5 15 0 10 Pout PAE Gain -5 5 -10 0 -30 -25 -20 -15 -10 -5 0 Pin (dBm) Fig. 6 Measured Pout, PAE, and Gain versus Pin at 58 GHz. 5 Gain (dB) A four-stage medium power amplifier is designed focusing on achieving high gain and high output power with reasonable power consumption and compact size. The chip micro-photograph is shown in Fig. 4 with a chip size of 0.36 × 0.8 mm2 including all the testing pads. The measurements are performed by on-wafer probing. The vector network analyzer Agilent E8361C and Anritsu 37397D used to measure the small signal S-parameters from 40 MHz65 GHz and 65-90 GHz respectively. As all four transistors are biased at class A operation for gain and linearity concern, total dc current consumption is 45 mA with V DD of 1.2 V and Vg of 0.8 V. Fig. 5 shows the measured small signal gain and return losses of this amplifier. This amplifier achieves a measured gain of more than 21 dB from 57-66 GHz with Pout (dBm) III. EXPERIMENTAL RESULT PAE (%) from 57-66 GHz to achieve the broadband matching. Since the input and output matching networks dominate the bandwidth of port return loss, both the input and output matching are designed with more sections than that in the inter-stage matching, which can be seen in Fig. 1. P1dB, Psat (dBm) 18 10 15 8 12 6 9 4 6 P1dB Psat 2 measured. This circuit achieves high gain and high output power at the same time with simple structure but careful broadband matching concern. From a 1.2-V supply, Psat of 10.3 dBm is achieved with PAE of 16 % at 58 GHz. The linear gain is above 21 dB from 57-66 GHz with 3-dB bandwidth of 14 GHz while consuming 54 mW. Peak PAE (%) 12 ACKNOWLEDGEMENT This work was supported in part by the National Science Council of Taiwan, R.O.C. (under NSC 98-2219-E-002-009), the Excellent Research Projects of National Taiwan University (98R0062-03, 99R80300) and the University Shuttle Program of Taiwan Semiconductor Manufacturing Company (TSMC), Hsin-chu, Taiwan. The EDA tools are provided by National Chip Implementation Center (CIC), Hsinchu, Taiwan. 3 Peak PAE 0 57 58 59 60 61 62 63 64 65 0 66 Frequency (GHz) Fig. 7 Measured power performance from 57-66 GHz, which including P1dB, Psat, and peak PAE. of this amplifier with increasing input power at 58 GHz. A Psat of 10.3 dBm and a peak PAE of 16% are achieved. Fig. 7 shows measured P1dB, Psat, and peak PAE from 57-66 GHz. It can be seen that the P1dB is larger than 4 dBm from 57-66 GHz with maximum P1dB of 6.2 dBm at 58 GHz. The Psat is larger than 8.2 dBm from 57-66 GHz, and a maximum Psat of 10.3 dBm is achieved at 58 GHz. From 57-66 GHz, the peak PAE exceeds 9.5 %. Table I summarizes the performance of reported 60-GHz CMOS medium power amplifiers. This work demonstrates highest gain with broadband feature while achieving a maximum Psat of 10.3 dBm and a peak PAE of 16 % with reasonable power consumption of 54 mW and compact chip size of 0.3 mm2. VI. CONCLUSIONS A wideband medium power amplifier at 60 GHz using 65 nm CMOS process has been designed, fabricated and REFERENCES [1] IEEE P802.15-05-0596-01-003c [2] M. Varonen, M. Kärkkäinen, M. Kantanen, and K. A. I. Halonen, “Millimeter-wave integrated circuits in 65-nm CMOS,” IEEE J. Solid-State Circuits, vol. 43, no. 9, pp. 1991– 2002, Sep. 2008. [3] Dan Sandström, Mikko Varonen, Mikko Kärkkäinen and Kari Halonen, “60 GHz amplifier employing slow-wave transmission lines in 65-nm CMOS,” NORCHIP, 2008. pp. 2124, Nov. 2008. [4] Wei L. Chan and John R. Long, “A 58–65 GHz neutralized CMOS power amplifier with PAE above 10% at 1-V supply,” IEEE J. Solid-State Circuits, vol. 45, no. 3, pp. 554-564, Mar. 2010. [5] Debasis Dawn, Saikat Sarkar, Padmanava Sen, Bevin Perumana, Matthew Leung, Navin Mallavarpu, Stephane Pinel, and Joy Laskar, “60GHz CMOS power amplifier with 20-dBgain and 12dBm Psat,” in IEEE MTT-S Int. Microw. Symp. Dig., 2009, pp. 537-540. [6] S. Pinel, S. Sarkar, P. Sen, B. Perumana, D. Yeh, D. Dawn, and J. Laskar, “A 90 nm CMOS 60 GHz radio,” in IEEE Int. SolidState Circuits Conf. Tech. Dig., Feb. 2008, pp. 130–131. TABLE I REPORTED PERFORMANCES OF 60-GHZ MEDIUM POWER AMPLIFIERS. Ref . [3] [4] [5] [6] This Work Process 65-nm CMOS 65-nm CMOS 90-nm CMOS 90-nm CMOS 65-nm CMOS Topology 3-stage CS 3-stage CS Diff. w/ 1-stage cascade+2-stage 1-stage cascode+2-stage baluns CS+1-stage combined 2 CS CS 4-stage CS Frequency [GHz] 51-67 58-65 57-65 57-65 57-66 (3-dB BW) (19.6) (8.5) (8) (8) (14) Peak Gain [dB] 13 16 20 17 23.7 OP1dB [dBm] 4 5 8.2 5.1 6.2 Psat [dBm] 7.9 11.5 12 8.4 10.3 Peak PAE [%] 8 15.2 9 5.8 16 PDC [mW] 54 50 146 (VDD) (1.2 V) (1 V) (1.2 V) Chip Size [mm2] 0.43 0.05* 0.65 1620 54 0.99 54 (1.2 V) 0.3