A 2.469~2.69GHz AlGaN/GaN HEMT Power Amplifier for IEEE 82.16e WiMAX Applications Weijia LI 1, Yan WANG 2, Giovanni GHIONE 3, Fellow, IEEE Department of Electronics, Politecnico di Torino Torino 1129, Italy 1 weijia.li@studenti.polito.it, 2 yan.wang@studenti.polito.it, 3 giovanni.ghione@polito.it Abstract This paper presents a 2.469~2.69GHz AlGaN/GaN HEMT power amplifier for IEEE 82.16e WiMAX applications operating at E mode under a single supply of +6v. At the central frequency point, the power added efficiency (PAE) can achieve 96.37%, the small signal gain is 2.81dB, and the output power is 2.39dBm. The paper describes the circuit design in detail, then shows the simulation results and discusses about the simulation results. In the end, the paper concludes the design. Index Terms- Power Amplifier; AlGaN/GaN HEMT; IEEE 82.16e; WiMAX; PAE I. INTRODUCTION WiMAX, with the whole name Worldwide Interoperability of Microwave Access, is a new broadband wireless access technique which is based on IEEE 82.16 standard. The basic object of WiMAX is that, under the environment of Metropolitan Area Network it makes the wireless devices from different manufacturers connect together. The standards defined by IEEE for different scenarios are showed in Figure 1. This technique mainly provides the last kilometer high-speed broadband access for family, enterprise and mobile communication networks, even for the personal mobile communication business in the future. It uses a lot of advanced technologies which represent the direction of the development of communication in the future. The key technologies are including OFDM/OFDMA, HARQ, AMC, MIMO, QoS Mechanism, Sleep Mode, Handover, etc. WAN Cellular Network MAN IEEE 82.16 WiMAX LAN IEEE 82.11 Wi-Fi PAN IEEE 82.1 Bluetooth There are three frequencies for WiMAX, i.e., 2.GHz, 3.GHz and.8ghz. The frequency periods assigned to 2.GHz are: 2.3~2.32GHz, 2.3~2.4GHz, 2.34~2.36GHz and 2.469~2.69GHz. The paper focuses on the design of a 2.469~2.69GHz AlGaN/GaN HEMT power amplifier for IEEE 82.16e WiMAX applications. The power amplifier operates at E mode under a single supply of +6V. At the central frequency point, the power added efficiency (PAE) can achieve 96.37%, the small signal gain is 2.81dB, and the output power is 2.39dBm. This paper is organized as follows: Section II describes the circuit design in detail. Section III shows the simulation results and discusses about the simulation results. Section IV we conclude the design. II. CIRCUIT DESIGN A. AlGaN/GaN HEMT Technology Si created the history of semiconductor as the first generation semiconductor material. Now the silicon VLSI has entered into a practical period. As the representative of the second generation semiconductor materials GaAs has a lot of good nature in physical and chemical aspects, such as the saturation rate, cutoff frequency, and so on. GaAs has been widely used in a microwave high-power devices, especially in the RF circuit of communication that making it an ideal material for LNA, PA, etc. In recent years, with the increasing requirement of low power, high frequency, high temperature properties of electronic devices, we need to find a better semiconductor material instead of GaAs. Under this background the GaN (gallium nitride) as the representative of the wide band gap semiconductor material is generated as the third generation semiconductor material. GaN has better chemical, electrical properties comparing with the first generation semiconductor Si, Ge and second generation compound semiconductor GaAs, InP. And comparing to the current semiconductor material, GaN has the following advantages: wide band gap, high electronic saturation rate, high breakdown field, high thermal conductivity, low dielectric constant, and so on. So GaN is an ideal material to produce high-temperature, high-power, highfrequency electronic devices. Figure 1. The standards used for different scenarios 1-4244-2424-/8/$2. 28 IEEE 147 ICCS 28
TABLE I. COMPARISION OF GAN AND OTHER SEMICONDUCTOR MATERIALS Material Width of band gap (Eg/eV) Breakdown field (V cm -1 ) Thermal conductivity (W cm - 1 K -1 ) Dielectric constant Electron mobility (cm /V 2 s) GaN 3.4 >. 1 6 1. 9. 9 4H-SiC 3.2 3. 1 6 4. 1. 1 GaAs 1.43 4. 1.4 12. 8 Si 1.12 3. 1 1.4 11.9 14 The comparison of GaN and other semiconductor materials is showed in TABLE I. With the breakthrough of the thin film growth technology and the developments of ohmic contacts, annealing, corrosion and etching technologies, a variety of GaN heterostructures have been successfully grown up, including MESFET(metal semiconductor field-effect transistor), HFET(heterojunction field Effect transistor), HEMT(high electron mobility transistors), MODFET(modulation doped field-effect transistor), MISFET(metal insulation field-effect transistor) and other microwave devices, among them HEMT is most widely used. AlGaN / GaN HEMT device has wide band gap, high breakdown field, high two-dimensional electron gas concentration, high transconductance, high cutoff frequency, low noise, fast switching speed, etc. With the comparison to GaAs devices, the power density has a substantial increase. It is ideal for PA and other high-frequency circuits. Therefore the design of AlGaN / GaN HEMT power amplifier becomes a focus in the current research. B. Switch Mode Operation for High Efficiency Comparing to the conventional linear classes operation of PA, the amplifiers with displacement of peak drain/collector voltage and current can achieve a much higher efficiency. This phenomenon only happens when the transistor behaves as an ideal switch which only allows either current or voltage peak to occur at one time. In such switching mode, power dissipation in the transistor can be totally eliminated, and hence, achieving 1% efficiency, theoretically. is a switching mode power amplifier [1]. C. Optimum and Suboptimum Operation In order to achieve a high efficiency, the peak of the current and voltage waveforms for the switch must be displaced in the time. When the switch is turned on, the current flows through it with no voltage drop across. On the opposite, there will be a voltage induced when the switch is off, blocking any current flow. Thus, the two waveforms behave like two pulse trains, both with fall and rise sections occupying % of the RF period, ideally. It is required that the rise section of one waveform occurs when the other one is in its fall to avoid peaking simultaneously. Since the voltage and current of the switch are the same as these of the transistor drain, respectively, drain voltage and drain current will be used from this point forward in the thesis. Figure 2 shows the ideal waveforms for 1% efficiency. Figure 2. Ideal Switch (or Drain) Voltage and Current Waveforms in to Achieve 1% Efficiency 1) The peak drain voltage and current do not exist simultaneously. 2) At the end of the rise section of the drain current waveform, it mustdecrease to zero before the rise section of the voltage waveform can start. In other words, the current reaches zero at the end of the ON interval right before the switch is turned off. The beginning of the rise section of the voltage waveform should be delayed until after the switch is turned off. 3) The slope of the current waveform at the end of its rise section must be zero to avoid power dissipation due to the existence of both current and voltage. The similar conditions apply to the drain voltage waveform at the end of its rise section. 4) It must return to zero at the end of the switch OFF interval (right before the switch is turned on) before the rise of the current waveform can start. The starting point of the rise section of the current waveform should be postponed until after the transistor is turned on. ) Its slope is zero at that moment to avoid power dissipation due to the simultaneous imposition of current and voltage. All the above five conditions are meant to eliminate the power dissipation of the transistor as much as possible during the class E operation so that to increase the efficiency. The realization of condition 1 reduces the majority power loss. Condition 2 & 3 and 4 & are aimed to decrease the power dissipation during the ON to OFF and OFF to ON transitions of the switching process. They prevent the energy loss from the coexistence of substantial current and voltage during the transitions. Even though the power dissipation during the transition intervals is small compared to that from the 1476
coexistence of peak voltage and current, it still can decrease the efficiency dramatically, thus, degrade the class E performance. The condition 2 and 3 are known as Zero Current Switching (ZCS) and Zero Slope Current Switching (ZsCS), respectively, from the concepts of switching regulator. Similarly, condition 4 and are called Zero Voltage Switching (ZVS) and Zero Slope Voltage Switching (ZsVS). Practically, ZCS and ZsCS are very difficult to implement for frequencies greater than a few decades of MHz. Therefore, switching condition set ZCS & ZsCS and that for ZVS & ZsVS can not be achieved at the same time for high frequencies. On the contrary, the latter set is much easier to design and implement [1,2]. D. Basic Circuit Schematic In this paper, the prototype of the RF class E power amplifier proposed consists of the circuit elements synthesized in the design block diagram below [3]. E. Driver What drives the MOSFET used in class E is the voltage across the input capacitance of the transistor, which is also the gate voltage. Due to the fact that the gate DC current is always zero, the gate voltage does not dissipate power, thus, no power loss. The ideal driver for class E PAs is a pulse train with rectangular waveform shape because it has the shortest transition interval between fall and rise sections among all the waveforms; hence, it has the lowest power loss during transitions. However, it is not realizable in practice. Thus, a trapezoidal waveform is the best choice for frequencies below MHz. For higher frequencies, a sine wave can be used as a usable approximation to the trapezoids [4]. F. Gate Bias Since the gate voltage variations will drive the switch on and off, the gate bias is important in supplying this swing. For a BJT acting as the switch in class E, the transistor operates in cutoff and active region for OFF and ON interval, respectively, each for a half RF switching period. The gate bias should be the DC offset of the voltage waveform, which swings among the values needed for cutoff and active, each for half time. For a FET switch, the transistor operates in cutoff and saturation regions when the switch is in the OFF and ON stage, respectively. Similarly, the gate bias should be the DC level of the swing which makes the transistor to go cutoff or deep saturation []. _SQR P=1 Z= Ohm AMP=2.8 V TR=.2 ns TF= ns TD= ns WINDOW=DEFAULT BIASTEE ID=X1 2 RF RF & DC DC f = _FREQ 3 DCVS ID=VGS1 V=-2.36 V 1 Figure 3. Basic Circuit Schematic CURTICE ID=CF1 AFAC=1.2 NFING=1 1 2 3 IND ID=L1 L=1e6 nh ID=AMP1 CAP ID=C2 C=.31 pf DCVS ID=VGS2 V=6 V ID=AMP4 V_PROBE ID=VP1 ID=AMP2 ID=AMP3 CAP ID=C1 C=.823 pf IND ID=L2 L=13.21 nh SUBCKT ID=S2 NET="output matching cct" 1 2 P=2 Z=49.27 Ohm G. Drain Shunt Capacitance The drain shunt capacitance, Cshunt, delays the starting point of the voltage rise section while the current is at the end of its fall section during the ON to OFF transition. It ensures that at the moment when the switch is turned OFF, the voltage across the switch still remains relatively small as it was still at the end of the fall section of the drain voltage, until after the drain current has reached zero. Thus, its purpose is to shape the drain voltage and current waveforms during the ON to OFF transition to make certain that there is as little power dissipation by the switch as possible. TLSC ID=TL2 Z= Ohm EL=1.26468863 Deg F=1 GHz TLIN ID=TL1 Z= Ohm P=1 EL=69.32419674928 Deg Z=49.278-j*1.2 Ohm F=1 GHz P=2 Z= Ohm Figure 4. Circuit Schematic of Power Amplifier Figure. Major Current Flows in a PA when Switch is ON When the switch is ON, there is no current flowing through Cshunt, or the voltage drop across it. The relation among the currents going through the RF chock on the drain DC supply 1477
path ( idc ), the switch ( is ) and the rest of the load network ( io ) is 8 Vds Vs Id 2 is = idc + io (1) 6 1 4 1 2 Figure 6. Major Current Flows in a PA when Switch is OFF.3.6.8 Time (ns) - III. SIMULATION RESULTS Vds (R, V) Itime(.AMP1,1)[11] (L, ma) 1 Characteristic Curve Vs Dynamic Load Line : Freq = 2.69 GHz : Freq = 2.499 GHz 8 6 p6 p Figure 8. Vds Vs Id 4 p4 8 Vds Intrinsic Vs Id Intrinsic 2 2 p3 p8 p7 6 1 -.137 4.86 9.86 14.9 18 Voltage (V) 4 2 1 IVCurve() (ma) Chara IVDLL(V_PROBE.VP1,.AMP1)[6] (ma) IVDLL(CURTICE.CF1@ds,CURTICE.CF1@ds)[11] (ma) : Vstep = -1 V : Vstep = -.8 V p3: Vstep = -.6 V p4: Vstep = -.4 V p: Vstep = -.2 V p6: Vstep = V p7: Freq = 2.69 GHz p8: Freq = 2.474 GHz Figure 7. Characteristic Curve Vs Dynamic load line s -.2.4.6.779 Time (ns) Itime(CURTICE.CF1@ds,1)[11] (L, ma) Vtime(CURTICE.CF1@ds,1)[11] (R, V) : Freq = 2.69 GHz : Freq = 2.69 GHz Figure 9. Vds Intrinsic Vs Id Intrinsic The dynamic load line measurement is used to plot the dynamic I-V trajectory on a rectangular graph. Typically, the load line measurement is used in conjunction with the IVCurve measurement as shown above in Figure 7. 1478
96.4 PAE and Eff -1 S22 96.2 96 9.8 2.49 GHz 96.37-3 9.6-4 9.4 2.469 2.19 2.69 2.619 2.669 2.69 PAE(_1,_2) DCRF(_2) Figure 1. PAE and Eff Here, The power-added efficiency is defined as Pout Pin PAE = 1% Pdc DCRF computes the DC to RF conversion efficiency of a circuit. The DC power is computed from the total DC power of all sources in the circuit. The RF power is computed using the specified measurement element. The DC to RF efficiency is defined as -1-3 P out DCRF = 1% Pdc S11 (2) (3) - 2.469 2.19 2.69 2.619 2.669 2.69 DB( S(2,2) ) output matching cct Figure 12. S22 S22 shows the match of output as is showed in Figure 12. IV. CONCLUSIONS This paper presented the 2.469~2.69GHz AlGaN/GaN HEMT power amplifier for IEEE 82.16e WiMAX application. The simulation showed that the power amplifier can achieve 2.81dB small signal gain with a PAE of 96.37% at the center frequency point. The output power is 2.39dBm at this point with 6v supply voltage, as showed in TABLE II. With these simulation results, it could be fit for IEEE 82.16e WiMAX applications. TABLE II. POWER AMPLIFIER SIMULATION DATA Parameter Simulation Data Frequency 2.469~2.69GHz Supply Voltage 6 V Small signal gain at central frequency point 2.81dB PAE at central frequency point 96.37% Output Power at central frequency point 2.39dBm ACKNOWLEDGMENT The design is built on lots of study in the fields of power amplifier which has been done by the researchers all over the world. Thank them for the already working!. -4-2.469 2.19 2.69 2.619 2.669 2.69 DB( S(1,1) ) output matching cct Figure 11. S11 S11shows the match of input as is showed in Figure 11. REFERENCES [1] N. O. Sokal and A. D. Sokal, Class-E A New Class of High- Efficiency Tuned Single-Ended Switching Power Amplifiers, IEEE Journal of Solid-State Circuits, Vol. SC-1, No. 3 pp 168-176, June 197. [2] F. H. Raab, Class-E, Class-C, and Class-F Power Amplifiers Based upon a Finite Number of Harmonics, IEEE Transactions on Microwave Theory and Techniques, Vol. 49, No. 8 pp 1462-1468, August 21. [3] B. Razavi, RF Microelectronics, Chapter 9, Prentice,1998. [4] Peter B. Kenington, High-linearity RF Amplifier Design, Artech House. [] Alireza Shirvani, et al., A CMOS RF power amplifier with parallel amplification for efficient power control, IEEE Journal of Solid-State Circuits, Vol. 37, pp. 684-693, June 22. 1479