Low cost, highly efficient, high power, microwave and RF amplifiers are required for many commercial and defense system applications. These include WLANs, wireless cable broadcast transmitters, cellular phones, and telecommunication systems, as well as advanced air-borne active phased array radar systems.

The majority of these applications are very cost driven, and, therefore, the choice of technology, design methodology, and manufacturing cycle time are the major cost contributors and should be contained.

This article provides a simple, yet accurate, design methodology for successful realization of switching mode, class-E high efficiency power amplifiers. Furthermore, a technique for modifying pHEMT large-signal model is described that yields a more accurate modeling of switching mode amplifiers. All aspects of circuit simulations, including time domain analysis, harmonic balance analysis, and large signal stability analysis were performed using an Agilent Technologies Inc. (www.Agilent.com) EEsof Advanced Design Systems' (ADS) circuit simulator¹.

The first monolithic version of a class-E switching-mode amplifier operating at 835 MHz was reported in 19941. The ideal class-E amplifier load circuit is essentially a tuned circuit as shown in Figure 1(a). The active device, in this case a pHEMT, acts as an ideal switch driven to ON and OFF conditions by the input RF signal. The ideal AC load line for the pHEMT transistor is shown in Figure 1(b).

The active device, in this case a pHEMT, acts as an ideal switch, driven to on and off conditions by the input RF signal. The ideal AC load line for the pHEMT transistor is shown in Figure 1(b).

It can be seen that the operating point moves along the Vds and Idss axes; meaning the device is either off (in the saturated region) or on (in the linear region). Therefore, the output voltage and current values at the device terminal do not exist simultaneously, and the dissipated RF energy in the device is zero, leading to 100 percent theoretical amplifier efficiency.

With the advent of active device performance and monolithic circuit technology in the last few years, significant progress has been noted towards the development of high efficiency RF and microwave components. In the case of class-E high efficiency amplifiers, the circuit designers have pushed the useful operating frequency of these types of amplifiers to ever-higher frequencies and efficiencies^2-4. Looking further ahead, we envisage by using innovative circuit topologies and device designs, it may be feasible to extend the useful operating frequency range of the switching mode amplifiers to Ku-band and possibly beyond.

Looking further ahead, we envisage that by using innovative circuit topologies and device designs, it may be feasible to extend the useful operating frequency range of the switching mode amplifiers to Ku-band and possibly beyond.

A typical design procedure would follow four main steps.

1. A simple ideal analytical design process based on a set of closed-form equations. The outcome of this step would yield the element values for the circuit topology shown in Figure 1a.
2. A time domain analysis and optimization of the above topology to assure optimum voltage and current waveforms across the switch terminals, as well as across Cds and the load terminals.
3. Development of a non-linear model for the pHEMT device and its modification for switching mode operation.
4. Harmonic balance analysis of the complete circuit including large signal stability analysis.

The detailed analysis and derivation of the following expressions are fully discussed elsewhere².

Knowing the device drain to source capacitance (C_ds) and the drain voltage (V_ds), an approximate maximum frequency (ƒ_max) for class-E operation can be obtained. Similarly, assuming a load resistance of 50Ω, one could obtain approximate values for the circuit elements of the wave-shaping network shown in Figure 1a. These expressions are:

Having obtained the starting values for the load network, it is worthwhile to perform a time domain analysis on the circuit shown in Figure 1a. The aim is to examine the current and voltage waveforms at appropriate terminals and optimize them for switching mode class-E operation.

Figure 2 shows the simulation results for the circuit after optimization of the load network.

The voltage waveform across the switch rises slowly at switch-off and falls to zero at the end of the half-cycle. It also has a zero rate of change at the end of half-cycle, ensuring a “soft” turn-on condition.

The voltage across the switch when it is off is defined by the integral of the current flowing through C_ds. The phase shift introduced by the LC circuit adjusts the point at which the current is diverted from the switch to the capacitor C_ds. Therefore, to ensure class-E operation, it is essential that the integral of capacitor current over the half-cycle is zero and that the capacitance current has dropped to zero by the end of the half-cycle. Figure 3 shows that the optimized current waveforms comply with the aforementioned criteria for class-E amplifiers.

The majority of the existing non-linear pHEMT models available in the commercial circuit simulators are not suitable for modeling class-E circuits. For accurate modeling of switching mode amplifiers, the model should have the following important properties:

• bias dependency of drain-to-source C_ds(V_ds,V_gs) and gate-to-drain C_gd (V_ds,V_gs) capacitances;
• bias dependency of input channel resistance R_i(V_ds,V_gs);
• bias dependency of output channel resistance R_ds(V_ds,V_gs); and
• a two current generator dispersion model for accurate simulation of R_ds.

The use of any non-linear models that model the dispersive behavior of the output resistance by a simple series resistor-capacitor network, connected in parallel to the standard output network, should be conducted with care. In such a case, the loading effect of the series resistor-capacitor network on the output resistance should be removed.

After careful observation of the available non-linear models, we decided on the EEsof GaAs HEMT (EEHEMT) model as a suitable choice for the non-linear simulation of class-E amplifiers. The most distinguishing features of this model for class-E are the ability to model R_ds(V_ds, V_gs) and its dispersion effect, as well as the bias dependencies of the device capacitances.

Having obtained the optimized load impedance from step two and the non-linear model coefficients from step three, the next step is to design the final amplifier topology.

At this stage, there are several design approaches one may follow, therefore, the suggested process is not unique, but has proven to be successful.

The design objective was to develop a class-E monolithic amplifier operating over 9 GHz to 11 GHz using a 0.3 um × 600 um pHEMT device. The design process starts by generating the large signal S-parameters of the device over the desired RF input drive and frequency band while the device stability is assured by conventional circuit techniques. The next stage is to design the input-matching network for the amplifier by providing a conjugate match to the large signal S11 over the frequency band of interest. Figure 4 shows the final schematic circuit of the amplifier.

Figure 5 depicts the simulated voltage and current waveforms at the pHEMT output terminals of the amplifier.

The waveforms confirm the switching mode behavior of the pHEMT, a condition that is necessary for class-E operation of the amplifier.

The completed monolithic amplifier chip is shown in Figure 6. A primitive layout was used in this first iteration to assure the accuracy of the complex load.

Figure 7 shows the measured output power, PAE, and gain versus input power at 10.6 GHz.

A maximum efficiency of 63 percent, and an output power of 24 dBm is obtained at P_in = 14 dBm. Figure 8 shows the measured amplifier output power for different values of RF input drive levels (0 dBm to 18 dBm) over 7 GHz to 12 GHz.

As it can be seen, a broadband output power is obtained indicating the broadband capability of class-E operation. Figure 9 depicts the measured amplifier PAE for different RF input drive levels.

PAE of greater than 40 percent over 9 GHz to 11 GHz, peaking to 63 percent at around 10.6 GHz is obtained.

The first successful design and fabrication of an X-band monolithic high efficiency class-E amplifier has been shown. In addition, a four-step design methodology is described. Based on this design approach, a monolithic amplifier that employs a 0.3 um × 600 um pHEMT device has been fabricated. The amplifier's measured performance shows a peak PAE of 63 percent at 10.6 GHz and a constant output power of greater than 24 dBm together with a gain of 10 dB over 9 GHz to 11 GHz.

1. T. Sowlati, et al. “Low Voltage, High Efficiency Class E GaAs Power Amplifiers for Mobile Communications,” IEEE GaAs IC Symposium Digest, pp.171-174, 1994.

2. T.B. Mader and Z.B. Popovic, “The Transmission-Line High-Efficiency Class-E Amplifier,” 1999 IEEE M&GWL, vol. 5, No. 9, pp. 290-292, September 1995.

3. P. Watson, et al., “Ultra-High Efficiency Operation Based On An Alternative Class-E Mode,” IEEE GaAs IC Symposium Digest, pp. 53-56., 2000.

4. R. Tayrani, “A Broadband Monolithic S-band Class-E Power Amplifier, “IEEE RFIC Symposium Digest, pp. 255-258, June 2002.

Reza Tayrani received his B.Sc., M.Sc. and Ph.D. degrees in electrical engineering from Kent University, Canterbury, England, in 1974, 1977 and 1985, respectively.

He is currently an engineering fellow at Raytheon Microwave Center, Space and Airborne Systems engaged in the research and development of GaAs and SiGe MMICs and their related devices. Tayrani has designed and developed many MMICs based on MESFETs, HEMTs, pHEMTs, and HBTs for microwave and millimeter-wave applications.

His current areas of interests are high efficiency switching mode monolithic power amplifiers, advanced SiGe MMICs, broadband sampling circuits, and miniature switched filters. Tayrani has published over 46 technical papers and holds six patents. He can be reached at rtayrani@raytheon.com