Technical Feature Designing A Broadband, Highly Efficient, GaN RF Power Amplifier J. Brunning and R. Rayit SARAS Technology, Leeds, U.K. A design approach for a broadband, linear, effjcient output back-off mode RF power amplifjer (RFPA) emphasizes the importance of minimizing design uncertainties. Using this approach, excellent agreement between modeled and measured performance is achieved with a fjrst-pass design. D emand for linear RFPAs cover- ciency greater than 54 percent over its oper- ing the frequency range from ating bandwidth. In back-off mode, the RFPA 1.5 to 2.8 GHz is driving new achieves an uncorrected linearity of 30 dBc design methods for broadband, and drain effjciency of 34 percent or higher linear and highly effjcient RFPAs operating when driven with a 2.5 MHz, 9.5 dB peak-to- in output back-off mode. Improving effj- average power ratio (PAPR) COFDM signal ciency in PAs has long been a challenge in the 2.0 to 2.5 GHz band. for designers, in part due to poor control of harmonic load impedances. The diffj- RFPA DESIGN FLOW culty measuring waveforms at microwave Device Selection frequencies makes it hard to determine if The fjrst step begins with a thorough optimum waveshaping has been achieved. device/technology selection process to de- Broadband design adds a challenge when termine the best candidate device to meet a harmonic of a lower operating frequency a specifjc set of criteria prior to the time- lies in the operating band. These inherent consuming tasks of load- and source-pull diffjculties can be compounded by impre- and network synthesis. Several candidates cise design techniques, leading to multiple are acceptable on the basis of claimed fre- time-consuming and expensive iterations. quency and power. In addition to the more In this article, a design fmow is described common characteristics such as V ds , gain, that uses NI AWR Design Environment, operating frequency and power rating, other specifjcally Microwave Offjce circuit design parameters such as C ds , C gs and transforma- software, as well as a measurement tech- tion ratio are considered. nique for determining the input and output impedances of the matching networks, prior Optimal Load Impedance Extraction to RFPA turn on. Several approaches to the Once a device is selected and a nonlinear problems inherent in PA design are present- model obtained, optimal source and load ed with the aim of minimizing uncertainty impedances are determined. The required and achieving fjrst-pass success. load impedances to achieve maximum The effectiveness of this approach is dem- power, effjciency and gain—or an accept- onstrated using a commercially available able trade-off between these performance discrete 10 W GaN on SiC, packaged, high metrics—are frequency dependent and vary electron mobility transistor fabricated with substantially over the operating bandwidth a 0.25 µm process (Qorvo’s T2G6000528) of a broadband design. and a 20 mil RO4350B printed circuit board. To determine the correct load imped- The fabricated RFPA achieves a peak power ance, a combination of load-pull plotting greater than 40 dBm and a peak drain effj- 72 MWJOURNAL.COM JUNE 2018
Technical Feature RFPA and conse- 0 0 1.0 quent diffjculties in Insertion Loss (dB) Return Loss (dB) achieving optimal 8 0.2 2.0 harmonic termi- 0.5 0.4 16 nations 1 without 0.6 24 using transmis- Zopteff 32 sion zeros in the 0.8 5.0 network. 2 Load- 1.0 40 0.2 1400 1900 2400 2900 pull at the second Frequency (MHz) harmonic is also (a) performed, with a 0.5 1.0 2.0 region of high ef- Load Network Efficiency Zoptpwr 1.0 fjciency identifjed 1 Power Gain that can be con- Transducer Gain –0.2 trolled in the net- 0.9 –5.0 work synthesis. Network Synthesis –0.5 N a r r o w b a n d 0.8 1400 1900 2400 2800 –2.0 RFPAs have the Frequency (MHz) advantage of little –1.0 (b) variation of the op- Fig. 2 Distributed load network Fig. 1 Fundamental frequency load-pull analysis showing timal load imped- loss and match (a) and transducer and power (red) and effjciency (blue) contours over the operating ance over their op- operational power gain vs. frequency (b). bandwidth. erating bandwidth, making the task of eral, EM simulation is seen as an im- network design less complex. This is at the fundamental and harmonic portant step in reducing uncertainty not to say that a low fractional band- frequencies and waveform engi- in the design fmow. width match is always trivial. Indeed, neering (circuit design techniques One design technique is to rep- an investigation of source and load based on shaping the transistor resent the conjugate of the optimal impedances will reveal that for very voltage and current waveforms) are impedance as that of a two-terminal high performance, the network fun- performed in Microwave Offjce. The generator (port 1), after which the damental impedance must often use of waveform engineering relies matching network design can be be precisely controlled to a single on having access to the intrinsic de- viewed as a problem of reducing the gamma point, with signifjcant sub- vice nodes across the current gener- mismatch loss that exists between optimal performance penalties if ator of the device plane, rather than this complex-valued load and a 50 Ω the network locus misses its target at the package reference plane. As- termination over the amplifjer’s oper- load impedance. Precise control of suming the nonlinear device model ating bandwidth. This mismatch can, harmonic termination impedances provides these nodes, a waveform however, be evaluated at the 50 Ω for F and F -1 amplifjer classes in- engineering approach enables the side (port 2) of the network, as shown creases the complexity of the task visual observation of voltage and in Figure 2a . As a passive network, beyond what is required for an aver- current swing, clipping and ampli- the output matching circuit has an op- age PA design. fjer class of operation. erating power gain less than 1, equal In the case of a broadband am- For this example, a load-pull sim- to its effjciency determined only by in- plifjer, particularly one with high ulation is run at V ds = + 28 V and I dq ternal dissipative loss. The necessarily performance specifjcations, the = 90 mA across the operating band, smaller transducer gain is the product network is required to control its and the impedances for optimal of this effjciency with the effect of loss impedance variation over a far power and effjciency are extracted, due to refmection at the input. These larger fractional bandwidth. After with the mid-band results shown in quantities are shown as percentage defjning optimal impedances and Figure 1 . A target load region based effjciencies in Figure 2b . The effjcien- target areas, the load network is on the overlap between P max ‐ 1 dB cy of the load network is calculated to developed using a simplifjed real- and drain effjciency max (eff max ) be 96.6 percent at 2800 MHz, close frequency technique (SRFT) 3 to ‐ 5 percent is defjned. Clearly, the to the value calculated from return design the ideal lumped-element larger this target area is, the easier loss at the same frequency. For com- network and convert it to a distrib- the matching problem becomes. In parison, the operational power gain, uted stepped-impedance format, 4 this case P max occurs on a tightly- which considers purely ohmic loss in before performing electromagnetic packed clockwise rotating locus over the network, is calculated to have an (EM) simulation. In this example, EM the operating bandwidth, which is effjciency of 97.7 percent. Although simulation results agree closely with helpful in the case of a broadband this does not directly include refmec- model predictions; however, for less amplifjer. Load-pull is performed tion losses, its value does depend on conventional matching topologies, at the fundamental frequency due the termination impedances, as these this might not be the case. In gen- to the broadband nature of the affect the distribution of current and 74 MWJOURNAL.COM JUNE 2018
Recommend
More recommend