23 Jan 2018

Building more efficient RF power amplifiers by terminating harmonics in the package

Abdul R Qureshi, Mustafa Acar, Sergio Pires, and Leo C N de Vreede look at how terminating harmonics in the package can improve RF power amplifier efficiency.

High data-rate mobile communication systems need RF power amplifiers (PAs) that offer high energy efficiencies, to help reduce the networks’ operating costs.

This is a challenge, since the complex modulation schemes used in the latest cellular standards have high peak-to-average-power ratios (PAR) that in turn demand high average efficiencies from the transmitters’ PAs. Many PA architectures have a ‘sweet spot’ at which they operate most efficiently, and operate at much lower efficiencies away from that spot. Achieving high average efficiencies therefore means building PA architectures that are efficient across a wide range of operating conditions.

We’ve seen some promising approaches to building such PAs, using GaN transistors in Doherty and out-phasing architectures. We think it is possible to achieve even greater efficiencies if the way in which the higher harmonics of the transmitted signal are terminated can be controlled more effectively without increasing the size or complexity of the PA board.

Our approach uses harmonically matched GaN transistors and a quasi load-insensitive (QLI) architecture to achieve the efficiencies of a Class-E amplifier in a standard RF package. The approach offers high-efficiency operation. Despite the way that Doherty and out-phasing PA architectures modulate their loads.

As a reminder, Figure 1 shows a simplified Doherty PA architecture.

A simplified Doherty PA architecture - Doherty gives high efficiency when compared to standard types

Figure 1: A simplified Doherty PA architecture

A simplified out-phasing PA architecture

Figure 2 A simplified out-phasing PA architecture.

Building more efficient PAs using QLI techniques

We use a finite-inductance implementation of a Class-E amplifier to achieve high efficiency from a simple circuit structure. Numerous operating modes arise as the relation between the load network elements and the input parameters varies as a function of the resonance factor q = 1/ω√LC, through L and C, as shown in Figure 3.

The quasi load-insensitive Class E PA

Figure 3: The quasi load-insensitive Class E PA, with its finite DC feed inductor L and lowpass LC section (L1C1) and related waveforms

At q = 1.3, the PA goes into a Class-E operating mode that offers the best efficiency over a wide range of load resistances – as required for systems that use dynamic load modulation.

In standard RF packages, size and cost constraints only allow simple matching network topologies. A series capacitor is particularly difficult to implement internally. Therefore, we derived a functionally identical transformed low-pass LC section (L1C1), as shown in the lower half of Figure 3.

Since the higher harmonics are matched inside the package, a conventional fundamental load-pull system is good enough to achieve the optimum impedance for maximum efficiency, maximum output power and back-off (e.g., 6dB). The measured data shows that maximum output power and efficiency are aligned on the real axis of the amplifier’s Smith chart. The peak efficiency is preserved while the output power decreases for an increasing real part of the load, which shows that the second-harmonic impedance required to achieve peak efficiency during load modulation is unaffected. This property is very useful to boost the average efficiency of Doherty and out-phasing PAs.

Prototype Doherty PA to improve efficiency

Figure 4: Prototype Doherty PA (a), and mixed-mode outphasing PA (b)

Applying QLI techniques to a Class E Doherty PA design

Our load-pull measurements of the power and efficiency of the packaged device suggest that it has a λ/4 internal signal rotation. This internal rotation can be taken into account in the design of the load network of the Doherty PA, so it’s not necessary to add compensation lines at the output. The fundamental load impedance required at the package leads is also high enough to allow the Doherty combiner to be connected directly without an extra matching network.

The fact that higher harmonics are terminated inside the package means that the load network for the Doherty PA can be simple, compact and that it does not need higher harmonic matching. Furthermore, the main device is biased in Class-AB mode while the peak device is biased in Class-C mode for their quiescent currents to ensure conventional Doherty operation, so that when driven hard, the device will enter Class-E like operation.

Applying QLI techniques to a dual-input, mixed-mode outphasing PA design

The mixed-mode out-phasing design is shown in Figure 4 (b). Chireix compensation has been incorporated in the two branches by adjusting their electrical length by ±Δ, instead of adding an area-consuming shunt susceptance. The value of Δ determines the required out-phasing compensation angle.

For mixed-mode out-phasing operation, a combination of phase and input-power control is used to achieve the maximum drain / PAE efficiency vs power back-off. The drive profile to achieve the best efficiency response is stored in a lookup table. This means the out-phasing PA can avoid a sharp efficiency /gain roll-off at larger out-phasing angles, and so maintain its high line-up efficiency.

QLI PA architectures in practice

We tested these two PA architectures using a dual-input measurement setup that could sweep both the input phase and amplitude of the signal. The devices were not pushed into high compression, to avoid them overheating when operating with continuous waves. This means that the peak power with modulated signals is at least 1dB higher than the static measured output power. A vector-switched general memory polynomial approach was used for linearization. An optimized digital pre-distortion strategy should give even better linearization.

The continuous-wave measurements of the Doherty PA in Figure 5 show that at 2.14GHz the peak output power reaches 46.2dBm with an efficiency of 68.79%, which is maintained above 58% at 6dB back-off. Figure 5 also shows the gain response.

Figure 5: Static measurement of Doherty PA at 25V

The Doherty PA was also tested with a single-carrier WCDMA signal having 7dB PAR. The test showed that the Doherty PA has 58.3% average efficiency and an average output power of 40.41dBm after linearization. The power spectrum of the Doherty PA after linearization is shown in Figure 6.

Dynamic measurement of a Doherty PA

Figure 6: Dynamic measurement of a Doherty PA using single-carrier WCDMA at 2.14GHz with a 7dB PAR, after digital pre-distortion

The measured efficiency versus output power response for a dual-input, mixed-mode outphasing PA using the package-integrated QLI Class-E approach discussed previously is shown in Fig. 6. Branch PA1 is biased in Class-AB while PA2 is biased in deep Class-AB, which helps to improve back-off efficiency by 3% when compared to Class-AB/Class-AB biasing. The color dots show the 2D sweep of input power and phase. The static measurement results show a peak output power of 49dBm with 77% efficiency, which is maintained above 60% beyond 6 dB back-off. The final optimum response, achieved by connecting all high efficiency points, shows more than 50% efficiency over a 9dB back-off range with a well behaved gain, as shown in Figure 7.

Static measurement of a mixed-mode outphasing PA

Figure 7: Static measurement of a mixed-mode outphasing PA versus output power at 28V showing efficiency (a), and gain (b)

The mixed-mode outphasing PA was also tested with a single-carrier WCDMA signal with a 7dB PAR. The measurement result shows that average efficiency of this mixed-mode outphasing PA is 66.6%, with an average output power of 42.68dBm after linearization. The spectrum after linearization is shown in Fig. 7.


This work shows that it is possible to build high-efficiency, load-modulation based PAs by terminating higher harmonics inside the RF package. This approach also means that the power-combining networks can be simple and compact.

Page 1 of 1

About the author

Abdul R Qureshi is from Delft University of Technology, Ampleon, and NED University of Engineering and Technology, Karachi.

Mustafa Acar and Sergio Pires are from Ampleon.

Created in 2015, Ampleon is shaped by 50 years of RF power leadership. Recently being spun-off from NXP Semiconductors, the company is set-out to exploit the full potential of data and energy transfer in RF. Ampleon has more than 1,250 employees worldwide, dedicated to creating optimal value for customers. Its innovative, yet consistent portfolio offers products and solutions for a wide range of applications, such as cellular base stations, radio/TV/broadcasting, radar, air traffic control, cooking, lighting, industrial lasers and medical.

Most popular articles in RF topics

  • Integrated PLLs and VCOs for Wireless Applications
  • Doherty Amplifiers for UHF
  • Measuring Noise Figure on RF systems
  • Demystifying PIM – Passive Intermodulation Products
  • Building more efficient RF power amplifiers by terminating harmonics in the package
  • Share this page

    Want more like this? Register for our newsletter

    The Developing Role of Electronic Component Distributors Ian Poole | Electronic Notes
    The Developing Role of Electronic Component Distributors
    The service that electronic component distributors has provided over the years has changed very significantly. Nowadays, distributors provide a very effective service, meeting the many needs of development, manufacturing and service organisations small and large.

    Radio-Electronics.com is operated and owned by Adrio Communications Ltd and edited by Ian Poole. All information is © Adrio Communications Ltd and may not be copied except for individual personal use. This includes copying material in whatever form into website pages. While every effort is made to ensure the accuracy of the information on Radio-Electronics.com, no liability is accepted for any consequences of using it. This site uses cookies. By using this site, these terms including the use of cookies are accepted. More explanation can be found in our Privacy Policy