1. Home
  2. Featured
  3. Articles
  4. How GaN Transistors Relieve Some of the PA Design Burden in Systems with Non-Constant Envelope Signals

How GaN Transistors Relieve Some of the PA Design Burden in Systems with Non-Constant Envelope Signals

75
0

by Fairview Microwave

Modern communications systems rely on non-constant envelope techniques with increasingly complex waveforms. This introduces a plethora of design challenges and is particularly burdensome on the power amplifier (PA) due to the high peak-to-average power ratios (PAPR) used. This causes a balancing act between linearity and efficiency. The use of the GaN substrate in PAs is no longer a nascent concept, but rather a rapidly developed field where the advantages of this semiconductor are exploited to offer smaller, power dense, and efficient PAs for the transmitter signal chain. However, utilizing GaN for systems with complex modulation schemes has unique considerations due to its relatively lower linear performance compared to Si-based PAs. Understanding the benefits of implementing GaN with several common design enhancements with linearization techniques can be helpful in understanding PA design considerations in modern communications systems.

The General Benefits of the GaN Substrate

Relative to other popular III-V semiconductor technologies such as Gallium Arsenide (GaAs) and Indium Phosphide (InP), as well as Silicon-based transistors such as Laterally Diffused Metal Oxide (LDMOS), GaN presents a higher band gap, breakdown electric field, power density, saturation velocity and thermal conductivity. The wide bandgap allows for better operation and higher temperatures with an increased ability to switch to large voltages­—thus the large breakdown field. A high saturation velocity translates the ability of the material to handle larger current densities. The higher voltage and current density lends itself to higher the relatively large power density. This leads to the final significant parameter of thermal conductivity; the GaN substrate can dissipate heat better than the bulk of substrate materials used for RF devices. This directly correlates to an increased Mean Time Between Failures (MTBF) due to the inherent improvements in thermal management. 

The improvement in efficiency is particularly relevant for battery powered systems where the power consumption of the PA is particularly relevant—the more gain an amplifier is able to provide with the same bias voltage, the more efficiently the unit is able to operate. The power density and general efficiency of a GaN PA allows for smaller sized devices for comparable performance. The characteristically smaller size permits the use of a higher impedance; in other words, impedance matching to 50 ohms is relatively straightforward (low transformation ratio) and there are generally less components required to accomplish a matching circuit. For instance, the large parasitic capacitors often used in LDMOS PAs create a low impedance at the die interface of 1 ohm or lower—increasing the transformation ratio and the number of stages to transform the impedance to 50 ohms in a wideband design [3]. The general increase in efficiency allows more design flexibility for systems that push the PA to its limits in terms of bandwidth, efficiency, linearity, and thermal management. This is particularly the case in systems with dynamic supplies, or non-constant envelopes that exhibit high peak-to-average power ratios (PAPR).  

Figure 1: Frequency versus power chart for various power amplifier semiconductor technologies

Balancing Efficiency and Linearity

Non-linear parameters of power amplifiers are defined by the signal distortion that occurs due to gain compression and saturation of the output power. Beyond a certain point, the non-linear voltage- and current-limiting phenomena found in the FET devices generate not only output power saturation, but also signal distortions that ultimately degrade signal quality below acceptable levels. Often, this boils down to the problem of maximizing power-added efficiency (PAE) by riding the average power of the transmitted signal close enough to the nonlinear zone of the PA. In essence, a PA with a high efficiency harnesses a strict DC power budget to generate an adequate RF output power. Parameters such as Psat, P1dB, and IP3 are all basic measurements that grant insight into the operation of a PA. 

The Challenges of a High PAPR 

However, in modern communication systems, a PA not only has to optimize efficiency, but also must fulfill constraints on linearity and spectral efficiency. Models with single-tone excitations are not viable with the modulated signals that are often used instead that result in a populated spectrum over the frequency band as opposed to the harmonic generation at 2ƒ and 3ƒ. Adjacent Channel Power Ratio (ACPR) accounts for this spectral regrowth due to out-of-band radiation in band-limited systems. On the other hand, in-band distortions will ultimately lead to a higher bit error rate (BER), often measured through error vector magnitude (EVM). Modern modulation formats utilize signals with time-varying envelopes that involve waveforms with a high Peak-to-Average Power Ratio (PAPR) in the range of 6 to 12 dB. A PA can clip these signals due to a limited dynamic range or nonlinearities causing signal degradation. Higher output backoff (10-20 dB backoff range), or operation well below the saturation point, is therefore required to ensure the PA is operating linearly so as to avoid sending the device into compression, ultimately failing ACPR requirements and increasing the BER. The problem of signal degradation due to the high PAPR is dealt with one of two ways: 

  • Reducing the PAPR 
  • Linearizing the power amplifier

Techniques for PAPR reduction include selected mapping (LSM), partial transmit sequence (PTS), and signal clipping and filtering. Power amplifier linearization can fall into one of three categories: 

  • Altering the input signal (digital predistortion, feedback linearization, linear amplification using nonlinear components )
  • Altering the output signal (feedforward linearization)
  • Dynamically altering amplifier characteristics (envelope tracking, envelope elimination and restoration)

In these types of systems, the average efficiency becomes a more pertinent operating parameter than instantaneous efficiency (based upon the peak envelope power (PEP)). The average efficiency, however, depends upon the probability density function (PDF)—a measure of the relative amount of time spent at each amplitude for an input signal with a non-constant envelope. The drain bias voltage varies in proportion to the amplitude of the input signal and PA efficiency is often improved with ET by optimizing the drain bias voltage to the maximum PDF, where the probability is the highest [1]. A PA that maintains a high efficiency over a wide drain bias range is preferred for systems with ET. GaN FETs tend to show a small variation of output capacitance with respect to drain voltage, thereby providing a high efficiency over the full range of output powers [2].

Figure 2: A normalized input power versus normalized gain chart comparing GaN HEMT and LDMOS power amplifier technology [7]

Linearizing GaN Amplifiers in 5G—Benefits Compared to LDMOS

The PAPR has been steadily increasing with each generation of cellular transmissions. In 3G, W-CDMA PAPR sat at approximately 3.5 dB. This increased to 8.5 dB with the OFDM modulation scheme for LTE. Some 5G installations can bring the PAPR up to 10.5 dB with cyclic-prefix orthogonal frequency-division multiplexing (CP-OFDM) modulation up to 256 quadrature amplitude modulation (QAM). 

In the recent past, Si-based LDMOS amplifiers have been used for base station PAs because of their high breakdown voltages and power density. A GaN PA, however, has a higher band gap at 3.4 eV compared to 1.1 eV for LDMOS transistors as well as nearly five times the power density at 10 W/mm compared to 2 W/mm for LDMOS. Still, the traditionally employed LDMOS PAs have an inherently higher linearity than GaN FETs, which is needed for systems with non-constant envelopes. In other words, a GaN PA would require more power back-off from the saturation power to achieve the same linearity in a LDMOS PA, nearly eliminating the efficiency benefits of leveraging the GaN substrate. 

However, linearization techniques reduce this power back-off necessary, thereby allowing an amplifier to operate closer to the more efficient saturation mode. Typically, DPD techniques are leveraged due to the ability to cost-effectively implement sophisticated digital signal processing (DSP) while consuming little power. This, combined with the popular Doherty amplifier configuration that allows for increased efficiency at output power back-off, generally mitigates the hurdle of linearity when implementing the GaN substrate. Moreover, it has been illustrated that GaN-based PAs are more straightforward to linearize relative to LDMOS-based PAs in both DPD and ET techniques [4-6]. 

In GaN PAs, the gain compression often begins at large power back-off and compresses slowly while LDMOS illustrates a characteristic “hump” in gain before output power saturation. In the simplest, memoryless models of DPD expressed as power series polynomials, a third-order polynomial is enough to adequately model a GaN PA while a fourth-order is required for LDMOS due to the hump. This ultimately leads to less computing resources to implement DPD [4]. For adaptive bias techniques such as ET and EER, the characteristic of strong gain variation with drain supply voltage of GaN transistors can be exploited to linearize the amplifier—a characteristic that LDMOS amplifiers invariably lack. In essence, as the amplifier reaches peak powers, any compressive behavior can be linearized by the wide gain variation when the drain bias is adequately increased [5]. Considering the numerous design benefits of implementing GaN PAs for communications systems, there is a general trend of GaN PAs replacing more legacy LDMOS amplifiers in basestation transmission despite the hurdle of GaN amplifier linearity. 

 GaN Amplifiers for Military COFDM UAV/UGV Data Links

The growing trend seen in unmanned aerial systems (UAS) is the introduction of advanced sensors/communications systems with multi-band requirements. Typically, UAS data links rely on line-of-sight (LoS) communications within the L-, S-, and C-bands. However, beyond LoS (BLoS) data links can occur at lower VHF/UHF bands that have the benefit of a long range due to the utilization of longer wavelengths. Nonlinear modulation schemes can also be found in military/aerospace applications; for instance, coded orthogonal frequency-division multiplexing (COFDM) data links are often used  for surveillance, airborne data and telemetry as well as unmanned aerial vehicles (UAV) and unmanned ground vehicles (UGV). 

Figure 3: Various GaN power amplifiers in assembled packages

The COFDM modulation scheme provides a more reliable and faster data link alternative to the traditional VHF/UHF radio transmissions. Some OFDM-based standards such as IEEE 802.11 a/g/n/ac wireless LANs, Digital Audio Broadcasting (DAB), and Digital Video BroadcastingTerrestrial (DVB-T) are found not only in 4G wireless systems but also in public safety and tactical communications. The inherent benefit of OFDM systems is the increase in reliable transmissions over noisy wireless channels. The COFDM scheme in particular, overcomes multipath and doppler effects that cause errors in communication channels with forward error correction (FEC) techniques as well as time/frequency interleaving. Once again, this comes with the setback of a high PAPR. As stated earlier, the GaN substrate can effectively replace silicon-based power transistors in military applications with wideband operation, allowing for higher data rates as well as better operation at higher temperatures and voltages. The highly efficient amplifiers can achieve a space savings relative to Si-based counterparts and the added benefit of high power density lends itself to aerospace applications where size, weight, and power (SWaP) are all critical metrics that are necessarily optimized. 

Conclusion

Regardless of the PAPR of the modulation scheme, GaN PAs can be implemented with linearization techniques as well as efficiency enhancing topologies (i.e. Doherty) to successfully meet the increasingly stringent system linearity, efficiency, and bandwidth requirements. Moreover, PA size is generally decreased, allowing for implementation in space-constrained aerospace military applications such as COFDM UAV/UGV data links. 

References

1. GaN Transistors for Efficient Power Conversion Alex Lidow (pages 301-302).

2. D. F. Kimball et al., “High-Efficiency Envelope-Tracking W-CDMA Base-Station Amplifier Using GaN HFETs,” in IEEE Transactions on Microwave Theory and Techniques, vol. 54, no. 11, pp. 3848-3856, Nov. 2006, doi: 10.1109/TMTT.2006.884685.

3. Analog Circuit Design: Smart Data Converters, Filters on Chip, Multimode… by Arthur H.M. van Roemund (pages 262-263).

4. Gracia Sáez, R.; Medrano Marqués, N. LDMOS versus GaN RF Power Amplifier Comparison Based on the Computing Complexity Needed to Linearize the Output. Electronics 2019, 8, 1260.

5. Z. Yusoff et al., “The Benefit of GaN Characteristics over LDMOS for Linearity Improvement Using Drain Modulation in Power Amplifier System,” 2011 Workshop on Integrated Nonlinear Microwave and Millimetre-Wave Circuits, Vienna, 2011, pp. 1-4, doi: 10.1109/INMMIC.2011.5773334.

6. M. Olavsbråten, D. Gecan, M. R. Duffy, G. Lasser and Z. Popovic, “Efficiency Enhancement and Linearization of GaN PAs Using Reduced Bandwidth Supply Modulation,” 2017 47th European Microwave Conference (EuMC), Nuremberg, 2017, pp. 456-459, doi: 10.23919/EuMC.2017.8230888.

7. Linearity of GaN HEMT RF Power Amplifiers – A Circuit Perspective 

(75)

print

LEAVE YOUR COMMENT