1. Home
  2. Featured Articles
  3. A First Look at Modern Macrocell PAs and Their Practical Implementation Considerations

A First Look at Modern Macrocell PAs and Their Practical Implementation Considerations


by Pasternack

The onslaught of 5G-related technologies has led to a plethora of material covering small form factor transceivers for user equirement and mobile devices. While many of these design constraints do not necessarily apply to macro BSs, some do. This article aims to provide some more details around the design considerations for PAs for sub-6 GHz macrocells. 

Macrocell-specific Amplifier Design Considerations

The power amplifier needed in macrocells such as mobile base stations is considerably different than those required in handsets. Maintaining a relatively high efficiency remains a significant concern, not primarily due to the battery constraints found in mobile devices, but due to power and cooling limitations within the system [1]. Furthermore, base station amplifiers do not have the size constraints of handheld devices where there is finite room to incorporate performance enhancing capabilities that must necessarily occur off-the-die (e.g.: bias circuitry, digital processing, self-testing, calibration, etc.). Power requirements for macro base stations can range from the tens of watts to the hundreds of watts. This requires amplifiers that have a high breakdown voltage and gain at a broad range of frequencies. 

Basic Base Station Architecture

As shown in Figure 1, the typical base station operates with multiple transceivers (TRX) for each individual antenna element; this can be applied to micro- and macro-cells alike. For macro BSs, however, the loss due the long antenna feed from the amplifier to the antenna must also be taken into consideration. More recently, attenuation has been mitigated through the use of Remote Radio Heads (RRH) that bring the PA closer to the antenna. There is, in addition, an RF processing unit which can implement a number of architectures including superheterodyne, low-IF, or zero-IF. Macro BSs will often implement a low-IF or superheterodyne architecture after BB processing while micro BSs (e.g.: picocell, femtocells) can go directly to BB digital signal processing in a zero-IF architecture. 

Figure 1: Basic BS transceiver block diagram [2]

The Power System and The Need For a Greener BS

The power and cooling system involves an AC-DC unit, converting direct supply from the electrical grid into DC for the BS circuitry. The DC-DC converter (e.g. buck, buck-boost) supplies stable voltage levels to the unit with a high level of efficiency. The cooling system can vary from thermal pads or small form factor heatsinks in low-power handsets to active air cooling or liquid cooling systems. More often than not, the macro BS will use convective cooling due to the high powers that are required. As a matter of fact, up to 50 percent of the total operational expenditures (OPEX) for a cellular operator consits of the energy consumption of the BS [3]. Power is also of particular concern for off-grid, often diesel-powered, base stations that require regular refills/charging in areas that are hard to access. There is an increasing amount of literature covering the energy efficiency of the BSs for a more “green” cellular network. Along these same lines, the efficiency of the system can be broken down at the network level, link level, and component level where much of energy wastage can be observed when the base station continues to consume large amounts of power at low traffic loads [3]. This can be seen at the network level with the high density of power hungry macro- and micro-cells that may be underutilized at times of low capacity. At the component level, this can be seen primarily by the PA where the nonconstant envelopes found in modern modulation schemes cause the device to operate far less efficiently, wasting all the consumed input power into heat energy. Since the PA accounts for the bulk of the power consumption of a BS (Figure 2), it is of high importance to optimize the balance between linearity and efficiency. 

Figure 2: Power consumption distribution in radio access technology [3]

PAPR and The Major Drop in Efficiency at Backoff

The balance between efficiency and linearity is a prominent concern for PA designers. Advanced modulation schemes involve high peak-to-average power ratios (PAPR or PAR) where more power is locked up in these high peaks ,causing clipping to occur and sending the amplifier into compression. It seems straightforward enough to simply implement an amplifier that can withstand these peak powers without falling into a nonlinear realm that includes unfilterable IMD products (often IM3). Parameters such as the third order intercept point (TOI), or output IP3, help to ascertain the how large a signal an amplifier can process before distortion will occur.  However, it is not this straightforward. 

As PAPR varies, the mean output power of the PA is necessarily “backed off” so that the peak envelope power swings up to a compression point at a manageable IM3 level. Backing off involves reducing the input RF power in order to reduce the output RF power. The amplifier falls into its linear region which actually prevents adjacent channel interference (ACI). Unfortunately, the solution of backing off output power wildly degrades the efficiency of the amplifier as the device operates at only a fraction of what it is capable of.

Evolving Technologies Causing Additional Design Constraints

Enhanced modulation schemes and optimization technologies are, in fact, causing far more design constraints on the PAs as compared to the past. For instance, the OFDM signal specified for the LTE downlink signal can involve signal waveforms with peaks up to 7 dB above the mean power. This is why the relatively low PAPR SC-FDMA signal was chosen for the uplink. Still, there is a signal increase in PAPR from 3G to LTE, and yet again with carrier aggregation (CA). 

Most mobile carriers do not have large contiguous blocks of spectrum beyond 20 MHz but rather random chunks of spectrum over a number of frequency bands. Combining these non-contiguous blocks of spectrum for much usable bandwidth is a growing necessity for more downlink and uplink capacity. LTE Advanced (LTE-A) utilizes carrier aggregation (CA) to increases data rates by aggregating channels. This can be done within the same frequency band with or without gaps (intra-band), or, over different operating frequency bands (inter-band). Each component carrier (CC) can have a bandwidth of 1.4, 3, 5, 10, 15 or 20 MHz. A total of five channels can be aggregated for up to 100 MHz of bandwidth. Major telecom companies are already touting peak speeds beyond 1 Gbps by leveraging five- to six-channel intra-band carrier aggregation, high-order QAM (256 QAM), and MIMO [4][5]. The downside of intra-band CA is the significant increase in PAPR. As the PAPR becomes greater, either the amplifier spends more time in compression which could cause more ACI, or requires more backing off to obtain the same IM3 level at the peaks, which degrades efficiency. 

Evolving wireless standards are continually highlighting the need for TRXs to have a high PAE, linearity, and multi-band operability—three often competing parameters. Fortunately, there is a wide array of classical and modern techniques that accomplish the task of achieving adequate efficiency and linearity. These techniques include the doherty amplifier, outphasing amplifier, envelope elimination and restoration (EER), envelope tracking (ET), crest factor reduction (CFR) and digital predistortion (DPD) techniques. It is beyond the scope of this article to dive into each individual technique so a couple of popular methods will be overviewed.  

Efficiency Enhancing Technique: The Doherty Amplifier

Generally, amplifiers are most efficient at a specific power level—typically near the maximum rated power—efficiency enhancing techniques attempt to maintain this efficiency under modulated envelope conditions. This techniques include: the doherty amplifier, outphasing amplifier, EER, and ET. The doherty amplifier may very well be one of the most popular configurations. Doherty amplifier configurations have become a popular choice, particularly for a maintained efficiency deep (6 to 8dB) into the output back off region. This is accomplished by combining two amplifiers: the main (or carrier) amplifier and the auxiliary (or peak) amplifier. Each of these amplifiers are operating in different power modes such as AB for main and C for auxiliary. Both amplifiers contribute to the output power until the input drive falls below a certain threshold. In this case, the main amplifier solely provides output power the load and the efficiency of its amplifier class defines the location of the bias point. Beyond this threshold, both the main and auxiliary amplifiers contribute to the output power. This is managed by connecting a quarter wave transformer that allows the output load impedance of the main amplifier to decrease due to the power outputted by the auxiliary amplifier. This way, the saturation voltage of the carrier amplifier can be maintained while more power is delivered through the auxiliary amplifier. The maximum efficiency can therefore be maintained in the backoff region and at peak powers. 

Linearization Enhancing Technique: Digital Predistortion and Adaptive Biasing Algorithms

Due to the relatively high spectral efficiency of data transmission, there are often tight requirements for allowable spectral leakage. PAs operating close to the saturated region consistently can cause significant in-band distortion. Linearization enhancement techniques such as digital predistortion (DPD) allow the PA to operate near saturation without causing nonlinearities by “distorting” the input signal in order to minimize the distortions found at the output. This increases linearity without compromising PAE too much. The approach can be open looped or close looped. The close loop approach adapts to predistortion so that it can model the nonlinear characteristics of the amplifier in real time. The open loop relies upon an accurate prediction of the non-linear behavior of the PA.  

Choice of Substrate: Si LDMOS, GaN on SiC, and GaN on Si

The above stated techniques can allow for a power efficient PA with a low adjacent channel leakage-power ratio (ACLR). This, combined with an optimal choice of transistor substrate and topology can help significantly with the design of a macro BS PA. Solid state power amplifiers (SSPA) in macrocell applications leverage III-V semiconductors due to the high frequency performance. However, it is important to select the right substrate and transistor topology to suitably handle the high transmit powers found in macrocells. A GaAs-based device, for instance, is impeded by its low supply voltages, making it difficult to achieve adequate signal gain. Table 1 lists some of the semiconductor material properties for popular III-V substrates used in PAs. Devices with a larger electron mobility and breakdown electric field tend to score higher in the various power device figures of merit (e.g.: combined, keyes, baliga, johnson).  Parameters such as bandgap determine the inherent ability of the substrate to function at higher temperatures and switch to larger voltages. In terms of critical parameters for power devices, GaN scores fairly high across the board. 

Typically, common source, class AB Si LDMOS or GaN amplifiers are used for SSPAs in macrocell applications. LDMOS device technology has, in the past, dominated sub-6 GHz BS PAs due its cost effectiveness, ability to be produced in bulk, and satisfactory linearity and PAE in the harmonically tuned Doherty configuration. GaN HEMTs, while performing electrically superior in nearly all categories, is still pricier than LDMOS technologies. This is due to the bulk substrate technology­—Silicon Carbide (SiC). The production of SiC includes wafer diameters limited to less than 150 mm while Si is available up to 300 mm. In some analysis, the relatively poor thermal performance of Si-based GaN transistors can be overcome with the optimized layout of heat sources over a larger active device area [6]. There is currently a great deal of research and development being done both on Gan-on-SiC and GaN-on-Si technologies to potentially drive down the cost of these devices. 


Macro BSs increasingly require PAs with a high level of efficiency and linearity with technologies such as carrier aggregation and OFDM. Various linearity and efficiency enhancing techniques such as Doherty and predistortion can be applied in combination with inherently power efficient substrate technologies such as GaN. 


1. Cripps, Steve C. RF Power Amplifiers for Wireless Communications. Artech House, 2006.

2. Hossain, Ekram, et al. Green Radio Communication Networks. Cambridge University Press, 2012.

3. L. M. Correia et al., “Challenges and enabling technologies for energy aware mobile radio networks,” in IEEE Communications Magazine, vol. 48, no. 11, pp. 66-72, November 2010.

4. https://www.fiercewireless.com/wireless/t-mobile-achieves-1-3-gbps-speeds-nokia-using-laa

5. https://www.fiercewireless.com/wireless/verizon-touts-1-4-gbps-lte-speeds-unlicensed-spectrum-carrier-aggregation

6. Iucolano, Ferdinando, and Timothy Boles. “GaN-on-Si HEMTs for Wireless Base Stations.” Materials Science in Semiconductor Processing, vol. 98, 2019, pp. 100–105., doi:10.1016/j.mssp.2019.03.032.