## 1. Introduction

The commercialization of 5G communication has extended the carrier frequency to the millimeter wave band and requires a radio frequency (RF) beamforming system consisting of array antennas and transceiver chains. Transceiver chains include millimeter-wave power amplifiers (PAs), low noise amplifiers (LNAs), phase shifters, mixers, etc.

Figure 1 shows a block diagram of 1 × 4 array antennas and transmitter chains as an example mainly used in 5G communication systems. As shown in

Table 1, the 3rd Generation Partnership Project (3GPP) has already defined equivalent isotropic radiated power (EIRP) specifications for power class 3 for handheld applications such as smartphones. Plus, it has defined higher EIRP specifications for power classes 1 and 2 that serve vehicles and fixed wireless access (FWA) systems [

1]. In the future, it is expected that more power will be required in 5G communications as in existing Long Term Evolution (LTE). Therefore, at present, millimeter-wave PAs in 5G mobile user equipment (UE) are mostly implemented as complementary metal-oxide- semiconductor (CMOS) PAs, but gallium arsenide (GaAs) PAs, which can generate higher output power due to the fact of its high breakdown voltage, is attracting attention again.

Meanwhile, 5G communication requires PAs that can operate at various millimeter-wave frequencies such as 28, 39, and 47 GHz [

2]. Because broadband millimeter-wave PAs can achieve this requirement in a single chip, they can reduce system complexity and form factor. Millimeter-wave PAs generally require a power combining technique, because single PAs cannot obtain the desired power level; however, conventional power combining circuits, such as parallel combining techniques using Wilkinson combiners, inevitably reduce power bandwidth due to the lossy power matching from very low optimum loads. Although a series power combining technique using stacked field-effect transistors (FETs) somewhat mitigates this problem, it also possesses some drawbacks such as easy oscillation and a limited number of stacked FETs at high frequencies [

3]. Recently, single-ended dual-fed distributed combining (SEDFDC) designs with zero-phase shifting (ZPS) transmission lines have been proposed for millimeter-wave CMOS PAs as shown in

Figure 2a [

4,

5]. As a similar approach, dual-fed distributed amplifiers (DAs) have been reported to enhance radiated gain and power at the leaky-wave antenna [

6,

7]. A distributed amplifier (DA) can inherently produce ultra-wideband characteristics. Moreover, ZPS transmission lines can optimize the output power of each power cell in the distributed structure [

4,

5,

6,

7]. However, References [

4,

5] still showed a P

_{sat} less than 18 dBm and a power bandwidth less than 30% in the millimeter-wave band. References [

6,

7] were implemented at low frequencies below 4 GHz. In relation to this, this work proposed a modified distributed power-combining structure as shown in

Figure 2b.

The proposed structure has the advantage of maintaining the wideband characteristics when more FETs are combined into the type of SEDFDC. The PA monolithic microwave-integrated circuit (MMIC) using the proposed structure exhibits saturated output power (P_{sat}) over 20 dBm from the full Ka- and Q-bands to the V-band (26–56 GHz).

## 2. Circuit Design

Figure 3 shows the simulated amplitude and phase differences of the inputs (in

_{1} and in

_{N} in

Figure 2a) and the outputs (out

_{1} and out

_{N} in

Figure 2a) between the first ZPS power cell and the last, according to the number of cells (

N) in the SEDFDC structure. In the simulation, 15 dBm of RF power was supplied as the input power. At

N = 2, each cell had an almost equal input/output signal amplitude and phase through the wideband, while at

N > 2, these symmetries clearly began to dissolve. This incongruity seriously reduced the power and gain bandwidth. To compensate for this input/output signal mismatch,

N/2-way power dividers/combiners were inserted into the distributed power-combining structure as shown in

Figure 2b.

Figure 4 shows the simulated results of the improved input/output amplitude and phase distribution when the 2-way and 3-way power dividers/combiners were each inserted in the case of the 2 × 4 combining and 2 × 6 combining, respectively. Both the input and output were improved by inserting the

N/2-way power dividers/combiners.

In

Figure 3a,b, the input amplitude and phase difference from 25 GHz to 60 GHz were within 3 dB and 80°, respectively, when

N = 2; but when

N = 4, they increased to 11 dB and 180°, respectively, and when N = 6, they increased to 22 dB and 180°, respectively. The output amplitude and phase difference were also within 6.5 dB and 40°, respectively, when

N = 2, as shown in

Figure 3c,d, but were 15 dB and 115°, respectively, at

N = 4 and increased to 16.5 dB and 170°, respectively, at

N = 6. If

N/2-way power dividers/combiners were added to the SEDFDC, as shown in

Figure 4a–d, the input amplitude and phase difference from 25 GHz to 60 GHz decreased within 8 dB and 120° at

N = 4, respectively. At

N = 6, they decreased to 5 dB and 30°, respectively. The output amplitude and phase difference also decreased to within 2 dB and 60°, respectively, at

N = 4 and within 4 dB and 60°, respectively, at

N = 6. Thus, this improvement increased the output power of SEDFDC in the wideband by uniformly inputting the signal to each power cell and uniformly merging the output power from each power cell.

The ZPS transmission lines are usually designed as a composite right-/left-handed (CRLH) type [

8]. The ZPS transmission line can combine the output power of each power cell inside the SEDFDC in phase to maximize the total output power [

9]. However, as

N increases, the layout complexity and DC biasing difficulty increases, and the cross-coupling induced by inductive lines, which consist of ZPS lines, becomes more severe. These factors degrade the power combining efficiency in this structure. Therefore, the ZPS lines need to be modified into a simplified structure without shunt inductances and series capacitances. As shown in

Figure 5, the modified ZPS line without a shunt inductance and a series capacitance can be equivalently replaced with a simple microstrip line. Although the simplified structure deviates from the zero-degree phase,

Figure 5 shows that it retards the phase change according to the frequencies.

The modified ZPS lines can extend the overall power bandwidth at the cost of marginally sacrificing the optimum output power at the zero-degree phase frequency.

Figure 6 compares the simulated output power of the proposed structure with that of its conventional counterpart according to the frequencies. The proposed PA exhibited a greater amount of wideband power characteristics than the conventional one without critical power degradation.

As the power divider/combiner causes extra insertion loss, the power combining efficiency is an important factor. We investigated it via a 2.5 dimensional electromagnetic (EM) simulation using Keysight’s advanced design system (ADS) momentum tool. From 26 GHz to 56 GHz, the insertion loss of the input power divider was approximately 0.9 dB to 1.8 dB and that of the output power combiner was approximately 0.7 dB to 1.2 dB. Therefore, the calculated power combining efficiency was approximately 50% to 69% from 26 GHz to 56 GHz.