A Single-Board Integrated Millimeter-Wave Asymmetric Full-Digital Beamforming Array for B5G/6G Applications

Engineering ›› 2024, Vol. 41 ›› Issue (10) : 38 -53. DOI: 10.1016/j.eng.2024.04.013

Research

Article

Qingqing Lin ^a
, Jun Xu ^a^,^c^,^*
, Kai Chen ^a
, Long Wang ^a
, Wei Li ^a
, Zhiqiang Yu ^a
, Guangqi Yang ^a
, Jianyi Zhou ^a^,^b
, Zhe Chen ^a
, Jixin Chen ^a^,^b
, Xiaowei Zhu ^a^,^b
, Wei Hong ^a^,^b^,^c^,^*

Author information +

History +

Received	Accepted	Published
2023-09-25	2024-04-08	2024-10-01
Issue Date	Revised Date
2024-10-01	2024-02-19

PDF (7049KB)

Abstract

In this article, a single-board integrated millimeter-wave (mm-Wave) asymmetric full-digital beamforming (AFDBF) array is developed for beyond-fifth-generation (B5G) and sixth-generation (6G) communications. The proposed integrated array effectively addresses the challenge of arranging a large number of ports in a full-digital array by designing vertical connections in a three-dimensional space and successfully integrating full-digital transmitting (Tx) and receiving (Rx) arrays independently in a single board. Unlike the traditional symmetric array, the proposed asymmetric array is composed of an 8 × 8 Tx array arranged in a square shape and an 8 + 8 Rx array arranged in an L shape. The center-to-center distance between two adjacent elements is 0.54λ₀ for both the Tx and Rx arrays, where λ₀ is the free-space wavelength at 27 GHz. The proposed AFDBF array possesses a more compact structure and lower system hardware cost and power consumption compared with conventional brick-type full-digital arrays. In addition, the energy efficiency of the proposed AFDBF array outperforms that of a hybrid beamforming array. The measurement results indicate that the operating frequency band of the proposed array is 24.25-29.50 GHz. An eight-element linear array within the Tx array can achieve a scanning angle ranging from −47° to +47° in both the azimuth and the elevation planes, and the measured scanning range of each eight-element Rx array is -45° to +45°. The measured maximum effective isotropic radiated power (EIRP) of the eight-element Tx array is 43.2 dBm at 28.0 GHz (considering the saturation point). Furthermore, the measured error vector magnitude (EVM) is less than 3% when 64-quadrature amplitude modulation (QAM) waveforms are used.

Graphical abstract

Keywords

Full-digital beamforming array / Asymmetric structure / Single-board integrated / Beyond fifth-generation and sixth-generation / Millimeter-wave communication / Complex modulation / Printed circuit board / Vertical connection

Cite this article

Download citation ▾

Qingqing Lin, Jun Xu, Kai Chen, Long Wang, Wei Li, Zhiqiang Yu, Guangqi Yang, Jianyi Zhou, Zhe Chen, Jixin Chen, Xiaowei Zhu, Wei Hong. A Single-Board Integrated Millimeter-Wave Asymmetric Full-Digital Beamforming Array for B5G/6G Applications. Engineering, 2024, 41(10): 38-53 DOI:10.1016/j.eng.2024.04.013

登录浏览全文

4963

注册一个新账户忘记密码

1. Introduction

Millimeter-wave (mm-Wave) technology provides the advantages of abundantly available spectrum resources and excellent reliability, which perfectly correspond to the ultra-high transmission rate and large capacity requirements of beyond-fifth-generation (B5G) and the upcoming sixth-generation (6G) wireless communication systems [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12]. Although mm-Wave technology offers multiple advantages, there are also several significant challenges that must be addressed; in particular, mm-Wave technology is employed at a higher frequency, which results in smaller device sizes, higher design density, and layout challenges [8], [13], [14], [15], [16]. Due to the desired high integration in a mm-Wave array, single-board designs are always preferred over those consisting of several pieces [9], [10], [15], [16], [17], [18], [19]. In general, mm-Wave active beamforming technology can be divided into three categories: analog beamforming (ABF) arrays [9], [10], hybrid beamforming (HYBF) arrays [20], [21], [22], and full-digital beamforming (FDBF) arrays [8], [15], [16], [23], [24], [25], [26], [27]. Due to the benefits of their low cost and simple implementation, the ABF and HYBF arrays are widely employed. However, their insufficient phase shifter accuracy and beam count restrict the further application of these two techniques in next-generation mobile communications [1], [8], [23]. In Ref. [8], a 4 × 16 multiple-input multiple-output (MIMO) FDBF array for 5G mm-Wave communication is reported. In order to maximize flexibility and performance, each radiating element of the array is connected to an independent radio frequency (RF) channel, and the phase shift and amplitude management functions are completed in the digital baseband. However, the FDBF array suffers from significant disadvantages including a complex baseband, high power consumption, and high cost. To reduce manufacturing costs and power consumption, novel FDBF array architectures with nonreciprocal transmitting (Tx) and receiving (Rx) beam patterns have been developed in Refs. [23], [24]. The FDBF array demonstrated in Ref. [23] operates at sub-6G and is constrained by limited spectrum resources. Each radiating element is independently connected to the Tx or Rx channel, and Tx-mode or Rx-mode switching is completed by a duplexer. In the Tx mode, all radiating elements are activated and operate as a Tx array, while a portion of the radiating elements are chosen to form a Rx array in the Rx mode. In Refs. [28], [29], the performance of an asymmetric full-digital beamforming (AFDBF) array is simulated and analyzed, with the reported AFDBF array achieving better performance compared with a conventional symmetric full-digital beamforming (SFDBF) array for spectral efficiency, with reduced system cost and power consumption. Both the FDBF arrays in Refs. [8], [23] are arranged as sub-array blocks in the x- and y-axes and stacked in the z-axis to establish a brick structure system, whose excessive size and weight dissatisfy the compactness and miniaturized requirements of B5G/6G communication systems. Numerous studies on beamforming arrays using a single board are also being conducted in an effort to reduce complexity and size; for example, wideband or dual-band mm-Wave phased arrays with a single-board structure have been reported in Refs. [9], [10]. Furthermore, a single-board digital beamforming array for radar application was developed in Ref. [15]; the array connects with the sampling board and digital baseband board through an intermediate frequency (IF) connector after all IF transmission lines are arranged on one side of the printed circuit board (PCB). The implementation of such an array becomes challenging when the array element number grows dramatically; an increase in the layout area and continuous bending of the RF transmission lines are required to complete the board design.

In this paper, a single-board integrated AFDBF array operating at the 24.25-29.50 GHz frequencies band is designed, implemented, and measured, and a further performance evaluation based on Ref. [23] is conducted. Possible application scenarios of the proposed AFDBF array at the base station (BS) are depicted in Fig. 1. In the downlink communication, the size of the Tx array in the AFDBF array is larger than that in the SFDBF array at the BS side, contributing to a higher antenna array gain and achieving a longer transmission distance. In the uplink communication, the antenna array gain is reduced because the size of the Rx array in the AFDBF array is smaller than that in the SFDBF array at the BS side. However, the beamwidth of the Rx array in the AFDBF array has been increased, thereby providing larger space coverage areas. At the same time, the size of the Tx array in the entire AFDBF array at the user terminal (UT) side is larger than that in the SFDBF array, which can compensate for the gain loss of the Rx array in the AFDBF array at the BS side. Nevertheless, several challenges remain in the implementation of the mm-Wave AFDBF array. Firstly, the devices and antenna have smaller sizes at mm-Wave frequencies than in the sub-6 GHz band; this renders the entire array more sensitive to the layout and size rationalization, making the design more challenging. Secondly, in contrast to phased-array architectures [9], [10], each antenna element of the proposed AFDBF array is connected to an independent RF channel. The total number of ports in the array dramatically grows as the number of array elements increases. Designing all ports reasonably within a constrained PCB space is very challenging. Finally, it is difficult to process the interference of various signals at different frequencies.

2. Performance analysis of the AFDBF array

The proposed single-board integrated AFDBF array integrates a Tx array with 64 elements (8 × 8) arranged in a square shape and an Rx array with 16 elements (8 + 8) arranged in an L shape, as shown in Fig. 2(a). Dummy elements are placed around both the Tx and Rx arrays to minimize the edge effect of the radiating elements. Fig. 2(b) depicts the PCB stack-up structure of the proposed AFDBF array. The full-digital multi-channel chips are located on the bottom layer of the PCB, while the radiating elements are positioned on the top layer. The chips and radiating elements are connected together by a vertical metallized via, and the IF ports of these chips are all connected to the IF connectors though a grounded coplanar waveguide (GCPW) transmission line. An IF adapter is employed to connect the array to the digital processing module more conveniently and effectively. In contrast to the design in Ref. [23], the Tx and Rx arrays of the proposed AFDBF array are designed to be separated; they are respectively composed of several four-channel Tx chips and four-channel Rx chips with the corresponding radiating elements. As a result, it is possible to achieve the most beneficial isolation between the Tx and Rx arrays while also effectively avoiding the delay caused by the RF switch used to choose either the Tx or Rx mode. The proposed AFDBF array is a single-board architecture, which is preferable to a brick structure in terms of size.

The proposed AFDBF array architecture is a configuration that requires independent transceiver chain and analog-to-digital converter (ADC)/digital-to-analog converter (DAC) links for each antenna element. Beamforming is accomplished in the digital domain. Compared with the phased-array architectures in Refs. [9], [10], the AFDBF array architecture is characterized by a flexible multi-beam ability, high beamforming precision, high precoding freedom, and fast beam steering speed. Only one beam is generated by a phased array that uses phase shifters. Poor accuracy, excessive cost, and significant losses are the drawbacks of analog phase shifters, compared with the tuning phase in the digital domain, and these have been bottlenecks in enhancing the performance of mm-Wave arrays [8], [10], [23], [28]. In summary, the AFDBF-based array has three major advantages, as follows: ① Higher capacity can be realized by an AFDBF-based array through the simultaneous transmission of several data streams. The AFDBF-based system supports a greater maximum number of spatial multiplexing streams compared with a phased-array-based system. ② An extremely high amplitude and phase resolution can be achieved by the AFDBF array, due to the utilization of digital precoding in the digital domain. ③ Furthermore, for multi-carrier signals such as orthogonal frequency division multiplexing (OFDM) signals, performance enhancement can be realized by the AFDBF array, which supports individual beam precoding at each resource block or subcarrier when the signal bandwidth is wide. The wideband wireless channel is characterized by frequency selectivity. There are different propagation characteristics in signals at different frequency bands. Independent magnitudes and phases to different subcarriers of the band can be assigned by the AFDBF-based system, while only the same phase can be assigned to all subcarriers by an analog phased-array system. The superior beamforming control enables the AFDBF-based system to achieve optimal transmission performance over wider frequency ranges.

2.1. Fundamental properties of the AFDBF and SFDBF array relationship

The proposed AFDBF array can be classified into two types (Fig. 2(c)) according to the number of antenna elements [23], [24], [28], [29]. In the downlink communication, the type-I array satisfies

N_{TBa}^{Asy} = a N_{TBa}^{Sym} (a > 1)

and

N_{RU}^{Asy} = (1 / a) N_{RU}^{Sym} (a > 1)

, where

N_{TBa}^{Asy}

and

N_{TBa}^{Sym}

are the number of Tx antenna elements in the AFDBF and the traditional SFDBF array at the BS side, respectively.

N_{RU}^{Asy}

and

N_{RU}^{Sym}

are the Rx antenna element in the AFDBF and the SFDBF array at the UT side, respectively. a is a factor of antenna element number. The gain of the type-I Tx array goes up as the number of antenna elements increases, making it more suitable for long-distance communications. In addition, the beamwidth of the Rx array becomes broader as the number of antenna elements decreases, facilitating alignment with the targets and enhancing the beam tracking and discovery performance in communications. Consequently, the type-I array can serve more users simultaneously. In contrast to the type-I array, the type-II array has the relationship

N_{TBa}^{Asy} < N_{TBa}^{Sym}

, which is ideal for high-throughput requirements in the uplink. In this paper, the type-I array is selected and analyzed, due to the fact that an ADC consumes more cost than a DAC when the sampling rates are equal. The gain relationship of the Tx and Rx arrays within the entire AFDBF and SFDBF arrays can be written as Eqs. (1), (2) [30]:

(1)

G_{t, i}^{ABa} = G_{t, i}^{SBa} + 10 \log (a) (dB)

(2)

G_{r, j}^{Au} = G_{r, j}^{Su} - 10 \log (a) (dB)

where

G_{t, i}^{ABa}

and

G_{t, i}^{SBa}

are the gain of the Tx array from the AFDBF and SFDBF arrays at the BS side, respectively; and

G_{r, j}^{Au}

and

G_{r, j}^{Su}

are the gain of the Rx array from the AFDBF and SFDBF arrays at the UT side, respectively. i and j represents the ith beam of the Tx array at the BS side and the jth beam of the Rx array at the UT side, respectively. When a communication link is established between the ith beam of the Tx array at the BS side and the jth beam of the Rx array at the UT side, the received signal power

P_{r, i}^{Au}

and

P_{r, i}^{Su}

from the UT side in the downlink communication can be written as Eqs. (3), (4) [30], [31], [32]:

(3)

P_{r, i}^{Au} = P_{t, i}^{ABa} + G_{t, i}^{ABa} + G_{r, j}^{Au} - {PL}_{ij}

(4)

P_{r, i}^{Su} = P_{t, i}^{SBa} + G_{t, i}^{SBa} + G_{r, j}^{Su} - {PL}_{ij}

where

P_{t, i}^{ABa}

and

P_{t, i}^{SBa}

are the total transmitted power of the AFDBF and SFDBF arrays at the BS side, respectively;

G_{t, i}^{ABa}

and

G_{t, i}^{SBa}

are the gains of the Tx arrays within the entire AFDBF and SFDBF arrays at the BS side, respectively;

G_{r, j}^{Au}

and

G_{r, j}^{Su}

are the gains of the Rx arrays from the AFDBF and SFDBF arrays at the UT side, respectively; and

{PL}_{ij}

is the transmission path loss. This demonstrates that, when the same signal (

P_{t, i}^{Au} = P_{t, i}^{Su}

) is delivered, the received signal power at the UT side is equal, whether in the AFDBF array or the SFDBF array. The received power at the BS side in the uplink communication reaches the same conclusion (i.e.,

P_{r, i}^{ABa} = P_{r, i}^{SBa}

, where

P_{r, i}^{ABa}

and

P_{r, i}^{SBa}

represents the received power of the AFDBF and SFDBF arrays at the BS side in the uplink communication, respectively), indicating that the uplink and downlink communications are equivalent in these two different array structures.

2.2. Comparison of the fundamental performance between the AFDBF and HYBF arrays

In the downlink communication, the number of radiating elements in the AFDBF Tx array is

N_{TBa}^{Asy}

, which is equal to the number of RF channels that can support K beams (

N_{TBa}^{Asy} \geq K

) at the BS side. Each beam generated by the AFDBF Tx array has a gain of

N_{TBa}^{Asy} G_{d, s}

, where

G_{d, s}

(

s = 1, 2, . . ., N_{TBa}^{Asy}

) is the gain of each radiating element, and here, we assumed that each radiating element has an equivalent gain. The HYBF array is typically divided into

L

(

L = K

) subarrays. Each subarray has

N_{TBa}^{Hy}

(

N_{TBa}^{Hy} \geq K

) radiating elements, where

N_{TBa}^{Hy}

represents the radiating element number of each HYBF Tx subarray at the BS side. The HYBF array can also generate K (each subarray generates one beam) independent beams simultaneously. The antenna aperture efficiency of the AFDBF array is enhanced because the FDBF array employs all of the radiating elements, whereas the HYBF array only uses part of them. In the HYBF Tx array,

G_{m, a}

(

a = 1, 2, . . ., N_{TBa}^{Hy}

) represents the gain of each radiating element (

G_{d, s} = G_{m, a}

), and the number of channels

L N_{TBa}^{Hy}

is equal to the number of radiating elements. The gain of the HYBF Tx subarray can be expressed as

N_{TBa}^{Hy} G_{m, a}

for each individual beam. Here, we assume that each Tx channel of the two types of Tx arrays in the downlink communication has the same transmitted power as

P_{ch}

(W). The maximum effective isotropic radiated power (EIRP) of each beam generated by the AFDBF Tx array (

{EIRP}_{Afd}

) and the HYBF Tx array (

{EIRP}_{hy}

) can be expressed as Eqs. (5), (6), as has been indicated in Refs. [30], [31], [32]:

(5)

\begin{matrix} {EIRP}_{Afd} = P_{tot} + G_{array} + G_{ant} - 10 \log K (dBm) \\ = P_{ch} {(N_{TBa}^{Asy})}^{2} G_{d, s} / K (W) \end{matrix}

(6)

{EIRP}_{hy} = P_{ch} {(N_{TBa}^{Hy})}^{2} G_{m, a} (W)

where

P_{tot}

is the total transmitted power of the Tx array,

P_{ch}

is the channel output power of the Tx chip; and

G_{array}

and

G_{ant}

are the gains of the antenna array and the single radiating element, respectively. When the EIRPs of the AFDBF Tx array and the HYBF Tx array are equal, the number of radiating elements between the two types of arrays satisfies

N_{TBa}^{Asy} = \sqrt{K} N_{TBa}^{Hy}

. This demonstrates that the AFDBF Tx array can achieve the same EIRP as the HYBF Tx array with fewer antenna elements.

In the uplink communication, we denote the transmitted power density as

S_{ut} = EIRP / (4 π R^{2})

from the UT side [32], where R is the transmission distance between the BS and UT. The received signal (

P_{r}^{Afd}

) in the AFDBF Rx array at the BS side with

N_{RBa}^{Asy}

radiating elements (

N_{RBa}^{Asy} = N_{TBa}^{Asy} / Q > K (Q > 1)

, where Q is a factor of antenna element number between Tx array and Rx array in the AFDBF array,

N_{RBa}^{Asy}

represents the radiating element number of AFDBF Rx array) can be expressed as follows [32]:

(7)

P_{r}^{Afd} = S_{ut} A_{r} = \frac{EIRP λ^{2} N_{RBa}^{Asy} G_{dr}}{{(4 π R)}^{2}}

where λ is the free-space wavelength,

A_{r}

is the effective area of the Rx antenna, and the gains of each Rx channel in the AFDBF and HYBF Rx array are all equal to

G_{dr}

. In the HYBF array, the Rx array has

L N_{RBa}^{Hy}

antenna elements, satisfying

N_{TBa}^{Hy} = N_{RBa}^{Hy}

(where

N_{RBa}^{Hy}

is the radiating element number of each HYBF Rx subarray), and is also divided into L subarrays. When the same signal is transmitted, the received signal power (

P_{r}^{hy}

) in the HYBF Rx subarray can be expressed as follows:

(8)

P_{r}^{hy} = \frac{EIRP λ^{2} N_{RBa}^{Hy} G_{dr}}{{(4 π R)}^{2}}

To achieve an equivalent received power by the AFDBF and HYBF Rx arrays, the number of antenna elements must satisfy

N_{RBa}^{Asy} = N_{RBa}^{Hy}

. The total number of RF channels in the AFDBF array and the HYBF array should satisfy the following relationship:

(9)

\begin{matrix} d = (N_{TBa}^{Asy} + N_{RBa}^{Asy}) - (N_{TBa}^{Hy} L + N_{RBa}^{Hy} L) \\ = (\sqrt{L} + 1 - 2 L) N_{TBa}^{Hy} \end{matrix}

where d is the relationship of total radiating number between the AFDBF array and the HYBF array. It is obvious that, when

L \geq 2

, the AFDBF array requires fewer channels than the HYBF array to achieve the same link budget.

Consider a point-to-point MIIMO system equipped with M Tx and N Rx antenna elements.

S (t) = {[S_{1}, . . ., S_{t}]}^{T}

are the data streams generated by the Tx array at BS side, transmitted to

t

users (

t \leq M

). The channel capacity

C

is derived as shown in Eq. (10), as has already been indicated in Refs. [28], [33].

(10)

C = B_{0} E_{H} [\log_{2} \det (I_{N} + \frac{P_{tot}}{M B_{0} N_{0}} {HH}^{H})]

where

E_{H} (\cdot)

denotes the expectation operator,

I_{N}

is an

N \times N

identity matrix,

H \in C^{N \times M}

is denoted as the

N \times M

channel matrix, H represents the conjugate transpose,

C

is complex domain,

B_{0}

is the system operation bandwidth, and

N_{0}

is the thermal noise density. In Ref. [28], the energy efficiency performance of the AFDBF is analyzed, showing that it outperforms the SFDBF and HYBF arrays. In contrast to the conventional SFDBF, the AFDBF array has an asymmetrical number of Tx and Rx radiating elements, allowing for the realization of a wide Rx beam and a high-gain pencil Tx beam. Moreover, the AFDBF array maintains the benefits of the conventional SFDBF array while drastically reducing the power consumption and system hardware costs.

3. Implementation of a single-board integrated mm-Wave AFDBF array

3.1. mm-Wave full-digital transmit and receiver chips

Figs. 3(a) and (b) present the block diagram and Figs. 3(c) and (d) show the package of the mm-Wave Tx and Rx full-digital chips fabricated in a 0.13 µm SiGe BiCMOS with a wafer level chip scale packaging (WLCSP) and 400 µm pitch balls from State Key Laboratory of Millimeter Waves, Southeast University, China. The independent design of the Tx and Rx chips offers developers a great deal of flexibility when sizing the transceiver arrays. Table 1 summarizes the performance metrics of the Tx and Rx chips.

3.2. Linearly polarization cavity-backed antenna design

A 45° linearly polarized antenna with a back-cavity structure is presented in Fig. 4(a); it is fed by a strip-line printed in layer 10 through an H-shaped slot in layer 12 to broaden the impedance bandwidth. The feeding stripline is connected to the pins of the chip through a metal quasi-coaxial vertical transition; similar designs are provided in Refs. [34], [35]. The entire antenna element is composed of metal layers from layers 9 to 16, as shown in Fig 2(b), and layers 1-8 are used to locate other lines of the transceiver, such as the power and control lines. An H-shaped slot similar to the design in Ref. [36] is etched in the metal ground in layer 12 as a coupling slot. A back-cavity structure similar to the design in Ref. [37] is used to effectively reduce the coupling between adjacent elements in the array. The structure is constructed by metallized vias connecting layers 9-12. Since the height of the metallized vias is close to a quarter wavelength at the center frequency of the antenna, four metallized vias that are placed at the center of four edges of the antenna function as four monopoles, helping to broaden the beamwidth [38], [39], [40], [41].

Fig. 4(b) shows the simulated reflection coefficient (|S₁₁|) and the realized gain of the proposed 45° linearly polarized cavity-backed antenna. A −10 dB impedance bandwidth of 28% covering 23.0-29.5 GHz is realized, which can meet the requirement of B5G/6G mm-Wave communications. The realized gain of the proposed antenna element is stable at about 5 dBi over the operating frequency band, with a fluctuation within 0.3 dB. Figs. 4(c)-(e) illustrate the simulated co- and cross-polarization radiation patterns in the elevation and azimuth planes (E- and H-planes) of the proposed 45° linearly polarized antenna at 24.0, 27.0, and 29.5 GHz, respectively. The normalized cross-polarizations are below -30 dB in both the E- and H-planes. The simulated 3 dB beamwidth of the E- and H-plane radiation patterns are maintained at 120° and 125° at different frequencies, which enables a wide-angle beam-scanning capacity for 5G mm-Wave BS applications, thereby increasing the space coverage.

In the proposed AFDBF array, each antenna element is connected to its corresponding RF and IF channels independently. Therefore, the total number of ports increases dramatically as the array scale becomes larger. Designing a single-board integrated AFDBF array within a constrained space becomes very challenging, as a huge number of ports should be arranged in a reasonable manner. Obviously, increasing the distance between array elements can benefit the design. However, it is crucial to consider not only the challenges associated with the PCB layout and fabrication but also the performance of the antenna array, such as mutual coupling and the beam scanning range, before determining a reasonable element spacing. In addition, the experimental costs and resource consumption are significant considerations when choosing the size of an array. A larger array size would result in excessive cost and resource consumption. Considering an additive white Gaussian noise (AWGN) channel, in order to meet the requirements of mm-Wave communication with 64 quadrature amplitude modulation (QAM) signals, the signal-to-interference-plus-noise ratio (SINR) should be greater than 22 dB. For the downlink, the sensitivity (S_min) of the UTs can be characterized by the following equation [42]:

(11)

S_{\min} = 10 \log (k T_{0}) + 10 \log B_{n} + {NF}_{0} + SINR

where 10log(kT₀) = −174 dBm·Hz⁻¹ is used, k is Boltzmann constant, T₀ is standard noise temperature, B_n represents the receiver bandwidth (in Hz), and NF₀ is the overall noise figure of the receiver (in dB). Each antenna element in the proposed array offers a gain of approximately 5 dBi, and each channel of the Tx chip provides a linear transmission power of 9 dBm (10 dB back-off from P_1dB), where P_1dB is 1 dB compression point. Assuming that the effective service range coverage of a BS cell is a radius of 200 m (a typical value) in mm-Wave cell communication, based on the Friis transmission formula (i.e., Eq. (12)), the path loss corresponding to a 200 m distance is about –107.1 dB at 27 GHz.

(12)

P_{r, i}^{AU} = P_{t, i}^{ABa} + G_{t, i}^{ABa} + G_{r, j}^{AU} + 20 \log (\frac{λ}{4 π R}) (dBm)

A power margin for the system is required in mm-Wave communication applications. Fig. 5 illustrates the link budget versus the number of antenna elements, indicating that a 64-element Tx array at the BS side can realize a power margin of approximately 8 dB. In order to control the experimental costs to a reasonable value and facilitate communication with various users within a 3D space, an 8 × 8-sized Tx array is selected.

The AFDBF array with a brick structure operates at sub-6G, as demonstrated previously [23], and a 16-element Rx array has been selected to verify the experimental principles. A single-board integrated AFDBF array is extended in Ref. [23]; thus, a 16-element Rx array is selected in this paper. Moreover, in order to realize a better Rx array performance, the Rx array is arranged in an L shape. According to Ref. [43], an L-shaped array demonstrates a greater potential with 37% accuracy, compared with other arrays with a simplistic structure. The maximum likelihood estimation of the wave directions can be realized in a computationally efficient way due to the properties of two uniform linear arrays in an L-shaped array. Thus, two eight-element uniform linear Rx arrays are arranged in the x- and y-directions, respectively, forming an L-shaped Rx array. The proposed asymmetrical architecture full-digital phased array provides a promising solution to reduce both the system cost and the hardware complexity. Such an architecture also offers a solution to lessen the computational burden that comes from processing a huge amount of uplink data at the cost of a moderate decrease in uplink transmission capacity, which is a tradeoff against the reduction of both expensive RF channels and high-rate ADCs (which are usually more expensive compared with DACs). Thus, we simulated a 64-element array, as shown in Fig. 6(a), with element spacing of 5.0, 5.5, and 6.0 mm (0.45λ₀, 0.50λ₀, and 0.54λ₀), respectively, where λ₀ is the free-space wavelength at 27 GHz. The simulated results of mutual coupling between adjacent elements and the beam-scanning performance are presented in Fig. 6, Fig. 7, respectively. The simulated results shown in Fig. 6 indicate that the mutual coupling between neighboring elements is up to −9.1, −11.8, and −13.5 dB when the element spacing is 5.0, 5.5, and 6.0 mm at 27.0 GHz, respectively. The mutual coupling between the adjacent elements decreases by approximately 1.7 dB when the element spacing increases from 5.5 to 6.0 mm. Another consideration that affects the selection of the element spacing is the beam-scanning range. As depicted in Fig. 7, maximum beam-scanning angles of ±55°, ±52°, and ±51° with a scanning loss of 3 dB are realized in both the xoz and yoz planes when the element spacing is 5.0, 5.5, and 6.0 mm, respectively. Moreover, ±69°, ±60°, and ±58° with a scanning loss of 5 dB are achieved in both the xoz and yoz planes with an element spacing of 5.0, 5.5, and 6.0 mm, respectively. The simulated results confirm that no vestige of the grating lobe falls into the visible region during scanning when the element spacing is set to be 5.0, 5.5, or 6.0 mm. The increased spacing between the array elements allows for greater flexibility in designing densely packed PCBs. Overall, an element spacing of 6 mm was chosen, after carefully considering the PCB design complexity, the isolation, and the beam-scanning range.

3.3. Vertical interconnections and key blocks

Different from phased arrays, such as those in Refs. [9], [10], the total number of ports dramatically increases as the array scale becomes larger because there is a one-to-one correspondence among the IF channels, RF channels, and radiating elements in the AFDBF array. When designing a single-board integrated AFDBF array in a constrained area, it is a significant challenge to arrange the massive number of ports in a reasonable manner. The proposed AFDBF array employs a vertical interconnection approach to realize the effective integration of both the Tx and Rx arrays. Fig. 8 depicts the vertical interconnection configuration between the chip and the antenna in the single-board integrated AFDBF array.

In the Tx array, Tx chips are uniformly mounted on one surface of the PCB, while antennas are evenly arranged on the other surface. First, the GCPW transmission line in layer 1 is utilized to connect the chips to the IF connectors. Then the IF connector is evenly positioned between two adjacent chips with the prescribed space, achieving a uniform IF port distribution and providing equal IF signal transmission paths. The IF signal of all channels in an array with an arbitrary scale can be efficiently and conveniently connected. Moreover, the PCB stack-up structure is significantly simplified, and the layout size is dramatically diminished. The IF connector and transmission line are conveniently utilized to establish a vertical connection between the mm-Wave RF board and the digital baseband board. Furthermore, the various functional boards are individually analyzed, making the array simple to assemble and disassemble for inspection. Serial-peripheral-interface (SPI) pins are also included in the IF connector. Each Tx chip can be controlled independently due to the corresponding connections between the SPI pins on the Tx chip and IF connector. Fig. 9 shows the measured transmission performance of the IF connectors. With reference to the figure, the reflection coefficients are below -15 dB regardless of whether the transmission line is bent or not, indicating their strong reliability and stability. A single Tx chip contains four RF channels, allowing for the simultaneous excitation of four radiating elements. The centers of the Tx chip and the rectangular area formed by the four adjacent radiating elements overlap in the vertical direction. In Refs. [15], [34], a quasi-coaxial vertical transition configuration consisting of a central metallized via and several surrounded ones (Fig. 8) is employed to transmit RF signals through different layers to the stripline connected with the radiating element while minimizing the insertion losses (ILs). However, there would be a conflict between the via from layers 1 to 10 and the another one from layers 9 to 16 during the manufacturing process. To address this issue, a metallized via that passes through all the layers—that is, from layers 1 to 16—is used to serve as the feeding via; under this circumstance, a section of the feeding via (i.e., from layers 11 to 16) remains as an extra stub. Then, a back-drilling technique is adopted to eliminated this stub from layer 11 to layer 16 in order to reduce the impedance mismatch. A number of ports and transmission lines that transmit different frequency signals are densely integrated in a single PCB and the question of how to reduce the isolation between them is another significant issue. Numerous metallized vias are placed to surround each transmission line in order to shield and reduce the signal interference, similar to the solution adopted in Ref. [19]. The simulation results of the isolation between different ports (as marked in Fig. 10(a)) reveal that it is below −40 dB at the operating frequency, as shown in Fig. 10(b), which indicates that excellent transmission performance has been achieved and the system requirements are satisfied.

Unlike the Tx array, which is arranged in a square shape with 8 × 8 radiating elements, the Rx array is in the form of two uniformly linear arrays. Fig. 8 presents a partial diagram of the entire Rx array with 1 × 4 radiating elements connecting to the same Rx chip; two IF connectors are shown in the figure. The same type of IF connectors are used in this Rx array as in the Tx array. However, the RF and the IF transmission lines may intersect in the same layout when the array scale is enlarged. To minimize the ILs and ensure phase consistency in all the different channel paths in the proposed single board, all the RF transmission lines are implemented to employ bends with the same length. The connection between the RF port and the feeding line of the radiating element is realized using GCPW transmission lines in layer 1 of the PCB and a quasi-coaxial via from layers 1 to 10. Bending and vertical transformation have less impact on the performance of the IF transmission lines than on the performance of the RF transmission lines operating at higher frequencies. The IF ports of the Rx chip are originally connected to GCPW transmission lines located at layer 1. To prevent intersection between the IF and the RF transmission lines, a well-shield transmission line—that is, a substrate integrated coaxial line (SICL)—is used and buried in the PCB to replace the partial path of the IF signal.

3.4. Local oscillator (LO) distribution networks

Two LO distribution networks are used to deliver LO signals to each channel of the Tx and Rx arrays, respectively. According to Fig. 8, the LO distribution network for the Tx array is a 1-to-16 power divider based on the SICL transmission line. However, it can be clearly observed from Fig. 8 that, if the LO distribution network of the Tx array is laid in layer 1, it will be obstructed by other high-density transmission lines. Similar to the RF connection presented in Fig. 8, a quasi-coaxial vertical transition composed of metallized vias is used to connect the LO distribution networks laid in layers 1 and 6. The input port of the LO distribution networks from the Tx array is connected to an End Launch connector (Southwest Microwave, Inc., Arizona, USA), and 16 outputs are uniformly distributed and connected to the corresponding Tx chips using GCPW transmission lines at layer 1 and the feeding via from layers 1 to 6. In order to enhance the isolation and prevent unnecessary modes on the parallel plate between layers 5 and 8 caused by an open-structure LO distribution network, a number of metallized vias are positioned around the transmission lines, forming a well-shielded SICL LO distribution network. In contrast to the Tx array, the LO distribution network of the Rx array is conveniently integrated into layer 1, as depicted in Fig. 8. A 1-to-4 GCPW power divider in layer 1 is formed by three cascaded Wilkinson power dividers. The input port of this LO distribution network is connected to the End Launch connector, and the four output ports are directly connected to the corresponding Rx chips, respectively. The LO signals are delivered to each Rx chip with equal phase and amplitude.

Similar to the design in Ref. [15], the vertical interconnection approach used in the proposed AFDBF array enables the complete integration of the Tx and Rx arrays in a single board. Conversely, one IF connector is placed in the middle space between two adjacent chips, providing four IF channels with equal transmission paths. Owing to the aforementioned methods, the challenges raised by the large number of ports in the FDBF array are effectively reduced. In contrast to the designs in Refs. [8], [23] with a brick construction, the proposed single-board AFDBF array is more suitable for B5G/6G mm-Wave applications due to its low cost, low weight, and compact size. Furthermore, the Tx and Rx arrays are separated in this design without any switch, thereby achieving high isolation and full-time operation.

4. AFDBF array prototype and measurement

The PCB of the designed AFDBF array was manufactured using standard processing. Subsequently, all Rx and Tx chips, as well as some other components, were accurately placed on the PCB, based on surface-mount technology (SMT). The fabricated AFDBF array was further measured to confirm the performance and validate the design. Figs. 11(a) and (b) present photographs of the manufactured mm-Wave AFDBF array. Because of the vertical connection design between the proposed AFDBF array and the IF adapter board using the IF connectors and transmission lines, it is extremely convenient to manage the large number of IF ports. Due to the limitation of the current experimental conditions (i.e., there are not enough high-speed ADC/DAC sampling boards available for numerous channels), it is not possible to directly evaluate the performance of the entire AFDBF array using common methods. The proposed array is measured by employing a beamforming network with eight independent phase shifters and a 1-to-8 power divider. A tiny fan is utilized to dissipate heat from the array and maintain a temperature balance during the measurement.

4.1. Calibration

Calibration is an essential procedure before the measurement of a phased array is carried out, due to the unavoidable channel differences that arise from the limitations in manufacturing precision, assembly techniques, unequal transmission line lengths, flaws in different independent devices, and other factors [8], [23]. A Keysight N5225B PNA vector network analyzer (VNA) and a standard gain horn antenna were implemented to perform the far-field calibration of the proposed mm-Wave AFDBF array in an anechoic chamber environment with a range of

D = 1.5 > 2 D_{\max}^{2} / λ_{0}

(unit: mm), where

D_{\max}

is the largest dimension of either antenna, as shown in Fig. 11(c). The measured root-mean-square (RMS) gain and the phase variation are within 5.85° and 0.75 dB at 27.00 GHz after calibration, as exhibited in Figs. 11(d) and (e).

4.2. Measurement of the radiation patterns

The simulated and measured xoz- and yoz-plane scanning radiation patterns of the Tx and Rx arrays at 27 GHz under uniform excitations are exhibited in Figs. 12(a)-(d). As mentioned above, a uniform eight-element linear array within the proposed Tx array is measured each time, due to the lack of enough high-speed DAC sampling boards. Fig. 12(a) shows the simulated and measured xoz-plane radiation patterns of the eight-element uniform Tx array (the element distance is 0.54λ₀ at 27.00 GHz) located in the x-axis direction, and a half-power beamwidth (HPBW) of 11.50° at 27.00 GHz can be observed. This Tx array achieves a scanning range of ±47° in the xoz plane with a scanning loss of 3-4 dB, and the sidelobe levels (SLLs) are lower than −10 dB at the maximum scanning angle. The scanning radiation patterns of the eight-element uniform Tx array located in the y-axis direction were also measured at 27 GHz, and the measured results agreed well with the simulated ones, as shown in Fig. 12(b). In the yoz plane, the measured HPBW is 11.2° at 27.0 GHz, and the measured scanning range is −47.0° to 47.0°, with SLLs lower than −10.0 dB at 27.0 GHz.

Similarly, the simulated and measured xoz-plane scanning radiation patterns of the eight-element uniform Rx array (the element distance is 0.54λ₀ at 27.00 GHz) located in the x direction are plotted in Fig. 12(c). Good agreement between the simulation and measurement is achieved. As the measured results indicate, an HPBW of 11.1° and a scanning range from −45.0° to +45.0° with a scanning loss of 3.0-4.0 dB and SLLs lower than −10.0 dB at the largest scanning angle are achieved at 27.0 GHz. Fig. 12(d) presents the simulated and measured yoz-plane scanning radiation patterns of the eight-element uniform Rx array positioned in the y-axis direction. In the yoz plane, a 3.0 dB beamwidth of 11.3° and a scanning range of ±45.0° with SLLs less than −10.0 dB at the maximum scanning angle are observed at 27.0 GHz.

4.3. Over-the-air (OTA) performance measurement

An OTA performance measurement of the proposed AFDBF array was carried out. The EIRP performance of the proposed AFDBF array can be obtained by comparing the results of the Tx array and a standard gain horn antenna. Figs. 13(a) and (b) show block diagrams of the EIRP performance measurement setups of the proposed array. In the measurement of the proposed array and an array using a standard horn as the Tx antenna, the received powers are

P_{Act}^{r}

and

P_{h}^{r}

, respectively.

(13)

P_{Act}^{r} = P_{sig} - L_{li}^{t} - L_{P} + G_{Act} - SPL + G_{hr} - L_{li}^{r}

(14)

P_{h}^{r} = P_{sig} - L_{li}^{t} - L_{P}^{h} + G_{ht} - SPL + G_{hr} - L_{li}^{r}

where

P_{sig}

is the input power level set at 1 dB compression point, and SPL is the space path loss (SPL).

G_{Act}

is the total gains of the proposed active beamforming array.

G_{ht}

and

G_{hr}

(

G_{ht} = G_{hr}

) are the gains of the standard gain horn antenna used in the transmitter and the receiver, respectively.

L_{li}^{t}

and

L_{li}^{r}

represent the insertion losses for each cable.

L_{P}^{h}

denoted the insertion loss for each channel of beamforming test board and

L_{P}

is the insertion loss for total beamforming test board, which can be expressed as follow:

(15)

L_{P} = 10 \log (\sum_{1}^{V} U_{v}) - P_{in} (dB)

(16)

L_{P}^{h} = 10 \log (\sum_{1}^{V} U_{v} / V) - P_{in} (dB)

where

U_{v}

(

v = (1, 2, . . ., V)

is the vth channel output power level of the beamforming test board in Watts, and V is the total channels number of the beamforming test board that is equal to the number of radiation elements. P_in is the input power of the beamforming test board. The total gain of the active beamforming array can be denoted as follows:

(17)

G_{Act} = (P_{Act}^{r} - P_{h}^{r}) + G_{ht} - (L_{P}^{h} - L_{P})

Then, the measured EIRP can be calculated by

(18)

\begin{matrix} {EIRP}_{P_{1 dB}} = P_{sig} - L_{li}^{t} - L_{P} + G_{Act} \\ = P_{sig} - L_{li}^{t} - L_{P}^{h} + G_{ht} + P_{Act}^{r} - P_{h}^{r} \end{matrix}

The measured EIRPs versus different frequencies are presented in Fig. 13(c). A maximum EIPR of 40.8 dBm is achieved at 28.0 GHz by the eight-element Tx array with uniform excitations when P_1dB is considered. When the saturation point (P_sat) is considered, the maximum value of EIRP is changed to 43.2 dBm at 28.0 GHz. The fluctuation of the EIRP is less than 3.5 dB within the 24.0-29.5 GHz frequency band. Due to the limitations of the current experimental conditions (i.e., there are not enough high-speed ADC/DAC sampling boards available for so many channels), an eight-element linear array within the proposed AFDBF array was measured and evaluated. The anticipated EIRP level for the entire 64-element Tx array could be maintained at about 61.2 dBm in an ideal condition. It is highly likely to achieve an EIRP level higher than 60 dBm using a 64-element Tx array when the transmission loss is taken into consideration. Moreover, Fig. 13(d) depicts the EIRP at P_1dB versus different scanning angles in the xoz plane at 27.0 GHz, and a maximum gain reduction of 3.5 dB is achieved within the scanning angle range from −47.0° to +47.0°. Overall, the proposed array maintains excellent performance within the scanning angle range, and the difference may result from the increased additional antenna loss, mismatch, or self-heating.

The error vector magnitude (EVM) and constellations measurement setup of the Tx array are shown in Fig. 14(a). The modulated signal with a center frequency of 3.5 GHz is initially generated by a Keysight M8190A arbitrary waveform generator (AWG) and amplified by the driver amplifier; it is then upconverted to a center frequency of 27.0 GHz by the proposed Tx array. To ensure high linearity, the output power of the driver amplifier is maintained at about 20.0 dBm at 3.5 GHz, which is equivalent to a 5.0 dB back-off from the P_1dB. In addition, the EIRP of the Tx array is adjusted by modifying the variable attenuators after the driver amplifier, ensuring that the nonlinearity of the array is the primary factor affecting the EVM performance, rather than the amplifier in the Tx chain. A Ka-band standard gain horn antenna is placed at a distance of 1.5 m, with an SPL of about 65.0 dB at 27.0 GHz. The radiated signals received by the horn are finally demodulated by employing a Keysight 50 GHz real-time digital storage oscilloscope (DSOZ504A) and Keysight Vector Signal Analysis software (VSA-89600); then, the equalization is done [10]. The EVM performance is measured using a modulation signal under uniform illumination at 27 GHz and is delivered by the RMS value of the constellation, as shown in Fig. 14(b). The eight-element Tx array achieved a maximum data rate of 6.00 gigabits per second (Gbps) with 2.45% EVM using 64-QAM waveforms, as shown in Fig. 14(c). It is possible to obtain a higher data rate by using high-order modulations (e.g., 128- and 256-QAM) with a large modulation bandwidth. Nevertheless, it should be noted that the EVM performance would be constrained by the AWG signal-to-noise ratio (SNR) and the LO phase noise.

The measured EVM performance of the Tx array versus the EIRP levels using a 200 Mbaud 64-QAM modulation signal at 27 GHz with a root-raised cosine pulse-shaping filter that has a roll-off factor and a peak-to-average power ratio (PAPR) of 7.7 dB is plotted in Fig. 14(d). The measured EVM performance is better than 5% within the EIRP value ranging from 20 to 33 dBm in the entire operating frequency band, which is only about 7 dB back-off from P_1dB. Due to the nonlinearity of the array, the EVM performance is degraded in the high-EIRP-level regions. Measurement of the EVM performance employing a 400 Mbaud 64-QAM waveform at 27 GHz was also conducted and is shown in Fig. 14(d); the measured EVM is better than 5% within the operating band. The EVM performance of the proposed Tx array corresponding to different scanning angles in the xoz plane at 25, 27, and 29 GHz is plotted in Fig. 14(e), respectively, to further exhibit the performance of the proposed Tx array in the AFDBF array. The measured EVM is obtained using a 400 Mbaud 64-QAM modulation signal. The EVM performance of the Tx array at all frequencies is better than 5% over the ±47° scanning range with a 10 dB back-off from the P_1dB. Since the link is restricted by the SNR and the antenna gain drops with the scanning angle, the EVM performance of the Tx array degrades at large scanning angles, as predicted.

Similarly, the EVM performance of the eight-element Rx array was measured using a complex modulation signal. Fig. 15(a) shows a block diagram of the link measurement of the proposed Rx array with a standard gain horn at a distance of 1.5 m. The AWG generates a complex modulation signal that is radiated by a standard gain horn at the operating frequency. The signal is received by the proposed Rx array and demodulated, and the constellation and EVM performance are extracted using an oscilloscope DSOZ504A. A data rate of 6 Gbps with 2.92% EVM using a 64-QAM modulation signal is realized by the Rx array at 27 GHz, as indicated in Fig. 15(b). Fig. 15(c) shows the EVM performance of the proposed Rx array measured at the operating frequency bands (i.e., 25, 27, and 29 GHz) for different scanning angles in the xoz plane, respectively, with a 400 Mbaud 64-QAM modulation signal. The EVM performance of the Rx array at all frequencies is better than 4% over the ±45° scanning range. The measurement setup and the SNR associated with the noise floor of the Rx system are the fundamental factors constraining the EVM.

5. Performance comparison with state-of-the-art mm-Wave communication systems

Tables 2 [8], [9], [10], [14], [15], [21], [23] offers a performance summary of the proposed single-board integrated mm-Wave AFDBF array and some other state-of-the-art mm-Wave communication systems. Compared with the brick-type arrays proposed in Refs. [8], [14], [23], the proposed single-board integrated mm-Wave AFDBF array exhibits the merits of a compact size, low weight, low system cost, and simple system design complexity. Unlike the ABF array reported in Refs. [9], [10] and the HYBF array reported in Ref. [21], the proposed AFDBF array uses the same FDBF architecture used in Refs. [8], [15], [23], which can provide a high beamforming accuracy, flexible beam control, fast beam steering speed, and the highest degree of precoding freedom. It is clear that the proposed AFDBF array can provide easy base implementation and has a low cost, and low power consumption, in contrast to the conventional DBF array structures proposed in Refs. [8], [15], [23]. Moreover, in Ref. [9], peak EIRPs of 39 and 42 dBm are achieved for beams in different operation frequency bands by a 32-element phased array. In Ref. [10], an EIRP of 54.8 dBm at P_1dB is achieved by a 64-element phased array. The extracted EIRP of the entire 64-element array is around 43 dBm in Ref. [23]. In the proposed AFDBF array, a measured maximum EIPR of 40.8 dBm is achieved by the eight-element Tx array with uniform excitations when P_1dB is considered. Nevertheless, when P_sat is considered, the maximum value of EIRP can be increased to 43.2 dBm.

According to Eq. (19), the EIRP of an array increases with the number of antenna elements [10], [30]:

(19)

EIRP = P_{ch} + G_{ant} + 20 \log M

A higher EIRP supports longer transmission distances. According to Eq. (19), the anticipated EIRP level for the entire 64-element Tx array may be maintained at about 61.2 dBm in an ideal condition. It is very possible to achieve an EIRP level higher than 60 dBm by means of a 64-element Tx array when the transmission loss is taken into consideration. Therefore, a longer transmission distance can be achieved by the proposed AFDBF array compared with the phased arrays in Refs. [9], [10], [23] when the same number of Tx channels are turned on. Furthermore, the Tx and Rx arrays of the proposed AFDBF array realized ±47° and ±45° scanning ranges in all planes, respectively. However, the arrays proposed in Refs. [8], [9], [10] exhibit a relatively narrow scanning range in the E-plane. In Refs. [9], [10], the beam scanning ranges are all within ±40° with a 3-4 dB scanning loss in the elevation plane in the entire operation band. In addition, the proposed AFDBF array can achieve 1 Gbaud 64-QAM waveforms transmission at 1.50 m with a 2.45% EVM. However, 200 Mbaud 64-QAM waveforms can be transmitted at 1.2 m with 4.1% (at 28 GHz) and 4.3% (at 30 GHz) EVM by the 32-element phased array in Ref. [9], and 400 Mbaud 64-QAM waveforms transmission at 1.38 m is realized with 3.20% EVM by the 64-element phased array in Ref. [10]. A 1.05% EVM using 50 Mbaud 64-QAM waveforms is measured in the array in Ref. [23]. Compared with the array in Refs. [9], [10], [23], a better EVM performance and enhanced transmission rate are achieved by the proposed AFDBF array when Tx complex modulation signals. Overall, the proposed mm-Wave AFDBF array exhibits a very competitive performance and is expected to be a promising candidate for B5G/6G communications.

6. Conclusions

A single-board integrated AFDBF array operating at 24.25-29.50 GHz for B5G/6G applications was presented in this article. The proposed array effectively integrates the Tx and Rx arrays independently into a single board by utilizing four-channel full-digital Tx/Rx chips, demonstrating a more flexible and compact structure compared with the conventional brick-type structure array. In comparison with the conventional analog phased-array architecture, the proposed AFDBF array architecture exhibits a flexible multibeam ability, higher accuracy of beam administration, and precoding freedom, as well as a fast beam-steering speed. The asymmetric architecture also exhibits advantages in terms of the system cost and power consumption, compared with a SFDBF array. Moreover, the energy efficiency performance of the proposed array is estimated to be superior than that of an HYBF array. The proposed Tx and Rx arrays are able to achieve a scanning range of ±47° and ±45° in both the xoz and yoz planes at 27 GHz, respectively. Furthermore, OTA link measurements were carried out, and data rates of 8 Gbps were achieved using 64-QAM waveforms with an EVM lower than 4% at 27 GHz. The excellent performance and its advantages make the proposed single-board integrated AFDBF array a highly promising candidate for beyond-5G and coming 6G applications.

Acknowledgments

This work was supported by the National Key Research and Development Program of China (2020YFB1804900 and 2022YFE0210900), the Fundamental Research Funds for the Central Universities (2242022k60008 and 2242022k30003), the National Natural Science Foundation of China (62301152 and 61627801), the Youth Talent Promotion Foundation of Jiangsu Science and Technology Association (TJ-2023-074), and the Start-up Research Fund of Southeast University (RF1028623286). The authors would like to thank Mr. Tianyi Huo, Dr. Chong Guo, Mr. Renrong Zhao, Dr. Yuanwei Zhu, Dr. Weiheng Chen, and other with the State Key Laboratory of Millimeter Waves, Southeast University, Nanjing, China, for their kingly help in measurement. The authors also thank the editors and the anonymous reviewers for their insightful and constructive comments.

Compliance with ethics guidelines

Qingqing Lin, Jun Xu, Kai Chen, Long Wang, Wei Li, Zhiqiang Yu, Guangqi Yang, Jianyi Zhou, Zhe Chen, Jixin Chen, Xiaowei Zhu, and Wei Hong declare that they have no conflict of interest or financial conflicts to disclose.

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	W. Hong, Z.H. Jiang, C. Yu, D.B. Hou, H. Wang, C. Guo, et al. The role of millimeter-wave technologies in 5G/6G wireless communications. IEEE J Microw, 1 (1) (2021), pp. 101-122.

[2]	C.X. Wang, X. You, X. Gao, X. Zhu, Z. Li, C. Zhang, et al. On the road to 6G: visions, requirements, key technologies, and testbeds. IEEE Commun Surv Tut, 25 (2) (2023), pp. 905-947.

[3]	W. Chen, X. Lin, J. Lee, A. Toskala, S. Sun, C.F. Chiasserini, et al. 5G-advanced toward 6G: past, present, and future. IEEE J Sel Area Comm, 41 (6) (2023), pp. 1592-1619.

[4]	W. Saad, M. Bennis, M.Z. Chen. A vision of 6G wireless systems: applications, trends, technologies, and open research problems. IEEE Netw, 34 (3) (2020), pp. 134-142.

[5]	W. Hong, Z.H. Jiang, C. Yu, J. Zhou, P. Chen, Z. Yu, et al. Multibeam antenna technologies for 5G wireless communications. IEEE Trans Antennas Propag, 65 (12) (2017), pp. 6231-6249.

[6]	K.M. Luk, B.Y. Duan. Advanced antennas push forward wireless connectivity. Engineering, 11 (4) (2022), pp. 1-2.

[7]	T. Tataria, M.F. Shafi, A.F. Molisch, M. Dohler, H. Sjöland, F. Tufvesson. 6G wireless systems: vision, requirements, challenges, insights, and opportunities. Proc IEEE, 109 (7) (2021), pp. 1166-1199.

[8]	B. Yang, Z. Yu, J. Lan, R. Zhang, J. Zhou, W. Hong. Digital beamforming-based massive MIMO transceiver for 5G millimeter-wave communications. IEEE Trans Microw Theory Tech, 66 (7) (2018), pp. 3403-3418.

[9]	S. Wang, G.M. Rebeiz. Dual-band 28- and 39-GHz phased arrays for multistandard 5G applications. IEEE Trans Microw Theory Tech, 71 (1) (2023), pp. 339-349.

[10]	Y. Yin, B. Ustundag, K. Kibaroglu, M. Sayginer, G.M. Rebeiz. Wideband 23.5-29.5-GHz phased arrays for multistandard 5G applications and carrier aggregation. IEEE Trans Microw Theory Tech, 69 (1) (2021), pp. 235-247.

[11]	H. Wang. Overview of future antenna design for mobile terminals. Engineering, 11 (4) (2022), pp. 12-14.

[12]	B. Duan. Evolution and innovation of antenna systems for beyond 5G and 6G. Front Inform Technol Electron Eng, 21 (1) (2020), pp. 1-3.

[13]	O. Kazan, Z. Hu, L. Li, A. Alhamed, G.M. Rebeiz. An 8-channel 5-33-GHz transmit phased array beamforming IC with 10.8-14.7-dBm P_sat for C-, X-, Ku-, and Ka-band SATCOM. IEEE Trans Microw Theory Tech, 71 (5) (2023), pp. 2029-2039.

[14]	R. Klimovich, S. Jameson, E. Socher. W-band endfire 2D phased-array transmitter based on ×9 CMOS active multiplier chips. IEEE Trans Antennas Propag, 68 (12) (2020), pp. 7893-7904.

[15]	E. Arnieri, L. Boccia, G. Amendola, S. Glisic, C. Mao, S. Gao, et al. An integrated radar tile for digital beamforming X-/Ka-band synthetic aperture radar instruments. IEEE Trans Microw Theory Tech, 67 (3) (2019), pp. 1197-1206.

[16]	C.X. Mao, S. Gao, C. Tienda, T. Rommel, A. Patyuchenko, M. Younis, et al. X/Ka-band dual-polarized digital beamforming synthetic aperture radar. IEEE Trans Microw Theory Tech, 65 (11) (2017), pp. 4400-4407.

[17]	S. Shahramian, M.J. Holyoak, A. Singh, Y. Baeyens. A fully integrated 384-element, 16-tile, W-band phased array with self-alignment and self-test. IEEE J Solid State Circuits, 54 (9) (2019), pp. 2419-2434.

[18]	B. Sadhu, A. Paidimarri, D.X. Liu, M. Yeck, C. Ozdag, Y. Tojo, et al. A 24-30-GHz 256-element dual-polarized 5G phased array using fast on-chip beam calculators and magnetoelectric dipole antennas. IEEE J Solid State Circuits, 57 (12) (2022), pp. 3599-3616.

[19]	K. Kibaroglu, M. Sayginer, T. Phelps, G.M. Rebeiz. A 64-element 28-GHz phased-array transceiver with 52-dBm EIRP and 8-12-Gb/s 5G link at 300 meters without any calibration. IEEE Trans Microw Theory Tech, 66 (12) (2018), pp. 5796-5811.

[20]	S.H. Wu, L.K. Chiu, J.W. Wang. Reconfigurable hybrid beamforming for dual-polarized mmWave MIMO channels: stochastic channel modeling and architectural adaptation methods. IEEE Trans Commun, 66 (2) (2018), pp. 741-755.

[21]	P. Taghikhani, K. Buisman, C. Fager. Hybrid beamforming transmitter modeling for millimeter-wave MIMO applications. IEEE Trans Microw Theory Tech, 68 (11) (2020), pp. 4740-4752.

[22]	B. Yang, Z. Yu, R. Zhang, J. Zhou, W. Hong. Local oscillator phase shifting and harmonic mixing-based high-precision phased array for 5G millimeter-wave communications. IEEE Trans Microw Theory Tech, 67 (7) (2019), pp. 3162-3173.

[23]	C. Guo, W. Hong, L. Tian, Z.H. Jiang, J. Zhou, J. Chen, et al. Design and implementation of a full-digital beamforming array with nonreciprocal Tx/Rx beam patterns. IEEE Antennas Wirel Propag Lett, 19 (11) (2020), pp. 1978-1982.

[24]	Hong W, Zhou J, Chen J, Jiang Z, Yu C, Guo C.Asymmetric full-digital beamforming mmWave massive MIMO systems for B5G/6G wireless communications. In: Proceedings of Asia-Pacific Microwave Conference; 2020 Dec 8-11; Hong Kong, China. New York City: IEEE; 2020. p. 31-2.

[25]	D.C. Kim, S.J. Park, T.W. Kim, L. Minz, S.O. Park. Fully digital beamforming receiver with a real-time calibration for 5G mobile communication. IEEE Trans Antennas Propag, 67 (6) (2019), pp. 3809-3819.

[26]	L. Kuai, J. Chen, Z.H. Jiang, C. Yu, C. Guo, Y. Yu, et al. A N260 band 64 channel millimeter wave full-digital multi-beam array for 5G massive MIMO applications. IEEE Access, 8 (2020), pp. 47640-47653.

[27]	C. Guo, L. Tian, Z.H. Jiang, W. Hong. A self-calibration method for 5G full-digital TDD beamforming systems using an embedded transmission line. IEEE Trans Antennas Propag, 69 (5) (2021), pp. 2648-2659.

[28]	X. Yang, S. Jin, G.Y. Li, X. Li. Asymmetrical uplink and downlink transceivers in massive MIMO systems. IEEE Trans Veh Technol, 70 (11) (2021), pp. 11632-11647.

[29]	J. Lu, J. Zhang, S. Cai, B. Li, X. Zhu, W. Hong. Downlink wideband channel estimation for asymmetrical full-digital system. IEEE Wirel Commun Lett, 11 (9) (2022), pp. 1830-1834.

[30]	R.C. Hansen. Phase array antennas. Wiley, New York City (1998).

[31]	S. Mumtaz, J. Rodriguez, L. Dai. mmWave massive mimo: a paradigm for 5G. Academic, San Diego (2016).

[32]	S. Haykin.Communication systems. ( 4th ed.), Wiley, New York City (2004).

[33]	S. Han, I. Chih-Lin, Z. Xu, C. Rowell. Large-scale antenna systems with hybrid analog and digital beamforming for millimeter wave 5G. IEEE Commun Mag, 53 (1) (2015), pp. 186-194.

[34]	S. Yang, Z. Yu, J. Zhou. A low-loss broadband planar transition from ground coplanar waveguide to substrate-integrated coaxial line. IEEE Microw Wirel Compon Lett, 31 (11) (2021), pp. 1191-1194.

[35]	Y. Shao, X.C. Li, L.S. Wu, J.F. Mao. A wideband millimeter-wave substrate integrated coaxial line array for high-speed data transmission. IEEE Trans Microw Theory Tech, 65 (8) (2017), pp. 2789-2800.

[36]	V. Rathi, G. Kumar, K.P. Ray. Improved coupling for aperture coupled microstrip antennas. IEEE Trans Antennas Propag, 44 (8) (1996), pp. 1196-1198.

[37]	Z.F. Ding, S.Q. Xiao, C.R. Liu, M.C. Tang, C. Zhang, B.Z. Wang. Design of a broadband, wide-beam hollow cavity multilayer antenna for phased array applications. IEEE Antennas Wirel Propag Lett, 15 (2016), pp. 1040-1043.

[38]	A. Clavin, D. Huebner, F. Kilburg. An improved element for use in array antennas. IEEE Trans Antennas Propag, 22 (4) (1974), pp. 521-526.

[39]	G. Yang, J. Li, J. Yang, S.G. Zhou. A wide beamwidth and wideband magnetoelectric dipole antenna. IEEE Trans Antennas Propag, 66 (12) (2018), pp. 6724-6733.

[40]	B. Feng, L. Li, K.L. Chung, Y. Li. Wideband widebeam dual circularly polarized magnetoelectric dipole antenna/array with meta-columns loading for 5G and beyond. IEEE Trans Antennas Propag, 69 (1) (2021), pp. 219-228.

[41]	K. Itoh, D.K. Cheng. A novel slots-and-monopole antenna with a steerable cardioid pattern. IEEE Trans Aerosp Electron Syst, 8 (2) (1972), pp. 130-134.

[42]	Q. Gu. RF system design of transceivers for wireless communications. Springer, New York City (2005).

[43]	Y. Hua, T.K. Sarkar, D.D. Weiner. An L-shaped array for estimating 2-D directions of wave arrival. IEEE Trans Antennas Propag, 39 (2) (1991), pp. 143-146.