Technology Assessment of Aperture Coupled Slot Antenna Array in Groove Gapwaveguide for 5G Millimeter Wave Applications

A compact triple-layer series fed slot unit array using Groove Gap Waveguide (GGW) technology for 5G millimeter-wave applications is presented in this paper. The design employed a two-step coupling mechanism for the radiating slots. The aperture coupling design when combined with the groove gapWaveguide (GGW) technology allows achieving three different goals: First, a reduction of the array size in the transversal direction, so more unit arrays can be added without grating lobes. Secondly, low insertion losses with high radiation efficiency. And third, a reduction of the manufacturing complexity with good stability against fabrication tolerances of up to 0.1 mm, making the design feasible to be massively deployed. The antenna was fabricated and measured showing good agreement with the numerical simulation results, the antenna provides a peak realized gain of 12.15dB at the central frequency of operation and a fractional bandwidth (FBW) of 9%.


I. INTRODUCTION
T HE increasing demand for high data speed, such as 8K video streaming and virtual augmented realities is expected to fold-1000 the data traffic in the upcoming decade [1]. Exploring new less-congested spectrum bands of high frequencies such as the millimeter-wave (mmWave) bands (30 -300 GHz) is a promising solution to increase network capacity. However, new architectures with densification of small-cell technologies are required to mitigate propagation losses, reduce interference and increasing coverage [2]. Such densification requires the extensive deployment of dense wireless networks such as backhaul, for aggregating and sending data traffic from the radio access to the (wired) backbone segment or the cost-effective 5G smart repeaters to ensure a complete wireless coverage in difficult urban spots, among many other requirements. In this regard, highgain antennas robust to manufacturing tolerances are required for the massive distribution of wireless network nodes. 5G mmWave band comprises the FR2 band, with frequencies ranging from 24.25GHz to 48.2 GHz where smart repeaters and many backhaul applications work on. However, the FR2 band for the latter one is not sufficient to deliver today's capacity peaks [3], offloading to E-band (71-76GHz and 81-86GHz) spectrum is taking place.
In addition, a minimum antenna gain of 43 dB at E-Band is imposed by the FCC to 5G Backhaul cells, which requires a large antenna size such as 1-foot parabolic antennas, compared to 5G smart repeaters only 22 dB is required. Companies such as Ericsson and Nokia, alongside Avit and Comsearch pushed to decrease the antenna gain requirement to 38 dB in order to support smaller antennas, so integration in equipment at the street level such as light poles can be easier [4]. There is a clear trend to reduce the antenna size, so a more compact solutions can be adopted. It is obvious that gain requirements fix the minimum antenna dimension but a technology with higher efficiency and smaller thickness requirements will be preferred. More integrated solutions have been proposed such as horn arrays [5], transmit arrays [6], and lens antennas [7] but the size is still too bulky to be easily fitted on urban equipment.
In this context, compact low-thickness, high-efficiency, high-gain antennas with good fabrication tolerance stability are required for 5G mmWave network deployments. When referring to very compact solutions traditional arrays are an attractive solution and can be mainly classified into corporate feeding networks and series fed arrays. The first is more commonly used for fixed point-to-point communications while with the latter technique, a phased array solution can be implemented making it more preferable.
Conventional series-fed arrays in a microstrip line can be found in [8] and [9]. However, the losses with dielectric and its uncertainties on the actual value of permittivity make it difficult to correlate the measured results with the simulations at these high frequencies. Waveguides can solve these problems as the dielectric is avoided by using four sealed walls to guide the electromagnetic wave through the air medium. Similar arrays in rectangular waveguide (RW) can be found in the literature by implementing slots, and they are mainly classified into two categories, vertical rectangular waveguide (V-RW) or narrow wall slots [10] and horizontal rectangular waveguide (H-RW) or broad wall slots [11]. Regarding V-RW, they offer more compactness in the antenna plane so beam steering capabilities using λ/2 separated unit arrays can be achieved without many dimensional restrictions as well as a considerable reduction in sidelobe level (SLL). However, the slots must be excited using a tilt angle as currents are in counter phase every λ g /2 [12], [13] , which is highly undesirable as cross polarization (CPol) appears. Many studies have been conducted in order to make these elements radiate in phase while maintaining low levels of CPol by using parasitic elements inside the waveguide to modify the current distribution [14]; however, a very demanding manufacturing precision is required and they are bandwidth limited. In addition, fast prototyping of waveguides needs to be manufactured on separate pieces, requiring perfect sealing when unifying them in order to avoid leakage [15].
In this context, gapWaveguide (GW) has attracted attention during the last decade, and the invention by S. Kildal offers a contactless full metal two parallel plates guiding structure solution by using an artifitial magnetic conductor (AMC) at both sides of the transmission line, so all the electromagnetic energy is confined by the parallel plates and the AMC [16]. The AMC can be achieved by rows of periodic nails, although other conformations can be adopted such as mushrooms type [17], with the only condition that the space between the AMC and the parallel plate is smaller than λ g /4. Clearly, in this configuration, there is no need for an electrical connection between the top and bottom conductive planes of the waveguide, leading to a low loss and relatively easy manufacturing process. In order to benefit from the advantages of the GW technology, different designs have been reported in the literature, which can be classified into two main groups, ridge gap-Waveguide (RGW) [18] which is the most popular design owing to its analogy to the microstrip line but in an inverted form and GGW [19] with a working principle similar to that of RW. There are two subgroups of GGW technology: named horizontal groove gapWaveguide (H-GGW) and vertical groove gapWaveguide (V-GGW). Some designs have been made on H-GGW [20] but the size is too long in the transversal direction of the array similar to H-RW. Only a few works have been published to solve the problems of slot excitation using V-GGW, in [21] parasitic dipole elements are used in the upper layer by means of an extra layer of substrate; however, it only shows 2% FBW.
In comparison to the antennas presented in the literature (see Table 1), we propose a new method for exciting narrow wall slots from V-GGW without added manufacturing complexity and a good fabrication tolerance stability of up to 0.1 mm (corresponding to the tolerances of a conventional manufacturing technique). The antenna consists of a triplelayer structure with a novel aperture coupling mechanism offering a linearly polarized, high efficiency (92.2%) with a high measured FBW (9%) while being compact in the transversal direction, so the gain can be simply increased by just adding more unit arrays along the transversal direction of the array without having large SLL.
The paper is organized as follows, in the next section II the antenna structure and working principle are described, section III shows a study of the tolerance in different GW technologies and a comparison with a conventional design of series fed array in RGW is made at simulation level, in section IV the results of the measured prototype and simulated ones are correlated and concluded in section VI. In section V the limitations of the work done in this study are also introduced.

II. ANTENNA DESIGN
The challenge in the development of new highly integrable antennas for urban 5G mmWave wireless nodes raises the demand for new compact solutions with low fabrication precision requirements, high efficiency, and high gain with beam-pointing requirements in some scenarios. Series fed slots in V-GGW technology were chosen as a good candidate because of three main reasons. First, with only one excitation port, all the elements radiate in phase, reducing the complexity of the feeding network when compared to corporate feeding. Second, the antenna is compact in the transversal direction of the array, so gain can be increased by simply adding 0.5λ spaced arrays in that direction without losing beam steering capabilities. Finnally, it offers an easy assembly process.

A. SLOT EXCITATION PROBLEM
The slot elements from a series fed design in the narrow wall of vertically oriented waveguide-related technologies cannot be directly excited as stated in the introduction.  The currents from the top wall of the V-GGW are in counter phase every 0.5λ g and oriented towards the transver-sal direction of the transmission line (see Fig. 1). Longitudinal slots can be implemented (see Fig. 1) (a) with a spacing of λ g in order to radiate in phase; however, this will result in large grating lobes. A conventional method to keep the elements radiating every 0.5λ g as shown in Fig. 1 (b) is performed by using tilted slots, however this will result in high levels of CPol. Many studies have been carried out using intrusion elements or an extra layer of substrate-based radiating elements as stated in the introduction an antenna with no grating lobes and low CPol level can be achieved at the same time. However, they are either bandwidth-limited or require a high manufacturing precision. An isometric and top view of the proposed unit array antenna for this study with 0.5λ g radiating slots in V-GGW technology can be seen in Figs  The V-GGW antenna is mainly composed of three layers: the feeding layer (bottom) L 1 , coupling layer (middle) L 2 and radiating layer (top) L 3 , whose dimensions are listed in Tab. 5. In the bottom layer L 1 there is a direct transition from WR-10 to GGW which consists of a slot (W W x L W x H W ) with corners of r c1 radius of curvature. An H-Plane power divider [24] is used to make a central feeding possible in order to achieve symmetrical fields at the radiating elements. The power divider is matched through a stepped transition inserted just below the middle layer Fig. 4. The groove line (W gr x L gr x H gr ) has rounded ends with r c1 radius of curvature. All the energy from the guiding line propagates through four slots (W Lslot x L Lslot ) from the middle layer L 2 with corners of tye r c2 radius of curvature. These slots are approximately spaced λ g from each other so they radiate in phase and each one of them, excites two of the eight λ g /2 spaced slots of (W U Slot x L U Slot ) from the top layer L 3 with a curvature radius of r c3 . In this manner, all the slots radiate in phase with no grating lobes or high CPol level problems. A symmetrical optimized distribution was considered for the radiating slots, where the intensity of the coupled fields was adjusted by L gr and the offset of the radiating slots dx in order to achieve the lowest SLL while maintaining a high FBW. Lower SLL distributions such as Chebyshev and Taylor can be implemented using a non-uniform offset dx of radiating slots from the top layer L 3 to adjust the coupled energy. The central elements from layers L 1 and L 2 are surrounded by a bed of nails with direct contact with the upper layer and a small spacing p1 with respect to the groove line. It acts as an AMC that prevents leakage through the sides of the transmission line. The dimensions of the bed of nails must be designed to ensure a good AMC performance at the frequencies of interest. A fast calculation of the modes of propagation can be performed by defining the unit cell of the bed of nails via the eigenmode solver from the CST simulator. In Fig. 5 the stopband of the AMC structure ranges from 60 GHz to 100 GHz if an air gap of 0.25λ is left between the upper layer and the nails, and 0-100GHz if they are in direct contact. If the nails are in contact with the upper layer, one single row of nails would be sufficient as will be demonstrated in the experimental section IV, resulting in a more compact unit array (0.57λ of minimum width x 1.91λ height). Two calibrated metallic screws (CalH) were used to prevent misalignment between the layers and two metallic screws (SH) to fix the separation between the layers.

III. TOLERANCE ANALYSIS
The massive deployment of antennas requires an important analysis of the manufacturing tolerances. In order to ensure a good radiation performance for N antennas with fabrication tolerances of up to 0.1mm ( which is a standard manufacturing tolerance for many industrial technologies) at high frequencies like E-Band is one of the main challenges that this paper has focused on. Gap waveguide technology was selected mainly because of its low loss and easy assembly properties.

A. DESIGN
A first test consisted of a comparison between the impedance stability of the three types of GW, the H-GGW, V-GGW and RGW when introducing dimensional errors of ±0.1mm in order to check the impedance stability of each technology.      The antenna also has eight series fed slots with a spacing of approximately 0.5λ g . Note that only a single layer of radiating slots is needed to directly excite the elements in phase. L end is approximately 0.25λ so all the remaining power is reflected back and a standing wave is created over the array axis. A uniform distribution of the delivered power was used with elliptical slots for bandwidth improvement. All the dimensions of each parameter are summarized at Tab. 3, for a complete description of a similar antenna, see [23].

B. ITERATION OF ERRORS
There are 13 dimensional parameters for the case of RGW design and 31 dimensional parameters for the case of V-GGW, assuming a uniform error distribution with only three discrete values of −0.1,0 and +0.1 mm, see Tab. 4, for every single parameter this would result in a total amount of 1.59 million and over trillions of combined errors respectively, which is obviously not feasible to simulate.    A first analysis of the critical dimensions was carried out to reduce this space of combinations. Errors of ± 0.1 mm VOLUME 4, 2016 were applied to each parameter separately for the RGW and V-GGW, resulting in 26 and 54 input reflection coefficients, the parameters used in this first study are marked with at least one * in Tab. 3 and Tab. 5 respectively. Some parameters were not included in the tolerance analysis such as the curvature radius (r c1 , r c2 , r c3 ) because it is related to the shape rather than dimensional error and, the height of the feeding port (H W ) because it does not affect the impedance of the structure. Parameters from the Tabs. 3 and 5 marked with a double * are the most critical ones in terms of reflection coefficients seen at Fig. 8 (a) and (b) respectively, resulting in a penalty of greater than 15% of FBW reduction if taking into account that the only valid bands for both designs are those marked by the input reflection coefficient when using nominal dimensions. For V-GGW the critical parameters are the length of the slots from each layer L Lslot and L U Slot , W nail and p1. For the case of RGW the results are much worse (>50% of the FBW penalty) with any parameter. Only these parameters are selected for the final study using combined errors and are plotted in Figs. 9 and 10.  For the case of the input reflection coefficients in this V-GGW design, 89% of the combined errors guarantee a frequency range from 74.25GHz to 80GHz, resulting in a 7.5% secured FBW. Clearly, V-GGW shows much better tolerance stability than RGW, as predicted previously in the Design subsection. Only pattern stability for the design in V-GGW is represented in Fig. 10 because the results are well-matched, showing good pointing stability as well as realized gain. An increase of up to 7.5dB in SLL was due to the errors related to the slot spacing dY up from the top layer L 3 . Of course, if considering misalignment errors between each layer, we will have a worse response in terms of SLL although it has almost no impact on the input reflection coefficients.

IV. EXPERIMENTAL RESULTS
The antenna was manufactured using two different methods. L1 and L2 were fabricated using a standard CNC milling machine from the UPC facilities. L3 was fabricated externally using laser cutting technology as the width of the slots was too thin to be performed at our facilities. Fabricated dimensional errors are listed in Tab. 5, which were measured using the Nikon Measurescope  The most deviated errors were caused by the displacement of the L3 layer (DispX3 and DispY 3), the connection of the WR10 (W W ) to the waveguide structure, the width of the slots W Lslot of L2 and dimensions (height x length) of the stepped matching element (H t1 x L t1 ) and (H t2 x L t2 ) with a deviation up to 0.15 mm. The reason for the large displacement off L3 was the positional error of the calibrated and screwed holes. Measurements were performed in a 4x4x3 m 3 room from the UPC facilities, with a proper time gating post-processing technique, the reflections inside the room can be filtered out [25]. The measurements and simulations using CST Studio software included the reflection coefficients and realized gain patterns, which were cut at 77 GHz, and are shown in figures 13,15 and 16. There is a good correlation between the measured input reflection coefficient (74.5GHz -81.25GHz) and the sim-  ulated one (74GHz -81.5GHz). A portion of mismatched frequencies ranges from 79.25 to 80.75 GHz at only -9.1 dB. The design was performed using a double row of nails with direct contact with the upper layer, which would result in large dimensions in the transverse direction of the array. However one single row of nails is sufficient, resulting in a more compact form (0.57λ width x 1.91λ height) without altering the results (see Fig 13). The stability of the gain versus frequency for the whole band is shown in blue in the right axis of Fig. 14, with a maximum gain reduction down to 11dB at 79GHz. In the same Fig. 14 we can also see the efficiency of the antenna in black color which was measured to be greater than 90% for the whole band ranging from 74GHz to 81GHz.
A realized gain of 12.15 dB was measured which also showed good agreement with the simulations for both planes at 77GHz. A simulated SLL of -18.5dB in H-plane was VOLUME 4, 2016  increased up to -16.5 dB at θ = -28 deg in the measurements caused by the misalignment of the radiating layer l 3 with respect to the other two ones, leading to a non-perfect symmetrical field distribution among the radiating slots.

V. LIMITATIONS OF THE STUDY
In this paper we show a new method for exciting series fed slot arrays in V-GGW, although the design is easy to be scaled to a larger aperture antenna for 5G mmWave applications by just adding more arrays, there are some remaining studies to be done.
First, when scaling to N arrays spaced at 0.57λ (minimum achievable inter-array space with this design), a compact distribution network needs to be designed.
Second, only eight elements were used in the series-fed array. For achieving high gain levels required by future 5G mmWave applications, further studies have to be carried out by using more array elements depending on the specific 5G mmWave application and see its impact on the FBW and the tolerances in order to simplify its scalability.

VI. CONCLUSIONS
A single compact series-fed array antenna using a novel aperture coupling mechanism in V-GGW technology has been proposed. The measured gain of the antenna was 12.15dB by comparing it to a standard horn antenna of 23.5dB. The efficiency at 77GHz was measured to be 92.25% by assuming a directivity similar to the simulated one (12.5 dB). The antenna offers a simulated FBW of 10% which is reduced to 9% when measuring it.
The antenna shows good stability against fabrication errors of up to 0.1 mm for both input reflection coefficients and radiation pattern, which is much better than the design in RGW. It provides a minimum realized gain of 12.1dB at 77GHz for the worst case of combined errors at the simulation level and offer good tilting stability. The radiating aperture of the antenna can be increased by adding more unit arrays to increase the directivity and the frequency can be scaled depending on the specific application of future 5G mmWave wireless networks requiring massive deployments. , and the Chairman of the EIT-Urban Mobility European Association. He has authored more than 200 scientific and technical articles, reports, and chapters in specialized volumes. His current research interests include antennas, electromagnetic scattering and imaging, and system miniaturization for wireless, and sensing industrial and bio-applications from microwaves to terahertz frequencies.
JORDI ROMEU (Fellow, IEEE) was born in Barcelona, Spain, in 1962. He received the Ingeniero de Telecomunicación and Doctor Ingeniero de Telecomunicaciòn degrees from the Universitat Politècnica de Catalunya (UPC), in 1986 and 1991, respectively. In 1985, he joined the Antennalab, Signal Theory and Communications Department, UPC, where he is currently a Full Professor and involved in antenna near-field measurements, antenna diagnostics, and antenna design. He was a Visiting Scholar at the Antenna Laboratory, University of California at Los Angeles, Los Angeles, in 1999, on an NATO Scientific Program Scholarship, and at the University of California at Irvine, in 2004. He holds several patents and has published 60 refereed articles in international journals and 80 conference proceedings. He was a Grand Winner of the European IT Prize, awarded by the European Commission, for his contributions in the development of fractal antennas, in 1998. More information can be found in http : //www.researchgate.net/prof ile/Jordi-Romeu. VOLUME 4, 2016