Low Cost AIP Design in 5G Flexible Antenna Phase Array System Application

In this paper, a low cost 28 GHz Antenna-in-Package (AIP) for a 5G communication system is designed and investigated. The antenna is implemented on a low-cost FR4 substrate with a phase shift control integrated circuit, AnokiWave phasor integrated circuit (IC). The unit cell where the array antenna and IC are integrated in the same plate constructs a flexible phase array system. Using the AIP unit cell, the desired antenna array can be created, such as 2 × 8, 8 × 8 or 2 × 64 arrays. The study design proposed in this study is a 2 × 2 unit cell structure with dimensions of 18 mm × 14 mm × 0.71 mm. The return loss at a 10 dB bandwidth is 26.5–29.5 GHz while the peak gain of the unit cell achieved 14.4 dBi at 28 GHz.


Introduction
Antenna technology is the latest breakthrough in design to accelerate cellular networks that aims to optimize the communication system in terms of smoothness and cost of the communications itself. Along with the development of the cellular network, currently the newest generation technology of the networks has arrived at the fifth generation (5G) network. The success of the 5G network paved its way to research for the newest technology product and to provide the best communication platform [1,2]. The 5G internet has become the center of research on how to improve its capabilities and novelty of the technology itself. The design of the Antenna in Package (AIP) is one field of research of 5G technology that can be improved to maximize the capabilities and functionalities of the technology [3].
The proposed research aims to optimize the bandwidth capabilities of the 5G technology by designing the AIP with models that exhibit low-cost array antenna design. The recent research on this field performed optimization of the 5G technology through space-frequency index modulation, spectral, energy, and economic fields [4,5]. On the other hand, many research also explored optimizing the scalability of the bandwidth of the 5G communication network. The first research was from K. Kibaroglu et al. [6] that has successfully designed a simple model 32-element (4 × 8) working at 28 GHz on the phased-array transceiver for 5G communication technology based on a 2 × 2 beamformer core chips. The research has achieved an effective isotropic radiated power (EIRP) of 43 dBm at P1dB, and the final state-of-art data rate was achieved in 1.0-1.6 Gb/s in a single beam using 16-QAM.

Patch Antenna Design
In this study, the micro strip patch antennas constructed the array system. The patch antenna is a kind of a resonant antenna that is like a resonant cavity. One important parameter of a resonant cavity is its quality factor (Q 0 ), which is defined as shown in Equation (1) [22].
where ω is the frequency, W e is the stored energy in the resonant cavity, and P l is the power loss of the resonant cavity. There are three kinds of losses in the resonant antenna namely radiation loss (P rad ), dielectric loss (P d ), and conducted loss (P c ). The formula is shown in Equation (2).
The antenna efficiency can be enhanced when the dielectric loss is reduced. This is due to the antenna efficiency (ξ), as shown in Equation (3) that is proportional to Q 0 when the conducted loss is fixed in the critical coupled condition.
This study used an air-filled cavity structure to design the patch antenna on a standard FR4 substrate [23]. This design constructed a metal patch that was located on the FR4 substrate with the open air cavity. The reference ground used the copper layer on the carrier board as illustrated in Figure 1. This design can reduce the dielectric loss and enhance the patch antenna performance with better radiation efficiency. The dielectric constant of air was 1.0006 and the loss tangent of air was 0, which can enhance the patch antenna performance with better radiation efficiency. The top and cross section views are shown in Figure 2.
Micromachines 2020, 11, x FOR PEER REVIEW 3 of 15 The antenna efficiency can be enhanced when the dielectric loss is reduced. This is due to the antenna efficiency (ξ), as shown in Equation (3) that is proportional to when the conducted loss is fixed in the critical coupled condition.
This study used an air-filled cavity structure to design the patch antenna on a standard FR4 substrate [23]. This design constructed a metal patch that was located on the FR4 substrate with the open air cavity. The reference ground used the copper layer on the carrier board as illustrated in Figure 1. This design can reduce the dielectric loss and enhance the patch antenna performance with better radiation efficiency. The dielectric constant of air was 1.0006 and the loss tangent of air was 0, which can enhance the patch antenna performance with better radiation efficiency. The top and cross section views are shown in Figure 2.  The return loss and radiation efficiency of the patch antenna are presented in Figure 3. Figure 3a illustrates that the return loss of the patch antenna with air cavity is better than 10 dB at 26.5-29.5 GHz. Figure 3b presents the radiation efficiency of the two types of antenna where the radiation efficiency of the patch antenna with air cavity was 92% while the radiation efficiency of patch antenna without air cavity was 66.25% at 28 GHz. The radiation efficiency was enhanced by 25.75% at 28 GHz. The maximum radiation efficiency of the patch antenna with air cavity was 93.28% while the maximum radiation efficiency of patch antenna without air cavity was 77.25%. The radiation efficiency was enhanced by 16.03%. The antenna efficiency can be enhanced when the dielectric loss is reduced. This is due to the antenna efficiency (ξ), as shown in Equation (3) that is proportional to when the conducted loss is fixed in the critical coupled condition.
This study used an air-filled cavity structure to design the patch antenna on a standard FR4 substrate [23]. This design constructed a metal patch that was located on the FR4 substrate with the open air cavity. The reference ground used the copper layer on the carrier board as illustrated in Figure 1. This design can reduce the dielectric loss and enhance the patch antenna performance with better radiation efficiency. The dielectric constant of air was 1.0006 and the loss tangent of air was 0, which can enhance the patch antenna performance with better radiation efficiency. The top and cross section views are shown in Figure 2.  The return loss and radiation efficiency of the patch antenna are presented in Figure 3. Figure 3a illustrates that the return loss of the patch antenna with air cavity is better than 10 dB at 26.5-29.5 GHz. Figure 3b presents the radiation efficiency of the two types of antenna where the radiation efficiency of the patch antenna with air cavity was 92% while the radiation efficiency of patch antenna without air cavity was 66.25% at 28 GHz. The radiation efficiency was enhanced by 25.75% at 28 GHz. The maximum radiation efficiency of the patch antenna with air cavity was 93.28% while the maximum radiation efficiency of patch antenna without air cavity was 77.25%. The radiation efficiency was enhanced by 16.03%. The return loss and radiation efficiency of the patch antenna are presented in Figure 3. Figure 3a illustrates that the return loss of the patch antenna with air cavity is better than 10 dB at 26.5-29.5 GHz. Figure 3b presents the radiation efficiency of the two types of antenna where the radiation efficiency of the patch antenna with air cavity was 92% while the radiation efficiency of patch antenna without air cavity was 66.25% at 28 GHz. The radiation efficiency was enhanced by 25.75% at 28 GHz. The maximum radiation efficiency of the patch antenna with air cavity was 93.28% while the maximum radiation efficiency of patch antenna without air cavity was 77.25%. The radiation efficiency was enhanced by 16.03%.

Array Antenna Design
The operation frequency band of the 5G system achieves the Ka-band. A small wavelength, small beam width, and high atmospheric attenuation are the shortcomings of this frequency band while its great advantages are its larger bandwidth and higher data rate. The multiple antenna techniques (MTA) is the solution that can solve wave shadowing of millimeter wave propagation [24]. The array antenna is an important development. The array antenna is composed of antennas that are arranged periodically as illustrated in Figure 4. The beam main lobe can be tilted by changing the phase of the antennas, which is called the beam steering technique. In this study, the four patch antennas constituted a 2 × 2 array antenna as shown in Figure 5. An AnokiWave phasor IC was set at the same side with the patch antennas. Such an arrangement makes the array antenna become a complete system. This modular system is more flexible and expandable, which is widely known as the Antenna-in-Package (AIP).

Array Antenna Design
The operation frequency band of the 5G system achieves the Ka-band. A small wavelength, small beam width, and high atmospheric attenuation are the shortcomings of this frequency band while its great advantages are its larger bandwidth and higher data rate. The multiple antenna techniques (MTA) is the solution that can solve wave shadowing of millimeter wave propagation [24]. The array antenna is an important development. The array antenna is composed of antennas that are arranged periodically as illustrated in Figure 4. The beam main lobe can be tilted by changing the phase of the antennas, which is called the beam steering technique.

Array Antenna Design
The operation frequency band of the 5G system achieves the Ka-band. A small wavelength, small beam width, and high atmospheric attenuation are the shortcomings of this frequency band while its great advantages are its larger bandwidth and higher data rate. The multiple antenna techniques (MTA) is the solution that can solve wave shadowing of millimeter wave propagation [24]. The array antenna is an important development. The array antenna is composed of antennas that are arranged periodically as illustrated in Figure 4. The beam main lobe can be tilted by changing the phase of the antennas, which is called the beam steering technique. In this study, the four patch antennas constituted a 2 × 2 array antenna as shown in Figure 5. An AnokiWave phasor IC was set at the same side with the patch antennas. Such an arrangement makes the array antenna become a complete system. This modular system is more flexible and expandable, which is widely known as the Antenna-in-Package (AIP).   In this study, the four patch antennas constituted a 2 × 2 array antenna as shown in Figure 5. An AnokiWave phasor IC was set at the same side with the patch antennas. Such an arrangement makes the array antenna become a complete system. This modular system is more flexible and expandable, which is widely known as the Antenna-in-Package (AIP).

Array Antenna Design
The operation frequency band of the 5G system achieves the Ka-band. A small wavelength, small beam width, and high atmospheric attenuation are the shortcomings of this frequency band while its great advantages are its larger bandwidth and higher data rate. The multiple antenna techniques (MTA) is the solution that can solve wave shadowing of millimeter wave propagation [24]. The array antenna is an important development. The array antenna is composed of antennas that are arranged periodically as illustrated in Figure 4. The beam main lobe can be tilted by changing the phase of the antennas, which is called the beam steering technique. In this study, the four patch antennas constituted a 2 × 2 array antenna as shown in Figure 5. An AnokiWave phasor IC was set at the same side with the patch antennas. Such an arrangement makes the array antenna become a complete system. This modular system is more flexible and expandable, which is widely known as the Antenna-in-Package (AIP).  The antenna spacing d is an important parameter in the design of the array antenna. In Figure 6, the ideal maximum array directivity (D) of a 2 × 2 array antenna is 6 dBi [25]. Basically, the single antenna gain (G) as shown in Equation (4) is proportional to the directivity of a single antenna. In fact, the antenna efficiency of each element does not need to be considered when taking into account the array gain. The array gain is equal to the array directivity. In this study, the estimated array gain is 5-6 dBi. Otherwise, the maximum scan angle must satisfy the condition in Equation (5). The θ max is the maximum angle to which the array can be steered. The steering can be reckoned by Equation (5). The maximum angle is listed in Table 1 with an operating frequency of 28 GHz.
Micromachines 2020, 11, x FOR PEER REVIEW 5 of 15 The antenna spacing d is an important parameter in the design of the array antenna. In Figure 6, the ideal maximum array directivity (D) of a 2 × 2 array antenna is 6 dBi [25]. Basically, the single antenna gain (G) as shown in Equation (4) is proportional to the directivity of a single antenna. In fact, the antenna efficiency of each element does not need to be considered when taking into account the array gain. The array gain is equal to the array directivity. In this study, the estimated array gain is 5-6 dBi. Otherwise, the maximum scan angle must satisfy the condition in Equation (5). The is the maximum angle to which the array can be steered. The steering can be reckoned by Equation (5). The maximum angle is listed in Table 1 with an operating frequency of 28 GHz.

= ξ •
(4) The ideally maximum steering of the array antenna was ±90°. With that the antenna spacing was half the wavelength. In this study, the minimum spacing was 9.4 mm since the phasor IC was set at the center of the proposed array antenna. The maximum steering of the proposed array antenna approached ±10°. The measured return loss of the simulated 2 × 2 array antenna of each port was better than 10 dB at an operating frequency of 26.5-29.5 GHz as shown Figure 7. The simulation results of each port were highly consistent, which is due to the structure of the array antenna that is in symmetry. The antenna peak gain was 14.4 dBi as shown through m1 in Figures 8 and 9. The 3 dB beam width that is shown through m2 and m3 on Figures 8 and 9, respectively, was 26°. The comparison of the simulation results of the single antenna and the array antenna is shown in Figure 10. The array gain was 5.78 dB, which was consistent with the estimative value. The ideally maximum steering of the array antenna was ±90 • . With that the antenna spacing was half the wavelength. In this study, the minimum spacing was 9.4 mm since the phasor IC was set at the center of the proposed array antenna. The maximum steering of the proposed array antenna approached ±10 • . The measured return loss of the simulated 2 × 2 array antenna of each port was better than 10 dB at an operating frequency of 26.5-29.5 GHz as shown Figure 7. The simulation results of each port were highly consistent, which is due to the structure of the array antenna that is in symmetry. The antenna peak gain was 14.4 dBi as shown through m1 in Figures 8 and 9. The 3 dB beam width that is shown through m2 and m3 on Figures 8 and 9, respectively, was 26 • . The comparison of the simulation results of the single antenna and the array antenna is shown in Figure 10. The array gain was 5.78 dB, which was consistent with the estimative value.              The filed pattern of beam steering can be simulated by changing the phase of the four patch antennas. The simulation results of the beam steering tilted angle at 28 GHz are shown in Figures 11  and 12. The maximum gain was 14.4 dBi for both X cut and Y cut. The beam steering tilted angle was ±34° in the X cut while the beam steering tilted angle was ±26° in the Y cut.

Antenna Manufacturing and Experimental Measurement
Progressive and lower loss materials were used to design a millimeter-wave antenna, such as Rogers (RO 4003C or RO 4350B), low temperature co-fired ceramics (LTCC), PTFE, and liquid crystal polymer (LCP). The manufacturing process of these novel kinds of materials is complex and their manufacturing costs are very expensive. The FR4 substrate has a lower cost compared to the other kinds of materials. The material cost of a Rogers material is three to five times more expensive than that of an FR4 material. Furthermore, the choice of the manufacturer, manufacturing quantity, design metal layers, and ordering options also affect the overall cost of the whole process. On the other hand, a low loss material process is 100 times more expensive than the manufacturing cost of a traditional FR4 PCB. However, the loss tangent of the low cost FR4 material is 0.01-0.04 at a frequency band of 26.5-29.5GHz, which restricts the performance of the antenna. The gain of the antenna that is designed on an FR4 substrate is approximately 4.5 dBi. The performance of the antenna that is designed on an FR4 substrate can be enhanced by using the air-filled cavity structure.
The antenna module proposed in this study was designed with a stack of three substrates and four metal layers (M1, M2, M3, and M4) as illustrated in Figure 13. The production process started by completing the circuit etching of the middle layer (M2 and M3) followed by the addition of two layers of PP (PP_1 and PP_2) on top and below the middle layer. During this step, the upper and lower materials of M2 (PP_1 and FR4_2) were laser precut as shown in the figure. The purpose of the laser precut is to leave a cutting path that will be used as a guide for the removal of the center substrate area later in the process. The next step was the lamination of M1 and M4, and the circuit etching for both metal layers. This was followed by creating laser holes from M1 to M2 and M3 to M4, and finally from M1 to M4. The final step involved mechanical drilling at the M4 surface towards the laser precut. Once the holes from the M4 surface to the laser precut were properly drilled and aligned, the center substrate could be removed therefore exposing the area of the entire air cavity.
The key point of the process technology is on the air-filled cavity structure. The tolerance of each airfilled cavity must be made as small as possible. If the tolerance turned out to be significantly large, it will lead to a significant difference in the gain of each patch antenna. In turn, the performance of the array will be affected. In addition, the reserved M1 layer and its supporting material FR4_1 must be designed to be thin in order to have a lossless air-filled cavity. Moreover, if the air-filled cavity is too large in terms of area, it will have an impact on the antenna gain due to the changed distance of the patch relative to the ground.

Antenna Manufacturing and Experimental Measurement
Progressive and lower loss materials were used to design a millimeter-wave antenna, such as Rogers (RO 4003C or RO 4350B), low temperature co-fired ceramics (LTCC), PTFE, and liquid crystal polymer (LCP). The manufacturing process of these novel kinds of materials is complex and their manufacturing costs are very expensive. The FR4 substrate has a lower cost compared to the other kinds of materials. The material cost of a Rogers material is three to five times more expensive than that of an FR4 material. Furthermore, the choice of the manufacturer, manufacturing quantity, design metal layers, and ordering options also affect the overall cost of the whole process. On the other hand, a low loss material process is 100 times more expensive than the manufacturing cost of a traditional FR4 PCB. However, the loss tangent of the low cost FR4 material is 0.01-0.04 at a frequency band of 26.5-29.5GHz, which restricts the performance of the antenna. The gain of the antenna that is designed on an FR4 substrate is approximately 4.5 dBi. The performance of the antenna that is designed on an FR4 substrate can be enhanced by using the air-filled cavity structure.
The antenna module proposed in this study was designed with a stack of three substrates and four metal layers (M1, M2, M3, and M4) as illustrated in Figure 13. The production process started by completing the circuit etching of the middle layer (M2 and M3) followed by the addition of two layers of PP (PP_1 and PP_2) on top and below the middle layer. During this step, the upper and lower materials of M2 (PP_1 and FR4_2) were laser precut as shown in the figure. The purpose of the laser precut is to leave a cutting path that will be used as a guide for the removal of the center substrate area later in the process. The next step was the lamination of M1 and M4, and the circuit etching for both metal layers. This was followed by creating laser holes from M1 to M2 and M3 to M4, and finally from M1 to M4. The final step involved mechanical drilling at the M4 surface towards the laser precut. Once the holes from the M4 surface to the laser precut were properly drilled and aligned, the center substrate could be removed therefore exposing the area of the entire air cavity. The key point of the process technology is on the air-filled cavity structure. The tolerance of each air-filled cavity must be made as small as possible. If the tolerance turned out to be significantly large, it will lead to a significant difference in the gain of each patch antenna. In turn, the performance of the array will be affected. In addition, the reserved M1 layer and its supporting material FR4_1 must be designed to be thin in order to have a lossless air-filled cavity. Moreover, if the air-filled cavity is too large in terms of area, it will have an impact on the antenna gain due to the changed distance of the patch relative to the ground. The proposed array antenna was manufactured on an FR4 substrate. Figure 14 shows a photograph of the array antenna assembly. The measured results of the return loss for each port were better than 10 dB at an operating frequency band of 26.5-29.5 GHz. The comparison of the simulation and empirical results are presented in Figure 15. The empirical results are shown to satisfy the requirement of a 5G system millimeter wave band. The proposed array antenna was manufactured on an FR4 substrate. Figure 14 shows a photograph of the array antenna assembly. The measured results of the return loss for each port were better than 10 dB at an operating frequency band of 26.5-29.5 GHz. The comparison of the simulation and empirical results are presented in Figure 15. The empirical results are shown to satisfy the requirement of a 5G system millimeter wave band. The proposed array antenna was manufactured on an FR4 substrate. Figure 14 shows a photograph of the array antenna assembly. The measured results of the return loss for each port were better than 10 dB at an operating frequency band of 26.5-29.5 GHz. The comparison of the simulation and empirical results are presented in Figure 15. The empirical results are shown to satisfy the requirement of a 5G system millimeter wave band. The proposed array antenna was manufactured on an FR4 substrate. Figure 14 shows a photograph of the array antenna assembly. The measured results of the return loss for each port were better than 10 dB at an operating frequency band of 26.5-29.5 GHz. The comparison of the simulation and empirical results are presented in Figure 15. The empirical results are shown to satisfy the requirement of a 5G system millimeter wave band.   Figure 16 shows an NSI-700S-360 antenna chamber [26]. Its measurement coordinates are shown in Figure 17. The gain measurement results of each patch antenna are shown in Figure 18 (X-cut) and    Figure 16 shows an NSI-700S-360 antenna chamber [26]. Its measurement coordinates are shown in Figure 17. The gain measurement results of each patch antenna are shown in Figure 18 (X-cut) and   Figure 16 shows an NSI-700S-360 antenna chamber [26]. Its measurement coordinates are shown in Figure 17. The gain measurement results of each patch antenna are shown in Figure 18 (X-cut) and

Conclusions
The design and simulation of a 2 × 2 low cost phase array antenna module for 5G applications operating at 28 GHz with 14.4 dBi antenna gain was proposed in this paper. The air-filled cavity used for patch antenna structure was with a FR4 PCB material for cost reduction instead of using a Roger or M6 material PCB. Moreover, it improved the antenna radiation efficiency by reducing the loss of the material. Furthermore, the designed array unit could be used and combined for a higher order array along two dimensions with a suitable surface mount technology (SMT) gap. It helps to easily and reliably implement a high order array. Therefore, the proposed array antenna is a promising candidate for the mm-wave 5G small cell applications. Table 2 summarizes the performance of this work and compares it with state-of-the-art mm-wave phased-array antennas [27][28][29][30][31][32][33][34][35][36][37][38]. The proposed patch shows around an 8.5 dBi antenna gain, which is better than [31,35,37,38], at a similar frequency. It describes that the air-filled cavity as a patch gap between the ground increased the antenna efficiency effectively instead of a lossy FR4 PCB material. The measured results of the single array unit show that the maximum radiation direction can be steered from -34 to +34° continuously in the X-cut and -26 to +26° continuously in the Y-cut at 28 GHz. The total dimension of the resulting design package was 18 mm × 14 mm × 0.71 mm. The gain of the array antenna achieved 14.4 dBi and the reflection coefficient of the array antenna was less than −10 dB from 26.5 to 29.5 GHz.

Conclusions
The design and simulation of a 2 × 2 low cost phase array antenna module for 5G applications operating at 28 GHz with 14.4 dBi antenna gain was proposed in this paper. The air-filled cavity used for patch antenna structure was with a FR4 PCB material for cost reduction instead of using a Roger or M6 material PCB. Moreover, it improved the antenna radiation efficiency by reducing the loss of the material. Furthermore, the designed array unit could be used and combined for a higher order array along two dimensions with a suitable surface mount technology (SMT) gap. It helps to easily and reliably implement a high order array. Therefore, the proposed array antenna is a promising candidate for the mm-wave 5G small cell applications. Table 2 summarizes the performance of this work and compares it with state-of-the-art mm-wave phased-array antennas [27][28][29][30][31][32][33][34][35][36][37][38]. The proposed patch shows around an 8.5 dBi antenna gain, which is better than [31,35,37,38], at a similar frequency. It describes that the air-filled cavity as a patch gap between the ground increased the antenna efficiency effectively instead of a lossy FR4 PCB material. The measured results of the single array unit show that the maximum radiation direction can be steered from -34 to +34° continuously in the X-cut and -26 to +26° continuously in the Y-cut at 28 GHz. The total dimension of the resulting design package was 18 mm × 14 mm × 0.71 mm. The gain of the array antenna achieved 14.4 dBi and the reflection coefficient of the array antenna was less than −10 dB from 26.5 to 29.5 GHz.

Conclusions
The design and simulation of a 2 × 2 low cost phase array antenna module for 5G applications operating at 28 GHz with 14.4 dBi antenna gain was proposed in this paper. The air-filled cavity used for patch antenna structure was with a FR4 PCB material for cost reduction instead of using a Roger or M6 material PCB. Moreover, it improved the antenna radiation efficiency by reducing the loss of the material. Furthermore, the designed array unit could be used and combined for a higher order array along two dimensions with a suitable surface mount technology (SMT) gap. It helps to easily and reliably implement a high order array. Therefore, the proposed array antenna is a promising candidate for the mm-wave 5G small cell applications. Table 2 summarizes the performance of this work and compares it with state-of-the-art mm-wave phased-array antennas [27][28][29][30][31][32][33][34][35][36][37][38]. The proposed patch shows around an 8.5 dBi antenna gain, which is better than [31,35,37,38], at a similar frequency. It describes that the air-filled cavity as a patch gap between the ground increased the antenna efficiency effectively instead of a lossy FR4 PCB material. The measured results of the single array unit show that the maximum radiation direction can be steered from -34 to +34 • continuously in the X-cut and -26 to +26 • continuously in the Y-cut at 28 GHz. The total dimension of the resulting design package was 18 mm × 14 mm × 0.71 mm. The gain of the array antenna achieved 14.4 dBi and the reflection coefficient of the array antenna was less than −10 dB from 26.5 to 29.5 GHz.