5.3.4.1 Substrate Selection and Thickness

The first step of the antenna design is the selection of substrate. The design frequency and operational bandwidth of the antenna govern the selection of substrate. The thickness, relative permittivity, loss tangent, conductive strip thickness, and finally, mechanical and thermal stabilities are the critical parameters of the substrate at 60 GHz frequency band. Considering the 5 mm wavelength at 60 GHz, precision in the design and substrate properties plays critical role in the design success. Taconic TLX-8 would be a suitable candidate as it satisfies many of the aforementioned parameters [26, 27]. Its low dielectric constant enhances the antenna bandwidth [18] and its favorable thickness is required for the design.

The selection of substrate thickness is bounded by two factors. A thin substrate is required for an efficient radiator. However, the large operational bandwidth needs a thick substrate. The substrate thickness should satisfy Equation (5.4) to ensure the suppression of higher order radiation modes on the dielectric surface [28]. At 60 GHz, the thickness of the substrate must be less than 1.06 mm.

5.4

In Equation (5.4), λ₀ is the wavelength and is between 4.68 and 5.26 mm. is the effective permittivity of the substrate. Considering the available commercial substrate, Taconic TLX-8 with 0.127 mm thickness is selected. This thickness adequately satisfies Equation (5.4) and ensures that only one surface wave, TM₀ with zero cutoff frequency, will exist in the substrate [29]. This thin substrate also enhances the CPL of the antenna [30], which is the critical aspect of the design.

5.3.4.2 Substrate Characterization

The knowledge of actual electrical parameters of the substrate is vital for accurate design, especially in the millimeter-wave range, where dimensions based on electrical characteristics such as substrate permittivity and loss tangent are tolerance-critical. Today, most substrate manufacturers specify the dielectric parameters of their materials at 10 GHz. Consequently, in realized designs at millimeter-wave, frequency shifts attributed to permittivity deviation commonly occur [31]. Microwave measurement techniques to determine substrate permittivity and loss tangent at specific frequencies are (i) transmission and reflection and (ii) resonance techniques. Transmission and reflection techniques can be used to obtain the real part of permittivity for a high- or low-loss specimen but lack the resolution to measure low loss tangents. Resonant technique is only suitable for low-loss materials since the resonance curve broadens as the loss increases [32]. Usually, dielectric property measurement by resonance technique provides higher accuracy than that obtained from transmission and reflection technique. It is therefore considered as the most accurate method for dielectric constant and loss tangent measurement of planar structures. Considering the structural dimensions at millimeter-wave, the substrate integrated waveguide (SIW) resonators approach is proposed for substrate characterization at V- and W-bands [33].

Taconic (www.taconic.com, accessed on May 6, 2014) published data sheet of TLX-8 at 10 GHz reveals the following information: ϵ_r = 2.55 and tan δ = 0.0019. In this work, SIW cavity resonators are designed and fabricated at 60 GHz (Figure 5.6). Propagation and attenuation constants of an SIW resonator are related to modal expansion theory of a rectangle waveguide [34]. Equation (5.4) shows the equivalent dimensions of the rectangle waveguide on which a, b, d, and s are defined at Figure 5.6(a). The photograph of a prototype SIW resonator is shown in Figure 5.6(b). The resonance frequency of the rectangle SIW resonator for mode relates to Equation (5.6).

5.5

5.6

In Equation (5.6), c is the speed of light in free space, and are found from Equation (5.5), and is the effective dielectric constant. To find the effective dielectric constant, the proposed approach by Zelenchuk et al. [33] is followed. Six prototypes of SIW resonators are fabricated and tested to measure the substrate parameters at 60 GHz (c.f. Figure 5.6). The average dielectric constant and loss tangent are measured to be 2.25 and 0.01, respectively, which are substantially different than those for published datasheet at 10 GHz. Table 5.3 summarizes the results. The measured data of the substrate are considered in designing the proposed DSPD array antenna.

5.3.5.1 Dipole Design

The DSPD element with its design parameters is shown in Figure 5.7. The length (L_a) of the dipole arm is a function of the thickness and dielectric constant of the substrate and mainly controls the resonance frequency of the dipole. As the substrate does not cover the whole structure, the resonant length is not scaled with as an antenna in a homogeneous medium [29].

The initial value of the dipole length comes from the following formula [35]:

5.7

5.8

In these formulas, v is the velocity of propagation inside the substrate and the rest of parameters are defined in Figure 5.7. Based on Equations (5.7) and (5.8), the required length of dipole arm is 1.2 mm. This initial length is entered in full-wave EM solver CST Microwave Studio and the design is optimized to meet the specification requirement. The simulated dipole arm length is 1 mm. The dipole arm width is independently optimized to match with 50 Ω input impedance although the optimized value of W_a would be different than that for the parallel feedline width, W_f. Matching of these two strip lines with different widths is manipulated through changing the chamfering depth, C_d. Finally, the parallel suspended feedline of the dipole arm is transformed to a microstrip line that is connected to a two-way power divider of the beam-forming network of the four-element array. To enhance the bandwidth and performance of the antenna, chamfering of the dipole element and tapered ground plane at the suspended stripline to microstripline transition are used. This technique yields more than 12% bandwidth at 60 GHz, while conventional design yields 7% bandwidth [20].

5.3.5.2 Dipole Chamfering

Dipole antennas are inherently unbalanced circuits and create distortion in radiation pattern. Connecting dipole to a balanced feedline requires a balun network to match balanced and unbalanced circuits. Chen and Sun [36] suggested a double-sided feed network on the substrate as the balun network for the planar dipole antennas. As shown in Figure 5.7, the double-side structure feedlines provide a smooth transition between the unbalanced and balanced currents of the line and the dipole feed. This configuration also simplifies the balun network. As discussed before, the chamfering of the dipole edge is required to match two different widths of the feedline and dipole arm. Moreover, the right angle in the microstrip line creates extra capacitance and also changes the electrical length of the line [37]. These two will affect the impedance matching of the structure that should be compensated by chamfering. For symmetry, both dipole arms at their right angle transitions are chamfered. However, from our detailed investigation in full-wave CST Microwave Studio simulation, we found that only top arm chamfering provides best results. Figure 5.8 shows the input impedance return loss (S₁₁ in dB) versus frequency for three cases – top arm, both arms, and no chamfering. Considering the effect of different chamfering scenarios on the input impedance, the maximum RL of 27 dB is obtained with top arm chamfering. As the bottom arm is directly connected to the ground, then it probably has minor effect on the impedance matching and only top arm chamfering provides the optimum impedance matching.

5.3.5.3 Ground Connection Tapering

The next investigation to improve the performance of the dipole is a tapered ground connection of the bottom feedline of the balanced feed network of the DSPD antenna. The ground plane acts as a reflector to the dipole, hence largely influences both the radiation pattern and input impedance. Therefore, the shape and distance of the ground plane from the dipole element determines the input impedance bandwidth, beamwidth, gain, and cross-polar level of the dipole element. The authors have investigated the effect of the ground plane on monopole and dipole elements' performance in Refs [38, 39]. We have investigated two types of ground planes – conventional (also called abrupt) and tapered ground planes. The effect of the tapered ground connection depends on its tapering angle. The tapering angle (α) is shown in Figure 5.9. The angle α is changed to find the best performance of the dipole through parameter sweeping. A tapered connection causes a resonance shift in return loss figure. To accurately compare the two scenarios, few parameters of the top arm are retuned to have the same frequency resonances. Figure 5.9 shows the input impedance return loss versus frequency for both abrupt and tapered connections. It confirms that the abrupt connection proposes better impedance matching on the resonance frequency while not affecting the bandwidth of the radiator. This effect was also experienced over the lower frequency range of 20–26 GHz [23].

5.3.5.4 Dipole Radiation Pattern and Cross-Polar Level

The radiation pattern of a single DSPD is simulated with full-wave EM solver CST Microwave Studio and the result is shown in Figure 5.10. The 2D radiation patterns of a DSPD are depicted for three sample frequencies – 58, 60, and 63 GHz in Figure 5.11. As will be discussed later, the final array configuration is a linear array while elements are oriented vertically. Through general knowledge of the array theory, one can conclude that the H-plane of the final array structure is similar to the single DSPD, while its E-plane would be different from that of a single DSPD.

The 3-dB beamwidth of the single DSPD on its H-plane, for three sample frequencies, is shown in Table 5.4. The values of the 3-dB beamwidth adequately satisfy the wide radiation pattern requirements of the final array as discussed in Table 5.2. This wide radiation pattern ensures that in the synthetic aperture approach, the tag's surface receives enough energy from the reader at different view angles. As is evident from the H-plane radiation pattern shown in Figure 5.11, the maximum radiation does not occur on the broadside direction (ϕ = 0) as expected. Instead, the pattern of a single DSPD is shifted to the left direction. This is because of existence of ground plan in one side of the substrate that pushes the radiation in the opposite direction. The direction of maximum radiation depends on the frequency and is approximately shown by the black dashed line in Figure 5.11; however, it varies from −345° to −350°.

As discussed earlier, the quality of the antenna polarization known as its CPL is a critical parameter in the proposed application. For the designed single DSPD, the CPL is simulated by having co- and cross-polar probes in CST Microwave studio and measuring the S₂₁ of the link. The simulated CPL for a single DSPD is depicted in Figure 5.12. Over the entire frequency range of operation, this ratio for a single dipole is less than −23 dB. There is the opportunity of decreasing the CPL by simply using the defected ground structure (DGS) [29]. However, the provided −23 dB CPL through the DSPD adequately satisfies the requirements of the proposed application.

5.3.6.1 Required Beam-Forming Array

The antenna specifications discussed earlier govern the array structure. As a fan-shaped radiation pattern, wide azimuth and limited elevation, is required hence a linear array-type antenna would satisfy this specification. Multiple vertically aligned dipoles restrict the radiation pattern in the elevation angle. It was calculated that four dipoles are enough to satisfy the 15–20° required beamwidth in the elevation angle. The wide H-plane radiation pattern of a single dipole as shown in Figure 5.11 also fulfils the 115° required azimuth beamwidth. To select a proper feeding network, the overall system structure for the spatial-based chipless RFID application will be considered. As the SAR-based signal processing is based on the relative position of the reader and tag, it is mandatory that the antenna has a unique and fixed radiation pattern in all positions. This is the reason why a phased array antenna type would not be suitable for this application as it creates various beam shapes on different angles [40]. From the array theory, it is clear that the uniform excitation of the array's elements creates the maximum radiation intensity and minimum 3-dB beamwidth in the boreside direction [41]. The requirement for a fixed fan-shaped radiation pattern suggests a linear array of dipoles with uniform excitation.

5.3.6.2 Corporate Feeding Network

The requirement for a uniform excitation of the array elements and also having two separate Tx and Rx antennas at the reader side ease the power divider arrangement. A T-junction power divider as one of the simplest type of power dividers [42] would be suitable for the proposed application. It is fully printable and of low cost without any lumped element in the structure. These specifications make the T-junction power divider very suitable for feeding the DSPDs on the array. However, the T-junction power divider suffers from two major limitations. First, it utilizes quarter wavelength transformer in its structure that restricts the operational bandwidth of the element. Second, it provides low isolation among output ports. The low isolation would be a major limitation for monostatic systems on which the same antenna is used for transmit and receive purposes. While in the proposed bistatic configuration, two separate antennas are utilized; therefore, the low isolation of the output is not an issue. The low bandwidth of the power divider, due to the quarter wavelength transformer, can be compensated by minor changes in the component design. Figure 5.13 shows a conventional two-way T-junction power divider and its dimensions. The length of the quarter wavelength transformer is fixed and it is intended to transform 50 Ω impedance to the 100 Ω. The length of the 50 Ω arm on each side of the power divider depends on the required physical distance between two connected dipoles. The chamfering depth is optimized through parameter sweep in the CST simulation tools.

It is normally accepted that for providing better impedance matching, insertion of a small V-notch in the dividing point of the T-junction structure is a good technique. This gap is shown in Figure 5.14(a). In the current work, however, it was found that a reactive stub acts better and can enhance the power divider bandwidth.

The corporate feed network of the proposed array antenna includes three two-way T-junction power dividers as shown in Figure 5.15. The S-parameter of the corporate feed network is depicted in Figure 5.16 that shows that S₁₁ in decibel versus frequency is adequately below −15 dB over the whole frequency range of operation. The minimum value of the reflection coefficient is −20.5 dB at 61.5 GHz. This is a noticeable result considering the narrowband aspect of the quarter wavelength sections in the power divider structure. Normally, four transmission coefficient curves are expected in Figure 5.16. However, as the transmission coefficients of each port (S_jI where j = 1, 2, 3, 4) are similar, only two of them are depicted as an example.

The values of the transmission coefficient vary between 6.4 and 6.55 dB over the band. While the reflection is adequately low, almost a quarter of the input power is delivered to each port. Therefore, the array elements are uniformly excited. Figure 5.17 shows the changes of the transmission coefficient phase on four different output ports. The phases of ports 1 and 2 are exactly equal. The same is also valid for the phases of ports 3 and 4 in Figure 5.15. There is a small difference between the phases of ports 1 and 2, and 3 and 4 as shown in Figure 5.17. As the difference is very small, 2–3°, it is not a serious issue in the total radiated power.

5.3.6.3 Mutual Coupling

Mutual coupling between the array's elements is a critical factor in the array design that can negatively affect the total performance of the array. Increasing the interelement distance reduces the mutual coupling; however, a large distance between dipoles results in grating lobes in the radiation pattern if the inner-element distance exceeds the wavelength, λ.

While the half-wavelength distance is normally selected as the inner-element space, larger distances (λ/2 < d < λ) are also feasible if narrower HPBW is desirable. In the proposed application, smaller HPBW on the elevation angle is desired as it focuses more energy on the tag surface and minimizes the reflections from surrounding objects. Therefore, the power divider's arm is tuned to provide 3.45 mm inner-element spacing equivalent to 0.65–0.73λ over the band (Figure 5.18). This large distance between the elements reduces the HPBW, as well as minimizes the mutual coupling.

To simulate the isolation between two adjacent dipoles, two ports are connected to the dipoles and the transmission coefficient, S₂₁, is simulated in CST. Figure 5.18 shows the simulated result. The mutual coupling between two adjacent dipoles is adequately below −22 dB. It is important to verify that the low value of S₂₁ is adequately showing the mutual coupling and it is not due to the high loss occurred in the feedline of the dipoles. This simply means that if the feedline causes a noticeable amount of loss, then a good S₂₁ is resulted. Hence, loss occurring on the feedline can be seen as an error on the transmission coefficient measurement. This feedline, which may degrade the S₂₁ result, is highlighted in Figure 5.18. The feedline length is only 0.8 mm (0.16λ) and therefore the occurred loss is very minor. By using the CST and/or advanced design system (ADS) facilities, it can easily be seen that the feedline loss only contributes 0.17–0.23 dB excess loss to the result, S₂₁ of Figure 5.18. This means that the isolation between adjacent elements is below −22 dB over the whole frequency band of operation.

5.3.6.4 Extended Feedline

The designed array of four dipoles and its feeding network has acceptable electrical performance over the frequency range of operation. However, four-element dipole array is too small that is impractical to connect the array with a V-type end launch connector as shown in Figure 5.19. A V-type connector that is suitable for the frequency range of operation has a much larger size than the designed antenna. To address this issue, the array's feedline is extended for 4λ ≈ 20 mm to provide (i) enough space for connector installation and (ii) minimize the cross-polar component.

Providing enough space for connector installation, it is possible to simply extend the microstrip line; however, induced surface wave results in higher feedline loss. To reduce the effect of long feedline on the antenna performance, the CPW extension is used instead of microstrip line. The grounded CPW line is immune to surface wave and also it provides better resistivity to the copper roughness than the normal microstrip line [43]. The prototype array is fabricated on TLX-8 using photolithographic fabrication process. The photograph of the assembled array antenna with the extended CPW feedline and the V-type connector [44] assembly is shown in Figure 5.20.

As it is clear from Figure 5.20, the length of the extended feedline seems to be longer than what is needed for the purpose of V-type connection. The feedline length is bound by two contradictory factors. First, the shorter line is preferred to minimize the loss occurring on the line and avoid unwanted radiation from the feedline. This implies that a 5–7 mm extension is enough. The increased CPL due to the back-lobe reflection by the large connector body is another affecting parameter. As shown in Figure 5.11, the dipoles have a noticeable back-lobe. If the array antenna is placed very close to the connector, its back-lobe energy is reflected back on the broadside direction by the connector structure. This may disturb the radiation pattern of the array antenna. However, the major issue would be related to the increased level of the array CPL. To minimize this effect, the length of the feedline is extended beyond what was mandated by the physical requirements for connection purposes. As a trade-off selection for the feedline length, 20 mm extension is selected for the length of the feedline. This provides enough distance between the connector and the array for minimum disturbance of the radiation pattern and degradation of the CPL. However, the loss occurring on the line will also be considered.

5.3.7.1 Reflection Coefficient Measurement

The designed array of four DSPDs is simulated using CST Microwave Studio. However, the V-type connector may have significant effect on the final performance of the structure due to its large size compared with array antenna. Therefore, the connector is also included in the simulation process as shown in Figure 5.21. The dimensions of the V-type connector are derived from its data sheet [45]; however, aluminum (lossy metal) is assumed for the connector body and silver for its central pin. The fabricated antenna through the photolithographic process shown in Figure 5.20 is also connected to the Agilent PNA for its return loss measurement after proper system calibration.

The simulated and the measured S₁₁ in decibel versus frequency are presented in Figure 5.21. It is important to reemphasize that the simulated and measured return loss reflects the effects of the array antenna, its long feedline, microstrip-to-CPW transition, and the V-type connector assembly. Three main aspects of the graphs in Figure 5.21 can be categorized as follows:

1. i. Satisfactory Return Loss. As shown in Figure 5.21, both simulation and measured results yield more than 10 dB return loss over the frequency band of operation with maximum RL of more than 30 dB at 63 GHz. This ensures the satisfactory performance of the fabricated array over the band 57–64 GHz. It also confirms that the effects of a long feedline and the connector as well as the transition between microstrip to CPW have been appropriately taken into account.
2. ii. Minor Simulation and Measurement Mismatch. The graphs in Figure 5.21 are not fully matched to each other. Regarding the high frequency range of operation and also the fabrication process, some minor mismatching would be normal. To have a better understanding of the problem, the fabricated array is investigated thoroughly with a digital microscope. The dimensions of the array in different sections are measured and compared with the original values. The percentage of errors in various parts of the array is shown in Figure 5.22. As one can see from Figure 5.22, errors in different parts of the structure have significantly different values. While the width of the design is more vulnerable to errors, the lengths of the various parts are more immune to fabrication errors. For example, while some widths are experiencing 11% error, the maximum error in the array lengths is found to be only 0.4%. Inaccuracy in the line width significantly affects the impedance matching as it changes the line impedance. The effect of errors in line lengths are discussed when the dielectric characterization is being considered. In addition to the errors occurring in the line widths, which result in impedance mismatch, one may suspect that residual surface waves, in the extended via grounded CPW line, and V-type connector assembly, cause impedance mismatch between the simulation and measured results. However, more than 10 dB return loss over the frequency band ensures no significant mismatch in the antenna and meets the specifications as discussed earlier.
3. iii. Accurate Dielectric Characterization. The resonance frequency of the array is mainly controlled by the length of the dipole arm and the dielectric constant. As shown in Figure 5.21, the simulated resonance frequency and the measured resonance are both concentrated around 63 GHz. This means that the simulation and measured results are adequately matched from the resonance frequency perspective. As was already explored in (ii), the lengths of the dipole's arms are not significantly affected by the fabrication tolerance. Therefore, having the same simulated and measured resonance frequencies confirms the accuracy of the dielectric characterization method as described earlier. In other words, the calculated mean value of the measured ϵ_r for Taconic TLX-8 at 60 GHz provides a good prediction of the substrate performance at the 60 GHz frequency band.

5.3.7.2 Antenna Fabrication Error

At 60 GHz, the design would be very sensitive to any error during fabrication process. The effect of fabrication error was shown in the previous section, where inaccurate widths in the array feed network and dipole arms resulted in mismatch between the measured and simulated reflection coefficients. At least two array antennas are required in the proposed technique; therefore, it is important to consider the design robustness to dimensional tolerances as process variation in photolithographic etching always exists. It was discussed earlier that the T-junction power divider provides higher bandwidths when a reactive stub is utilized compared to a conventional V-notch in its structure. Moreover, it was found that the reactive stub significantly influences the design robustness to process errors.

Based on this finding, five samples of fabricated array antennas as shown in Figure 5.23 are measured by a digital microscope. The values of errors in different sections of the array structure are summarized in Table 5.5. As discussed earlier, the width of lines in the printed circuit is more vulnerable to process error than the line length.

The measured input impedance, return loss (S₁₁ in dB) versus frequency for the five fabricated array samples and the simulated S₁₁ based on the accurate design values are shown in Figure 5.24. While the measured return losses are not completely matched with the simulation result due to the fabrication error, all five samples have acceptable performance (S₁₁ > 10 dB) in the frequency range 57–64 GHz. Moreover, one may notice that all of the fabricated samples acquire almost the same resonance frequency as suggested through simulation, around 62.8–63.2 GHz. It can also be concluded that irrespective of fabrication error through conventional photolithographic printed circuit board (PCB) fabrication technique, appropriate antenna design at 60 GHz millimeter-band is feasible.

5.3.7.3 Radiation Pattern Measurement

After satisfactory input impedance RL measurement, the radiation patterns of the fabricated array are measured. The array antenna under test (AUT) is set as a receiver antenna and a horn antenna [46], manufactured by Ainfoinc, is used as a standard gain transmitting antenna. Agilent PNA E8361A in its receiver mode is used. Figure 5.25 shows the photograph of the test setup with the standard gain horn antenna mounted on a fixed Perspex pole and the AUT mounted on a turn table. The radiation pattern is measured in every 10° steps for full 360° for both E- and H-plane cuts. Some conical absorbers are placed around the antennas to reduce surrounding interferences. Both measured and simulation copolar and cross-polar radiation patterns at 58, 60, and 63 GHz are shown in Figure 5.26. The measured radiation patterns are in congruence with the simulation results with minor shifts in sidelobe levels.

The measured cross-polar result is shown in Figure 5.26. As is clear from Figure 5.26, the CPL is adequately below −20 dB. This is a very satisfactory result considering the effects of the long feedline and the reflection from the large V-type connector.

As it is clear from Figure 5.26, the E-plane pattern is narrow compared with H-plane as the array is vertically aligned. Therefore, the 10° measurement would not be sufficient for the E-plane and to accurately measure the HPBW of the antenna in E-plane, measurements are repeated for every 1° for the ±10° of the boresight direction as shown in Figure 5.27. By considering the manual rotation of AUT in every 1°, small errors may be expected. To minimize this error, the measurement is repeated three times and the average value is considered. Figure 5.27 shows the copolar radiation pattern in ±10° of boresight direction of the array antenna at 58 GHz. It shows that the measured HPBW is smaller compared with the simulated HPBW. Table 5.6 summarizes the HPBW of the array that shows that, for all cases, the measured HPBW are smaller than the simulation result.

5.3.7.4 Gain Measurement

The measurement of the array gain is a combination of measurement, simulation, and calculation process. The standard horn antenna is used as transmit antenna and AUT receives the signal, while the Agilent PNA measures the S₂₁ of the link. The gain of the array relates to other parameters by

5.9

where G_Tx is the gain of horn antenna, which is 10.6 dB_i. L_cable is the cable loss and can be measured directly with the Agilent PNA E8361A using full two-port error correction calibration. In measuring L_cable, the loss of connectors and adaptors are also considered. Free space loss or path loss, LFS, is calculated using Friis transmission formula [41]. The feedline loss should be considered as the CPW feed extension is 4λ long and contributes significantly to the gain result. This loss has been separately simulated in CST and shown in Figure 5.28. The feedline is primarily designed at 60.5 GHz, which is the center frequency of 57–64 GHz. Therefore, the feedline shows the minimum insertion loss at 60.5 GHz and more loss in other frequencies of the band. The 1.85 mm Southwest Microwave connector has a VSWR of 1.25 [44] means that the impedance mismatch of the connector should not have significant effect on the gain measurement. Therefore, the connector's performance is not excluded from the gain measurement. By considering Equation (5.9) and by combining the measured data with the LFS and simulated L_feedline, the final value of the gain versus frequency is shown in Figure 5.28. As can be seen from the figure, the feedline loss and antenna gain are congruent to each other over the frequency band. There is about 1.5 dB gain drop from 7 dB_i due to 1.5 dB insertion loss variation of the feedline. The absolute gain of the array antenna varies from 5.5 to 7 dB_i over the frequency band of operation.