I. INTRODUCTION
Microwave power amplifiers (PA) play a crucial role in modern communication systems, including wireless communications. However, under increased operation frequency, the traditional microstrip PA employed in conventional microstrip line technology becomes ineffective due to the inherent high radiation loss of the microstrip lines, leading to poor performance of the PA [1]. In this context, substrate integrated waveguide (SIW) technology has proven promising for the design of components, such as filters, power combiners, or diplexers, operating in the microwave and higher frequency ranges, such as the millimeter-wave region and the sub-THz region [2–5]. SIWs exhibit low loss in very high-frequency regions, good power capability, and high thermal conductivity [6, 7], which are critical features for PA design. However, limited research on SIW MPA has been reported in the literature [8–12]. Among the few conducted, Abdolhamidi and Shahabadi [8] introduced an X-band SIW amplifier employing iris-type inductive discontinuities to fabricate the impedance matching networks of the amplifier. Notably, iris-type inductive discontinuities resemble the double-stub matching network used in traditional matching methods. However, this PA was evaluated only for small-signal performance or S-parameters, whereas important large-signal parameters of PA, such as output power, efficiency, and linearity, were not tested. Using a similar impedance matching method, Wang and Park [9] presented a high-power SIW amplifier operating at 2.14 GHz using a 10-W GaN HEMT device. Here, inductive metalized posts were introduced within the SIW transmission line to realize impedance matching functions. However, the theory behind the operation of the impedance matching network was not clearly explained in the paper. Moreover, this PA worked at a low frequency of 2.14 GHz, and its size was relatively large.
Recently, a new impedance matching method based on filter synthesis techniques, such as the coupling matrix method, has been proposed. For instance, Gao et al. [10] presented an X-band SIW amplifier design using the active coupling matrix technique. Notably, filter synthesis techniques help increase the bandwidth of matching networks. Nevertheless, since the abovementioned method was applied to design a low-noise amplifier, only small-signal performance was evaluated in the paper. In addition, the size of the entire amplifier was quite large. Meanwhile, Pech et al. [11] introduced an SIW PA operating in the X-band by employing the passive coupling matrix technique to design matching circuits that provide a filtering response. The theory behind using a coupling matrix to design a filtering matching circuit is clearly explained in the paper. In addition, the researchers experimentally evaluated the large-signal performance of the PA, such as its output power and power gain efficiency. However, small-signal performance and linearity were not tested, and the size of the entire PA was still large. Furthermore, Yang et al. [12] introduced a class-F SIW PA based on the harmonic termination technique using an SIW stub. The use of harmonic termination significantly improved the efficiency of the proposed PA, which achieved a drain efficiency of 77%. However, the PA was able to work only at a relatively low frequency of 2.45 GHz. Moreover, its linearity and small-signal performance were not tested. Overall, it is evident that most published SIW amplifiers are either small-signal amplifiers, fit only for low-frequency regions, considerably large in size, or lacking in complete performance evaluation.
In this paper, a novel tri-band SIW PA is proposed. An impedance matching method based on the passive coupling matrix technique, as presented in [11], is employed to realize the input and output matching networks. The novelty of the current work lies in its use of the evanescent mode, where the operating frequency of the filter is below the cut-off frequency of the transmission lines, to significantly reduce the sizes of the proposed tri-band PA’s components, such as triplexers and matching circuits. Furthermore, to ensure the feasibility of the proposed method, a complementary split ring resonator (CSRR) is etched onto the SIW structures to attain a negative reflective index. Consequently, a tri-band PA featuring a compact SIW-CSRR triplexer and three single-stage SIW-CSRR PAs is proposed. The PA is experimentally tested for both small-signal and large-signal performance, including linearity in terms of third-order inter-modulation (IMD3) and adjacent channel power ratio (ACPR). Notably, the proposed PA works concurrently at three operating frequencies—6 GHz, 8 GHz, and 10 GHz. More importantly, its size is significantly reduced by employing evanescent mode operation.
The rest of the paper is organized as follows: Section II introduces the proposed SIW-CSRR technology, Section III presents the design of the proposed single-band PA, Section IV presents the design of the proposed tri-band PA, and Section V concludes the paper.
II. SIW-CSRR TECHNOLOGY
Fig. 1 presents a block diagram of the proposed tri-band MPA. The triplexers incorporated at the input and output help separate the signal at each frequency band at the input and combine signals at the output, allowing for independent optimization of performance at each band. They also contribute to removing undesired output spectrum caused by the cross-intermodulation effect when the MPA operates concurrently on multiple bands [13–15]. In this context, it is important to note that critical components of the proposed PA, such as triplexers and single-stage PAs, were implemented using SIW-CSRR technology based on evanescent mode operation. SIW-CSRR refers to a combination of SIW transmission lines and CSRR, which are etched on the surface of SIW transmission lines. This technique was presented in [16], where it was applied to design filters. It has since been proven that this technique can be employed to create a passband response below the cut-off frequency of an SIW transmission line, thus reducing the circuit size [17–20].
Fig. 2 illustrates a unit cell of the SIW-CSRR structure. The unit cells are configured face-to-face to ensure low loss and sharp selectivity, as detailed in [16].
To employ evanescent mode, the cut-off frequency of the SIW section must be higher than the passband or the resonant frequency of the SIW-CSRR filter. This can be achieved by calculating the width of the SIW transmission line and adjusting the CSRR dimensions—A, F, D, and W, as shown in Fig. 2—that determine the CSRR’s resonant frequency [21]. The cut-off frequency of the fundamental mode, TE10, in the SIW transmission line can be expressed as follows [22]:
where c0 is the speed of light in vacuum, εr refers to the relative reflective index, w is the width of the SIW transmission line, d indicates the diameter of the via hole, and p refers to the spacing between adjacent via holes of the SIW transmission line. The resonant frequency of the CSRR can be determined by the following equation [21]:
where Lr and Cr denote the equivalent inductance and capacitance of the CSRR, respectively, determined based on the dimensions of the CSRR. The relationship between Lr and Cr and the dimensions of the CSRR are presented in detail in [21]. Overall, w and the main dimensions of the CSRR—A, F, W, and D—define evanescent mode operation.
Since the PA proposed in the present study was expected to work at the three frequency bands of 6 GHz, 8 GHz, and 10 GHz, it was necessary to produce three passbands in the CSRR below three cut-off frequencies of the SIW transmission lines. In other words, three different SIW-CSRR unit cells were developed for evanescent mode operation. Consequently, these SIW-CSRR cells were employed to design the proposed PA’s components, such as the triplexer and the single-stage PA.
Fig. 3 presents the layout of the SIW transmission line and its detailed dimensions, which were calculated at each frequency band using Eq. (1). Here, it must be noted that only WSIW was adjusted to realize the corresponding cut-off frequency at each band, with the other dimensions kept constant.
The simulated frequency responses of the designed SIW transmission lines are presented in Fig. 4, showing that each transmission line has its own cut-off frequency—8.8 GHz, 10.7 GHz, and 12.2 GHz—which was then employed for the filters centered at 6 GHz, 8 GHz, and 10 GHz, respectively. Notably, two tapers were added at the two ends of the SIW transmission lines to realize impedance matching between the SIW waveguide and the standard 50 Ω transmission line.
After accurately designing the SIW waveguide, the SIW-CSRR filter was implemented. The SIW-CSRR filter was realized by simply excavating the CSRR on the top layer of the SIW waveguide, as shown in Fig. 2, resulting in an SIW-CSRR unit cell or filter. The center frequency of the filter was primarily determined by the dimensions of the CSRR, including A, G, F, and W. The values of these dimensions were calculated based on equations pertaining to the relationship between resonant frequency and CSRR dimensions, as presented in [21]. The other dimensions, such as Lm, Lt1, Wt1, T, Wt2, Lt2, and Wm, were then determined by conducting an electromagnetic (EM) simulation to obtain the desired S-parameters, including S11 and S22. Notably, these procedures were conducted using a full-wave CST simulator. Fig. 5 depicts the simulated S11 and S22 of the designed SIW-CSRR filters, confirming that the center frequencies of each SIW-CSRR filter at 6 GHz, 8 GHz, and 10 GHz are below the cut-off frequencies of the corresponding SIW waveguides, which are 8.8 GHz, 10.7 GHz, and 12.2 GHz, as shown in Fig. 4.
To verify the operation of the filters below the cut-off frequencies, the relative permittivity of each SIW-CSRR filter was calculated using the CST simulator. The extracted results are shown in Fig. 6, where the permittivity extracted from all the filters is observed to be negative, thereby confirming that all the SIW-CSRR filters can operate below the cut-off frequency.
III. SIW-CSRR SINGLE-STAGE POWER AMPLIFIER
1. SIW-CSRR Impedance Matching Network
Fig. 7 shows the complete impedance matching network for the SIW-CSRR single-stage PA, which consists of an SIW-CSRR part and a few other parts—the bias network, transmission line sections, and the DC-block capacitor. Here, the optimum source and load impedances ZS and ZL, respectively, which were determined by running a source/load pull simulation in an ADS simulator, were shifted to Zin and Zout because of the effects of the bias network, transmission line sections, and the DC-block capacitor. The values of Zin and Zout were calculated by accounting for the effects of the other parts, excluding the SIW-CSRR part. Subsequently, the SIW-CSRR part transformed the standard 50 Ω impedance into the desired Zin and Zout. Notably, this impedance transformation method was implemented based on the technique introduced in [11].
Fig. 8 presents a layout and an equivalent circuit of the SIW-CSRR part of the matching network as a narrow-band filter. Here, it must be noted that the SIW-CSRR part refers to the SIW-CSRR filter, as mentioned in Section II. This means that the SIW-CSRR filter can be employed for the impedance matching purpose. Therefore, the evanescent mode operation can also apply to the impedance matching network. In other words, the center frequency of the matching filter must be lower than the cut-off frequency of the SIW transmission line. In Fig. 8, the source and load impedance of the filter are referred to as Z1 and Zi, which are complex, while J0,1 and J1,2 are the impedance inverters, and B1(ω) represents the parallel resonator or the CSRR. The order of this filter is one, meaning that it involves only one resonator or has a narrow-band characteristic. Its center frequency was the same as the resonant frequency of the CSRR. Notably, the fractional bandwidth (FBW) of the matching network employed in the present study was calculated as less than 2%, with the absolute bandwidth being 100 MHz for all SIW-CSRR filters and Z1 = 50 Ω. Furthermore, Ri and Xi denote the real and imaginary parts of the load impedance Zi, which is Zin (Zout).
All the design parameters of the matching networks, as presented in [11], were calculated directly. These calculations were performed for all the input matching networks (IMNs) and output matching networks (OMNs) at the three operation bands of 6 GHz, 8 GHz, and 10 GHz. Subsequently, these calculated ideal values were used to derive the physical dimensions of the SIW-CSRR filter. Here, it is important to note that the dimensions of the CSRR—A, F, G, and W—defined the resonant frequency of the SIW-CSRR filter, whereas its external Q-factor defined the spacing between each CSRR ring or T. The remaining parameters, as shown in Fig. 3, were determined by fitting the impedance results of the EM simulation to those of the ideal filter’s simulation. The final values of the physical dimensions of all the SIW-CSRR matching networks are listed in Table 1. Notably, the dimensions in Table 1 are slightly different from those shown in Fig. 3 since the physical dimensions of the SIW-CSRR had to be modified for the purpose of impedance matching.
Fig. 9 shows photos of the fabricated SIW-CSRR parts of the complete IMN and OMN at the three frequency bands. Notably, all the SIW-CSRR matching networks were designed and fabricated on Rogers RO4003C substrate characterized by a dielectric constant of 3.55, loss tangent of 0.0027, and substrate thickness of 0.508 mm. Furthermore, the sizes of all the SIW-CSRR matching networks are the same—1.4 cm × 1.4 cm. The impedance of the matching networks was measured using a vector network analyzer (VNA) from Keysight. When measuring impedance, the port extension function embedded in the VNA was applied to de-embed the delay effect of the SubMiniature version A (SMA) connectors.
Fig. 10 compares the impedance results attained based on the filter synthesis method, EM simulation, and measurement for the IMNs and OMNs at 6 GHz, 8 GHz, and 10 GHz. It is observed that the EM simulation and measurement results obtained using the ideal filter synthesis method are highly consistent, thus validating the accuracy and effectiveness of the proposed impedance matching method.
Next, the impedance realized by the complete matching network, including both the input and output sides, at all three operating frequencies was evaluated. Fig. 11 shows the resulting impedance—the optimum source and load impedances ZS and ZL—realized by the complete matching circuit (depicted in Fig. 7). Notably, these values differ from those shown in Fig. 10, which indicates the resulting impedance of only the SIW-CSRR part. It is observed that the simulated results are consistent with the measured results, thus confirming the effective-ness of the impedance matching method. Slight deviations between the simulation and measurement can be attributed to the effects of PCB fabrication tolerance.
2. SIW-CSRR Power Amplifier
After all the IMNs and OMNs at the different operating frequencies were correctly designed, the single-stage PA was fabricated by incorporating all necessary components. Fig. 12 shows the schematic for EM co-simulation of the single-stage PA, which was conducted using an ADS simulator. It consists of an active device, a GaN HEMT device, and the complete IMN and OMN. The GaN HEMT device was provided by Qorvo [23], and its nonlinear model was obtained from Modelithics [24]. The PA was designed and simulated for both small-signal and large-signal performance using the ADS simulator. Fig. 13 displays the fabricated prototypes of the SIW-CSRR single-stage PA at 6 GHz, 8 GHz, and 10 GHz, showing that the sizes of all the PAs are relatively compact—44.5 cm × 18.5 cm, 41 cm × 15.7 cm, and 38.5 cm × 12.4 cm, respectively. This was achieved due to the SIW-CSRR employing evanescent mode operation. Even in the case of the lower frequency band at 6 GHz, the size of the PA is compact. This is especially relevant for designing circuits operating at lower frequencies, considering that traditional SIW technology is not appropriate for this purpose as it leads to a large circuit size.
3. Small-Signal Evaluation
The evaluation of small-signal performance of the single-stage PA included measuring its input and output return losses (or S11 and S22, respectively) and its power gain (or S21). These parameters were measured using VNA N5242A from Keysight. Fig. 14 shows the simulated and measured small-signal parameters— S11, S22, and S21—of the designed single-stage SIW-CSRR PAs. It is observed that the measured gain is 14.7 dB, 12.51 dB, and 5.19 dB at 6 GHz, 8 GHz, and 10 GHz, respectively. The measured input return loss is −16.7 dB, −13.4 dB, and −17.7 dB, while the output return loss is −6.5 dB, −10 dB, and −8 dB at 6 GHz, 8 GHz, and 10 GHz, respectively. Overall, the simulated results are in good agreement with the measured ones. Notably, the measured result at 10 GHz shows some degradation compared to the simulated results, possibly due to the effects of low-quality RF-bypass capacitors, which were not self-resonated at 10 GHz, and the SMA connectors in the PCB.
4. Large-Signal Evaluation
To evaluate the large-signal performance, power gain and efficiency were calculated. Fig. 15 shows the simulated and measured Pout, power added efficiency (PAE), and gain of the 6-GHz, 8-GHz, and 10-GHz single-stage PAs. Notably, all the PAs were biased at Vds = 32 V, and a quiescent current of 25 mA, equivalent to Vgs = −2.57 V, was applied. Fig. 15 shows that the Pout, gain, and PAE of the 6-GHz PA are 36.96 dBm, 14.1 dB, and 53.2%; those of the 8-GHz PA are 35.57 dBm, 11.6 dB, and 47.2%; while the values for the 10-GHz PA are 30.5 dBm, 5 dB, and 31.7%, respectively. As pointed out previously, the low-quality RF-bypass capacitors in the 10-GHz PA resulted in a degradation of the measured performance compared to the simulation. Overall, the simulations accurately predicted the measurements at the large-signal level.
IV. SIW-CSRR TRI-BAND POWER AMPLIFIER
The tri-band PA was realized by connecting the SIW-CSRR triplexers and the SIW-CSRR single-stage PAs, as indicated in Fig. 1. All components were realized using SIW-CSRR technology based on evanescent mode operation.
1. SIW-CSRR Triplexer
Fig. 16 shows a photograph of the fabricated triplexer and its simulated electric field distribution at the different operating frequencies. Each branch of the triplexer comprises an SIW-CSRR unit filter, which has already been described in Section II. To effectively realize multiband operation, the triplexer needed to exhibit a wide passband ratio, low loss, and compactness. In this context, the T-junction plays a critical role in determining this passband ratio. As shown in Fig. 16, the triplexer was designed using three separate SIW-CSRR unit filters exhibiting three passbands at 6 GHz, 8 GHz, and 10 GHz, respectively. The three filters were connected to each other using a T-junction. The characteristic impedance in each arm of the T-junction was calculated and optimized to maintain impedance matching between the input arm and the three filters. Each SIW-CSRR filter was designed to ensure low loss by exploiting the high unloaded Q-factor of the SIW cavity. In addition, the dimensions of the CSRR were theoretically calculated and then optimized using a 3D full-wave EM simulator—CST Microwave Studio—to exhibit the passband within the desired operating frequency band. The triplexer was then fabricated on RO4003C substrate with a dielectric constant of 3.55, a loss tangent of 0.0027, and a substrate height of 0.508 mm. The fabricated prototype of the triplexer was measured using a VNA from Keysight, with the upper frequency limit maintained at 26.5 GHz. The performance of the triplexer is as follows: the minimum measured insertion losses achieved were 1.5 dB, 1.4 dB, and 1.9 dB, while the minimum measured return losses were −9.33 dB, −11.3 dB, and −12.4 dB at 6 GHz, 8 GHz, and 10 GHz, respectively. The measured isolation between the output ports was observed to be better than −17 dB within the frequency range of 5 GHz to 11 GHz. Moreover, the proposed triplexer occupies little space, with its size being 35 mm × 25 mm. The detailed design of the triplexer is described in [25].
2. Tri-Band Power Amplifier
Fig. 17 shows a photograph of the fabricated prototype of the proposed SIW-CSRR tri-band PA. Notably, all critical components of the PA, such as the triplexers and the single-stage PAs, were implemented on the SIW-CSRR by employing evanescent mode. The PA was fabricated on the same RO4003C substrate as the triplexers and single-stage PAs.
3. Small-Signal Evaluation
The measured small-signal gain as well as the input and output return losses at the three operating frequencies are presented in Fig. 18. The measured small-signal gain or S21 at 6 GHz, 8 GHz, and 10 GHz is 14.1 dB, 11.6 dB, and 4.92 dB, respectively. Notably, the low gain attained at 10 GHz can be attributed to low-quality RF-bypass capacitors, as pointed out previously. Furthermore, it is observed that there are three peaks in the S21 response, implying that the designed PA correctly exhibits triband operation at the small-signal level. In addition, the input return loss or S11 is −9.1 dB, −8.9 dB, and −22 dB, whereas the output return loss or S22 is −27.1 dB, −6.2 dB, and −18.5 dB at 6 GHz, 8 GHz, and 10 GHz, respectively. Taken together, the measured return losses ensure that the proposed PA works well at the three operating frequency bands.
4. Large-Signal Evaluation
4.1 Output power and efficiency
Fig. 19 shows the simulated and measured efficiency in terms of the PAE and transfer characteristics of the three individual SIW PAs at 6 GHz, 8 GHz, and 10 GHz. Good agreement between the simulations and measurements is observed for the 6-GHz and 8-GHz PAs. Notably, all the measured results accounted for the effect of SMA connectors. At 6 GHz, the measured PAE, saturated output power, and power gain are 52.2%, 36.57 dBm, and 11.4 dB, respectively; these values at 8 GHz are 47.2%, 35.51 dBm, and 11.3 dB, respectively; while the values at 10 GHz are 30.3%, 34.57 dBm, and 6.1 dB, respectively. The simulated and measured frequency response of the proposed PA is illustrated in Fig. 20, confirming that the simulations accurately predicted the measured performance within the frequency range around the center frequency of each PA.
4.2 Linearity
The linearity performance of the proposed PA was evaluated using a 16-Mbps QPSK modulation signal with a bandwidth of 10.72 MHz. As depicted in Fig. 21, all the designed PAs achieved an ACPR level better than −35 dBc, with the output power being 32 dBm at the center frequencies of 6 GHz, 8 GHz, and 10 GHz.
The final test conducted on the designed PA evaluated its ability to reject undesired spectrum at the output. Fig. 22 presents the simulated and measured output spectrums of the PA, which were obtained by simultaneously exciting three input signals to the PA at 6 GHz, 8 GHz, and 10 GHz. Significant rejection of undesired output spectrum is observed, resulting in a highly clean output spectrum.
Finally, the performance of the proposed PA was compared with that of previously reported PAs, including SIW PAs and tri-band microstrip-line PAs, as presented in Table 2 [8–12, 26–31]. It is observed that the proposed PA exhibits high-frequency multiband operation at the C-band and X-band, as well as good linearity, achieving an ACPR better than −35 dBc, compared to previously reported works that employed both microstrip-line and SIW technologies.
V. CONCLUSION
This paper proposes a methodology for designing a tri-band PA that can operate at the 6-GHz, 8-GHz, and 10-GHz bands. To ensure both compactness and effectiveness, SIW-CSRR filtering structures were incorporated into the proposed PA to realize impedance matching networks. Furthermore, the size of the PA was reduced using evanescent mode mechanism in the SIW-CSRR structures. The PA was designed, simulated, and tested for both small-signal and large-signal characteristics. For large-signal evaluation, it was tested using both continuous wave (CW) and modulated signals. In CW mode, the measured PAE, power gain, and output power at 6 GHz, 8 GHz, and 10 GHz were 52.2%, 11.4 dB, and 36.57 dBm; 47.2%, 11.3 dB, and 35.51 dBm; and 30.3%, 6.1 dB, and 34.57 dBm, respectively. In the modulated mode featuring a 16-Mbps QPSK and 10.72-MHz bandwidth modulation signal, the ACPR level achieved by the designed PA was observed to be better than -35 dBc, considering an output power of 32 dBm, at the center frequencies of 6 GHz, 8 GHz, and 10 GHz. The proposed PA also features a compact size of 6 cm × 8 cm, even when using additional triplexers. In addition, the PA proved capable of eliminating undesired signals at the output, owing to the good performance of the SIW-CSRR triplexer.
Compared to other SIW PAs and traditional microstrip-line PAs reported in the literature, the proposed PA exhibits multiband capabilities at high frequencies, while also maintaining compactness and good linearity performance. Moreover, the measured results agreed well with the simulated results, thereby validating the effectiveness of using SIW-CSRR technology to design multiband PAs for modern wireless communications.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
Print
