Low NF and High P1dB Wideband Quasi-Circulator With Unequal Power Split and Reconfigurable Inter-Stage Matching

In this article, a wideband quasi-circulator (QC) with the low noise figure (NF) and high 1-dB compression point (<inline-formula> <tex-math notation="LaTeX">$\text{P}_{\text {1 dB}}$ </tex-math></inline-formula>) is proposed. The slotline-based 180° hybrid is introduced to achieve the wideband high transmitter-receiver (TX-RX) isolation. Meanwhile, the bidirectional in-phase stage and the nonreciprocal out-of-phase stage are utilized to obtain specific power-split ratios for the low insertion loss and high antenna-transmitter (ANT-TX) isolation. By using the unequal power split, the proposed QC is without the fundamental 3-dB loss seen in reciprocal ANT interfaces. Besides, to further enhance the TX-RX isolation within a wideband, a reconfigurable inter-stage matching network based on a varactor-tuned capacitor is implemented. To verify the mechanisms mentioned above, a reconfigurable wideband QC operating at 1.75–2.9 GHz is implemented and fabricated. The total power consumption is 224 mW. The measurement exhibits in-band TX-RX isolation of 27–58 dB. The TX-ANT and ANT-RX insertion losses are 4.3–5.0 dB and 1.0–1.9 dB, respectively. The QC also achieves the minimum in-band ANT-RX NF of 3.4 dB. Meanwhile, the measured TX-ANT and ANT-RX IIP3s are 39.4 and 32.5 dBm, respectively. The TX input <inline-formula> <tex-math notation="LaTeX">$\text{P}_{\text {1 dB}}$ </tex-math></inline-formula> is greater than 28.78 dBm, while the RX input <inline-formula> <tex-math notation="LaTeX">$\text{P}_{\text {1 dB}}$ </tex-math></inline-formula> is 23 dBm. The TX-induced ANT-RX <inline-formula> <tex-math notation="LaTeX">$\text{P}_{\text {1 dB}}$ </tex-math></inline-formula> is measured to be 28 dBm.

spectral efficiency of wireless systems. The FD radio is widely known as a simultaneously transmit and receive capable radio since it enables the transmitter (TX) and receiver (RX) to operate at the same time and over the same frequency. Such operation substantially alleviates the radio spectrum allocation burden, thereby yielding uncompromised communication data capacity [3]- [7]. One of the entire challenges of an FD transceiver is that transmitted signals can directly feed into the RX chain, which will cause self-interference and severely diminish the RX's dynamic range. To avoid such self-interference, a pair of antennas (ANTs) in the RF/analog domain is introduced with the drawback of large size [8]. Then, the FD transceiver with a shared ANT is investigated. Such operation seems to be more feasible in the FD system for the minimized ANT size and can be easily used in multiple-input, multiple-output applications. The electrical balance duplexers (EBDs) are utilized to achieve the high TX-RX isolation [9]- [11]. However, such duplexers feature a fundamental minimum 3-dB loss due to their reciprocity.
Recently, circulators and quasi-circulators (QCs) have been widely developed as a transceiver front-end with a shared ANT interface. The conventional ferromagnetic circulator [12] is presented with good isolation and return loss, however, in the sacrifice of high cost and bulky size. To minimize the size of the circuit, active QCs are developed [13]- [15], which are with the drawback of relatively high noise figure (NF) and limited power handling. To improve the NF and power handling performances, passive circulators are proposed [16]- [20]. Nevertheless, the requirement of the additional modulated LO signals increases the system complexity. Recently, QCs using the combined active and passive circuits are developed with advantages of design flexibility and good performance [21]- [25]. As shown in Fig. 1(a), a tunable QC using the compact fully reconfigurable 180 • hybrid and amplifier stages is proposed to improve the isolation performance [25]. However, such a combined-type of QC suffers from the fundamental 3-dB path loss introduced by the reciprocal passive devices, while the operation frequency range still needs to be further expanded. Therefore, the circulator or QC design with the merits of the wide bandwidth, low NF, and high power handling still remains a great challenge.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ (a) Configuration of the conventional combined-type QC [25]. (b) Topology of the proposed wideband QC.
As depicted in Fig. 1(b), a wideband QC with the low RX NF and high 1-dB compression point (P 1 dB ) is proposed in our conference article [26]. Such QC utilizes a nonreciprocal network connecting to a 180 • hybrid to achieve the ANT-TX reverse isolation instead of using an amplifier. Thus, the amplifier's limitation of TX power handing could be avoided. In this work, the wideband 180 • hybrid is presented to achieve the wide operation band and high TX-RX isolation. Meanwhile, a bidirectional in-phase stage based on the common-gate (CG) stage and a nonreciprocal out-of-phase stage based on the common-source (CS) stage are introduced, which obtain specific power-split ratios for the low insertion loss and high ANT-TX isolation. To the authors' best knowledge, it is the first time that the QC uses the unequal power split without additional modulated signals to omit the fundamental 3-dB loss seen in reciprocal ANT interfaces, while achieves low NF and high P 1 dB .
Compared to the previous work in [26], a deep analysis using the power transmission method is proposed and the simplified model of the architecture is extracted in this article. Based on the theoretical analysis, a reconfigurable inter-stage matching network and a step-impedance 180 • hybrid are utilized to further improve the TX-RX isolation and the operation bandwidth, respectively. This article is organized as follows. In Section II, the architecture and operation principle of the proposed QC using unequal power split is presented. A prototype QC is implemented by a wideband 180 • hybrid, a CG-stage, and a CS-stage in Section III. The proposed reconfigurable wideband QC is measured and compared with state of the art in Section IV. The conclusion is summarized in Section V. Fig. 2 shows the principle and configuration of the proposed reconfigurable wideband QC. It is composed of a wideband 180 • hybrid, a reconfigurable bidirectional in-phase stage, and a nonreciprocal out-of-phase stage. Port (1) is the sum port and Port (4) is the difference port of such hybrid, which connect to the TX and RX Ports, respectively. Besides, the bidirectional in-phase stage and nonreciprocal out-of-phase stage are connected to the hybrid by the inter-stage matching networks, respectively. Note that from left to right, the gain of the outof-phase stage is larger than that of the in-phase stage and the signals are out-of-phase at the right outputs. Differently, from right to left, the signal is blocked at the out-of-phase stage due to the nonreciprocity and could only transmit through the in-phase stage.

A. System Architecture
The proposed QC utilizes the unequal power split to reduce the trade-off between the TX insertion loss and RX NF. Here, the unequal power-split situations appear at the 180 • hybrid and the T-junction of the ANT Port, respectively. For transmit mode operation, ideally, the TX input signal is split into two paths with unequal magnitude by the 180 • hybrid. Then, by optimizing the load impedance of Port (3), a larger than half portion of the TX signal is transmitted to Port (2), and the remaining portion is transmitted to Port (3), respectively. Thus, the lower TX insertion loss could be achieved. Since the signal transmits from Port (1) is suppressed at Port (4) of the 180 • hybrid, the TX signal is canceled at the RX Port. For receive mode operation, the signal from ANT Port is divided into two paths, which are defined as RX path1 and RX path2, respectively. Then, larger than half of the RX signal is transmitted to the RX path1 and the rest is transmitted to the RX path2. Here, the signals in RX path are depicted at each critical node shown in Fig. 2 (i.e., N 1 , N 2 , N 3 , N 4 , N 5 , and N 6 ). In RX path1, due to the non-inverting for the signal of the in-phase stage, the signals are in-phase at the N 1 and N 2 . Besides, in RX path2, the signal is inverted at the output of the out-of-phase stage. Thus, at the N 2 and N 5 , the signals are out-of-phase. By implementing the out-of-phase stage with specific transmission gain, the differential-mode signal could be canceled at Port (1), which represents the high ANT-TX isolation. Meanwhile, it is worth pointing out that the more power transmit to the RX path1, the larger gain of the outof-phase stage is required to obtain the high ANT-TX isolation, which leads to lower RX NF. Therefore, due to the unequal power split with the out-of-phase stage's power compensation, both lower TX insertion loss and RX NF of the QC could be achieved. Incidentally, the RX-ANT isolation is relatively low due to the in-phase stage reciprocity. However, this is not so important considering that the input signal is usually not from RX Port [27].
Therefore, based on the aforementioned principles, the QC function is achieved by combining the wideband 180 • hybrid, bidirectional in-phase stage, and nonreciprocal out-of-phase stage. The TX-RX isolation is provided by the 180 • hybrid, and the ANT-TX isolation is introduced by the bidirectional in-phase stage with nonreciprocal out-of-phase stage, respectively. Besides, the unequal power split is utilized to reduce the trade-off between the TX insertion loss and RX NF. To investigate the mechanism of the proposed reconfigurable wideband QC, the three-layer dielectric substrate (i.e., ε r1,3 = 3.66 and ε r2 = 3.55 with thickness of h 1,3 = 0.508 mm and h 2 = 0.1 mm), full-wave EM-simulator IE3D, and ADS are used.

B. Operation
The block diagram of the proposed QC and the power at critical nodes are shown in Fig. 3, where the matching networks are omitted. The P ih is power at terminal (i ) of the hybrid (i = 1, 2, 3, 4). Besides, the P 1t is power at the ANT Port (i.e., T 1 ). The P 2t and P 3t are power at inputs of the in-phase stage and out-of-phase stage (i.e., T 2 and T 3 ), respectively. Meanwhile, the a i are incident waves and the b i are reflected waves. For simplicity, assuming the QC is lossless and all terminations are matched, while the wideband 180 • hybrid has high isolation.
Input signal a 1 at the TX Port is split into upper and lower branches. Then, a portion of the upper signal is transmitted to the ANT Port by the in-phase stage and a T-junction. The transmitter power at the ANT Port could be derived as where H 21h,TX is the power transmission coefficient from Port (1) to Port (2) of the hybrid in TX path, H IS is the power transmission coefficient of the in-phase stage, and H 12t,TX is the power transmission coefficient from T 2 to T 1 of the T-junction in TX path. Note that the TX-ANT transmission parameter (i.e., P b 2 ,TX / P a 1 ,TX ) is related to the H 21h,TX of the hybrid and H 12t,TX of the T-junction. The |H 21h,TX | and |H 12t,TX | could be expressed as For a signal a 2 entering the ANT Port, assuming the interconnected mismatch is ignored, the power received by the TX Port can be derived by Here, H 21t,RX is the power transmission coefficient from T 1 to T 2 of the T-junction in RX path, H OS is the power transmission coefficient of the out-of-phase stage, and H 12h,RX is the power transmission coefficient from Port (2) to Port (1) of the hybrid in RX path. The |H 21t,RX | and |H 12h,RX | could be expressed as Similarly, the |H 31t,RX | and |H 13h,RX | could be obtained. To achieve the high ANT-TX isolation, ideally, the P b 1 ,RX in (4) should be 0, which leads to the following condition: Due to the reciprocity of the hybrid, the relation between the power transmission coefficients in the TX path and RX path of the hybrid could be as where i = 2 or 3. Then, (7) can be rewritten as Since the ANT-TX is isolated, the received signals are recombined constructively at the RX Port. Thus, the power received by the RX Port can be derived as Therefore, to achieve the lower RX insertion loss, the power-split ratio of the T-junction represented by k 1 should satisfy the following condition: Meanwhile, according to (1) and (2), to obtain the lower TX insertion loss, the power-split ratio of the hybrid represented by k 2 should meet the following requirement: Then, the relationship between the transmission parameters of the QC (i.e., |S 21 | and |S 32 |) and the power-split ratios (i.e., k 1 and k 2 ) can be simply expressed as Note that for equal power split (i.e., k 1 = 1 and k 2 = 1), the calculated |S 21 | is −6 dB and the |S 32 | is 0 dB, which indicates the inherent 3-dB path loss provided by the reciprocal approaches. Fig. 4 shows the calculated |S 21 | and |S 32 | versus power-split ratios k 1 and k 2 according to (13) and (14). Here, (9) is satisfied and |H IS | = 1. It can be seen that by properly choosing the k 1 and k 2 , the TX insertion loss could be lower than 3 dB and the RX gain is larger than 0 dB simultaneously. However, it doesn't exhibit zero TX-ANT loss fundamentally because the signal from TX to ANT through the out-of-phase path is completely blocked. The proposed structure may also exhibit a system insertion gain due to the large |H OS |. The overall circuit stability  directly depends on the gain of the out-of-phase stage and the isolation of the 180 • hybrid (i.e., |S 23,h |). The internal loop may exhibit an open-loop gain higher than unity, which leads to the potential onset of oscillations. The loop gain of the positive feedback system could be simply expressed as |S 32,n ||S 23,h |, where the |S 32,n | is the gain from in-phase stage to out-of-phase stage. If the loop gain is smaller than 1, the overall system gain is finite, leading to the following condition: Due to |S 32,n | is positively correlated with |H OS |, the value of the |H OS | cannot be too large. As shown in Fig. 5, two circuits (i.e., a general T-junction and a divider with λ/4 transmission lines) with different load impedance relationships are utilized to obtain the required power-split ratio. The relations between the power-split ratio and the load impedance of the two circuits are expressed as follows: where Z i,t is the load impedance of the T-junction and Z i,d is the load impedance of the divider (i = 2, 3). The detailed derivation is given by (19)- (24) in the Appendix. It can be seen that the k t is unaffected by the ANT voltage standing  wave ratio (VSWR). Note that as shown in Fig. 6, the k t and k d show the opposite trend while the load impedance changes. The simplified model of the proposed QC is depicted in Fig. 7 to further analyze the operating principle. Such a model consists of ideal transmission lines, no-noise amplifier models, and impedance matching circuits. To obtain the larger k t and k d simultaneously, the design procedures are summarized as follows: Firstly, the Z 3,t is designed for the larger k t to achieve lower RX insertion loss. Secondly, the Z 3,d is implemented for the larger k d to achieve lower TX insertion loss. Finally, the gain of the out-of-phase amplifier (i.e., |H 21,OA |) is chosen to compensate the unequal power-split ratios k t and k d for high ANT-TX isolation. Thus, the trade-off between the TX and RX insertion losses could be reduced. The C 1 , C 3 , and R 0 are designed for desired power-split ratios k t and k d . Besides, the capacitor C 2 is introduced to tune the phase-imbalance in the RX path and the impedance Z 2,d for higher isolation. Meanwhile, the L 1 and C 4 are utilized for ports matching, respectively. Fig. 8(a) shows the simulated TX-ANT transmission parameter (|S 21 |) and RX NF of the QC using the proposed simplified model. Here, the in-phase amplifier is assumed to be lossless (i.e., H 12,IA = H 21,IA = 1). Note that the TX insertion loss and RX NF could simultaneously be lower than 3 dB at 2.3 GHz, which indicates the proposed QC is without fundamental 3-dB loss seen in reciprocal ANT interfaces. Meanwhile, as shown in Fig. 8(b), the return loss of all ports is better than −10 dB. Fig. 9 shows the effects of the gain |H 21,OA | on the load impedance Z 3,d , the ANT-RX stability factor (K 23 ), and S-parameters with RX NF, respectively. As shown in Fig. 9(a), the real part of Z 3,d decreases sharply as the |H 21,OA | increases, which leads to the larger k d from (17). Besides, the larger |H 21,OA | could obtain lower RX NF. Thus, lower TX-ANT insertion loss and RX NF could be achieved simultaneously. However, as depicted in Fig. 9(b), the ANT-RX stability factor (K 23 ) becomes worse with the increasing of |H 21,OA |. Meanwhile, due to the unequal power split of the proposed hybrid, there will be an impedance mismatch in the TX path while looking into the out-of-phase amplifier output. The TX leakage signal reflected back by the out-of-phase amplifier could degrade the TX-RX isolation. As shown in Fig. 9(c), the increase of the |H 21,OA | leads to the TX-RX isolation reduction. Therefore, there is a trade-off between the low insertion loss and high TX-RX isolation.

III. CIRCUIT IMPLEMENTATION
Based on the principle investigated in Section II, a reconfigurable wideband QC based on a wideband 180 • hybrid, a reconfigurable bidirectional in-phase stage, and a nonreciprocal out-of-phase stage is implemented in a three-layer dielectric substrate (i.e., ε r1,3 = 3.66 and ε r2 = 3.55 with thickness of h 1,3 = 0.508 mm and h 2 = 0.1 mm). The design procedures of the proposed QC are summarized as follows: Firstly, a wideband 180 • hybrid is implemented for the wide operation band and high TX-RX isolation. Secondly, a bidirectional in-phase stage based on the CG-stage and a nonreciprocal out-of-phase stage based on the CS-stage are utilized to obtain the high ANT-TX isolation. Finally, a reconfigurable inter-stage matching network based on the varactor-tuned capacitor is introduced to  further enhance the TX-RX isolation under different operating frequencies.

A. Wideband 180 • Hybrid
The layout of the wideband 180 • hybrid is shown in Fig. 10(a), which is formed on a three-layer substrate. Note that a slot is etched in the middle metal-ground. The Ports   Such a hybrid can be regarded as an assembly of outof-phase and in-phase dividers/combiners. For the out-of-phase case in Fig. 10(b), the signal from Port (4) is coupled to Ports (2) and (3) through the slot. Due to the inherent property of the slot [12], the divided signals at Ports (2) and (3) are outof-phase. For the in-phase case in Fig. 10(c), the signal from Port (1) is split into two lines at the T-junction. Then, due to the symmetry of the two lines, the divided signals at Port (2) and Port (3) are in-phase. Once the hybrid is operated as a combiner, the in-phase signal from Ports (2) and (3) is canceled at Port (4), and similarly, the out-of-phase signal is disappeared at Port (1). The width of the slot is optimized based on the analysis in [28] and the slot is designed as a meander-shape for size miniaturization. The parameters w 4 , l 1 , and l 4 are optimized for improving the return loss of Port (4). Besides, the parameters w 1 , w 5 , w 6 , l 2 , and l 3 are optimized carefully to improve the power dividing performance. Here, the step-impedance design method (i.e., w 5 is not equal to w 6 ) is introduced to further wider the operation bandwidth. The simulated S-parameters of the proposed 180 • hybrid are depicted in Fig. 11. It is notable that high isolation within the wide frequency range of 1.6-3 GHz is achieved. Fig. 12 shows the amplitude-and phase-imbalances responses of Ports (2) and (3), while the hybrid is operated as a divider. The amplitude-imbalances of the in-phase case and out-of-phase case are both less than 0.2 dB and the phase-imbalances are both less than 0.6 • , which indicates a good in-band balance performance.
As shown in Fig. 13, the varied load impedance Z 2,h and Z 3,h at Port (2) and Port (3) show negligible influence on the isolation |S 32,h | of the hybrid. Meanwhile, the insertion losses |S 21,h |, |S 24,h |, and isolation |S 41,h | of the hybrid are affected by the load impedances of Port (2) and Port (3). Note that the path losses of the hybrid and the inter-stage impedance matching will affect the TX power efficiency and the RX NF of the QC [12], which should be taken into consideration in the design of the proposed QC with the optimized performance.

B. CG-Stage and CS-Stage
The CG-stage and CS-stage are employed to realize the bidirectional in-phase stage and nonreciprocal out-of-phase stage, respectively. The CG-stage is utilized to produce 0 • phase shift for signal and wideband matching for ANT Port [29]. Besides, the CS-stage is introduced to produce a 180 • phase shift for signal and nonreciprocal amplify. Meanwhile, the CG-CS noise-canceling technique is used to cancel the thermal noise of the CG-transistor in the RX path [30], which further reduces the RX NF. Fig. 14 shows the configuration of the CG-stage and CS-stage. The GaAs high electron mobility transistors are widely used in microwave amplifier design due to their good characteristics. Thus, the GaAs FET ATF-34143 from Avago, high Q capacitors, and RF inductors from Murata are utilized to implement the CG-stage and CS-stage.
The bias voltage V G1 and V D1 are configured for low conduction loss of the CG-stage. The optimized bias V G2 and V D2 are chosen to achieve the required gain of the CS-stage. Besides, the input impedance of the CG-stage is implemented for wideband ANT Port matching, while the input impedance of the CS-stage is designed for required power-split ratio k 1 and low RX NF. The capacitor C 3 could be used to trade the TX insertion loss and RX insertion loss. Meanwhile, the output impedances of the CG-stage and CS-stage are matched to the proposed wideband 180 • hybrid, which obtains the required power-split ratio k 2 and low interconnected loss. By properly designing the CG-stage and CS-stage, a good phase-balance performance at the CG-CS stage outputs could be achieved. Thus, the received signals are recombined constructively at the RX Port, while the thermal noise of the CG-transistor could be canceled. Moreover, the source negative feedback resistor R 3 of the CS-stage is introduced to enhance the stability. Since the gain of CS-stage can be controlled by the gate voltage V G2 of the CS amplifier, the potential instability can be taken care of by reducing the V G2 of the CS amplifier. Due to the principles of the proposed hybrid are similar to a rat-race coupler, the balance condition to achieve high isolation could be expressed as Therefore, a tunable C t is utilized to obtain higher TX-RX isolation versus operating frequency by tuning the input impedance of the P 2. Fig. 15 shows the simulated phase-imbalance of the outputs P 2 and P 3 versus the C t . The phase-imbalance is less than 8 • and the best balance performance could be tuned by the C t . Besides, as depicted in Fig. 16, the input impedance of the P 2 varies with the tuning capacitor C t . Note that at a relatively low frequency range, the larger C t leads to the lower difference between S 22,h and S 33,h . However, this trend is opposite at the higher frequency range. Thus, to achieve the high TX-RX isolation versus different operating frequency, the larger C t should be chosen at a relatively low-frequency range, while the lower C t should be chosen at a relatively high-frequency   range. Fig. 17(a) depicts the configuration of the C t used in the inter-stage matching network, which is a dc-block capacitor C D in series with a varactor (i.e., BB857 from Infineon). The control voltage V C is fed by a 10-k resistor R B . It is worth pointing out that there is a trade-off between the tuning range and quality factor of the C t [25]. Then, an optimized C D = 0.6 pF (i.e., GRM1555 series capacitors from Murata, with Q D = 124 at 2.35 GHz) is chosen for the tunable C t implementation. The simulated characteristics of the C t are shown in Fig. 17(b).

C. Reconfigurable Wideband QC
The configuration of the proposed QC is depicted in Fig. 18. Here, the 10-and 0.3-pF capacitors are added for TX and RX Ports matching due to the effects of parasitism, respectively. Fig. 19 shows the simulated TX-RX isolation versus control voltage. To achieve the maximum isolation at the different operating frequency, the V C is chosen as 0 V   Fig. 20, the RX NF is lower than 3.6 dB within the frequency range of 1.6-3 GHz. To further investigate the performance of the proposed QC in the FD wireless application, the NF of the RX-path is evaluated while the transmitter is operating. Fig. 21 shows that the in-band RX  NF degenerates slightly with the increasing of TX power. The probable reason is the signal-noise ratio degeneration due to the TX leakage.
IV. FABRICATION AND MEASUREMENT Based on the mechanisms and circuits mentioned above, a reconfigurable wideband QC is fabricated and measured.   The photograph of the proposed QC is illustrated in Fig. 22 and the total circuit size is 61.7 mm × 32 mm (i.e., 0.925 λ g × 0.479 λ g , where λ g is the microstrip guided wavelength at 2.35 GHz). Fig. 23 shows the measurement setup. The S-parameters are measured by the N5230A network analyzer. Besides, the NF is performed using the N4002A noise source and N8975A NF analyzer. Meanwhile, the large-signal performances are measured by the Rohde & Schwarz SMA100B signal generator and FSW43 spectrum analyzer.
As depicted in Fig. 24, the maximum TX-RX isolation varies with the control voltage V C . To obtain the higher TX-RX isolation, the V C is chosen as 7 V at a lower frequency range, while the V C is chosen as 28 V at a higher frequency range. Fig. 25 shows the measured S-parameters of the proposed QC with optimized control voltage V C . The measured TX-RX isolation is 27-58 dB over the operating frequency range of 1.75-2.9 GHz. Besides, the in-band TX insertion loss is 4.3-5.0 dB and the RX insertion loss is 1.0-1.9 dB. Here, the difference between simulation and measurement is mainly caused by the component tolerances and the solders used in fabrication. Meanwhile, the TX reversed isolation is greater than 10 dB and the return loss of the TX, ANT, and RX Ports are all better than −10 dB, respectively. The total power   consumption is 224 mW (i.e., 220 mW at CS-stage and 4 mW at CG-stage). Fig. 26 shows the measured TX-RX isolation under various loaded impedances at ANT Port. The ANT VSWR of 1.3:1 is estimated from the measurement plot to support 20-dB TX-RX isolation. Note that since the impedance tuner is unavailable, the loaded impedance at ANT Port is varied by the different combinations of resistors, inductors, and capacitors. The TX-RX isolation under different ANT VSWR can be improved by introducing an ANT impedance balancer or tuner [20], [25]. The SCAL 70CTR3 is used for RX NF measurement under different temperatures. As depicted in Fig. 27, the measured in-band RX NFs are 1.9-3.5 dB, 3.4-4.5 dB, and 3.8-5.1 dB under different temperatures of 250, 300, and 330 K with optimized control voltage V C , respectively. To further improve the losses and NF performances, a bidirectional nonreciprocal out-of-phase    stage could be analyzed and the noise-canceling technique of the CS-stage can be utilized.
As shown in Fig. 28, the measured TX-ANT input P 1 dB is greater than 28.78 dBm (limited by the measurement setup), at which point the compression is only 0.5 dB, while the measured ANT-RX input P 1 dB is 23 dBm. Fig. 29 shows the TX-RX isolation is 29.8 dB for TX input power is 28.78 dBm. Meanwhile, as depicted in Fig. 30, the measured TX-induced ANT-RX P 1 dB is 28 dBm. Fig. 31 shows the measured TX-ANT and ANT-RX IIP3s are 39.4 and 32.5 dBm, respectively.
The performance summary of the proposed QC and comparison with the state of the arts are shown in Table I. Compared to the relevant researches, the proposed QC has a competitive performance of the −20-dB operating frequency range, low RX NF, and high power handling.
V. CONCLUSION In this article, a reconfigurable wideband QC using a wideband 180 • hybrid, a reconfigurable bidirectional in-phase stage, and a nonreciprocal out-of-phase stage is proposed. The slotline-based wideband 180 • hybrid is introduced to achieve the wideband high TX-RX isolation. Besides, the bidirectional in-phase stage and the nonreciprocal out-of-phase stage are utilized to obtain a specific power-split ratio for the low insertion loss and high ANT-TX isolation. By using the unequal power split, such QC is without the fundamental 3-dB loss seen in reciprocal ANT interfaces. Meanwhile, a reconfigurable inter-stage matching network is implemented to further enhance the TX-RX isolation. Then, a reconfigurable wideband QC is implemented, fabricated, and measured. The measured in-band TX-RX isolation is greater than 27 dB over the 49.5% fractional bandwidth (i.e., 1.75-2.9 GHz). The RX insertion loss is 1.0-1.9 dB and the minimum in-band RX NF is 3.4 dB. Meanwhile, the measured TX input P 1 dB is greater than 28.78 dBm, while the RX input P 1 dB is 23 dBm. The TX-induced ANT-RX P 1 dB is measured to be 28 dBm. APPENDIX To investigate the power-split characteristic of the circuits in Fig. 5, an asymmetrical three-port network is introduced from [12]. As shown in Fig. 5(a), assuming the voltage at the T-junction is V 0 and the output powers are in a k t :1 ratio. Then, the output powers at Port t2 and Port t3 could be expressed as (20) where P in is the input power to the matched T-junction, Z i = R i + j X i , and V i,R is the voltage at the resistive portion (i = 2, 3). Therefore, the load impedance of Port t2 and Port t3 should meet the following condition: For a divider with λ/4 transmission lines, as shown in Fig. 5(b), the input impedance looking to Port d2 is Note that, for θ C = 90 • , the Z 2,d is simplified as Meanwhile, the Z 3,d could be obtained similarly. Assume the output powers are in a k d :1 ratio, thus the relation between the power-split ratio and load impedance could be as