Broadband Microwave Photonic Mixer with Flexibly Tunable Phase Shift and Supporting Dispersion-Induced Power-Fading-Free Fiber Transmission

Abstract

A broadband microwave photonic mixer with tunable phase shift and supporting dispersion-induced power-fading-free fiber transmission is proposed and demonstrated based on a dual-polarization dual-parallel Mach–Zehnder modulator (DP-DPMZM). In this scheme, the intermediate-frequency (IF)/radiofrequency (RF) signal and the local oscillation (LO) signal are applied to the four sub-MZMs biased at their minimum transmission points via a power splitter and a 90° hybrid coupler, respectively. Through mutual beating between the IF/RF and the LO modulation sidebands in a high-speed photodetector at the remote site, high-efficiency frequency conversion is achieved. The dispersion-induced power fading over long-distance fiber transmission is eliminated through setting the biased-induced phase difference between the parent-MZMs in the two sub-DPMZMs of the DP-DPMZM to be π/2. In addition, the phase shift of the frequency-converted signal can be continuously tuned over 360° through synchronously adjusting the bias voltages of the parent-MZMs in the two sub-DPMZMs. The proposed scheme is experimentally demonstrated, where a microwave photonic mixer with a 6-dB operation bandwidth of 40 GHz and supporting dispersion-induced power-fading-free transmission over 20 km SMF is realized. Meanwhile, a continuously tunable phase shift over 360° in the frequency range of 0.1 GHz to 29.9 GHz is achieved, where the power variation during phase tuning is smaller than 4 dB.

1. Introduction

Distributed antenna array (DAA) technology is a promising candidate to enhance the measurement accuracy and sensitivity in modern radar systems, where the signal transmission loss must be low enough to support a wide range distribution, and the signal phase in each antenna should be precisely controlled to enhance the synthesis efficiency [1,2]. Benefiting from its broadband signal processing ability and the ultra-low transmission loss in optical fibers, microwave photonics technology is recognized as an effective solution to realize wide-range distributed radar systems based on DAAs.

In the DAA-based radar system, frequency mixing and phase shift are two critical signal-processing functions, which are generally achieved by using independent electronic components such as mixers and phase shifters. The greatest limitation of the electrical solution is its small octave. In order to solve this problem, various broadband microwave photonic mixers with the ability to achieve full-range phase tuning have been proposed and demonstrated in recent years [3,4,5,6,7,8,9,10]. These schemes can be divided into two categories. The first category realizes the functions of the frequency mixing and the phase shift by employing separate hardware. In these schemes, the intermediate-frequency (IF)/radiofrequency (RF) signal and the local oscillation (LO) signal are loaded onto two orthogonally polarized light, respectively, by using a polarization-division multiplexing Mach–Zehnder modulator (PDM-MZM) [3] or a dual-polarization dual-parallel Mach–Zehnder modulator (DP-DPMZM) [4,5,6]. After photodetection, the up/down-converted signal is obtained, whose phase can be tuned within a full range by inserting polarization-sensitive devices before the photodetector, such as a polarization controller (PC) [3,4], a polarization-dependent phase modulator (PM) [5] and a dual-channel phase modulator (DPM) [6]. The second category achieves frequency mixing and phase shift by using a single electro-optic modulator such as a dual-drive Mach–Zehnder modulator (DDMZM) [7], a dual-parallel Mach–Zehnder modulator (DPMZM) [8], a dual-polarization dual-drive Mach–Zehnder modulator (DP-DDMZM) [9] and a DP-DPMZM [10]. Thereinto, the phase of the up/down-converted signal is tuned within a full range by simply adjusting the bias voltages of the electro-optic modulator. However, an optical bandpass filter (OBPF) [8] or a fiber Bragg grating (FBG) [7,10] is essential in most of those schemes to eliminate the optical carrier or the modulation sidebands, which deteriorate the stability and the flexibility of the system.

The most prominent advantage of the microwave photonic mixers lies in that the optically carried IF/RF signal, and the LO signal can propagate in an optical fiber with an ultra-low loss to achieve remote frequency conversion, which is favorable for realizing wide-range DAAs. However, the transmission distance and the operation bandwidth are restrained by the frequency-dependent power fading induced by the group-velocity dispersion (GVD) in the optical fiber, which puts a limit to the application of the existing microwave photonic mixers in the broadband and wide-range DAA. In past years, various approaches have been proposed and demonstrated to eliminate the dispersion-induced power-fading effect in the fiber transmission [11,12,13,14,15,16,17,18,19,20,21,22,23,24]. In [11,12,13,14,15,16,17,18], the frequency response of the fiber link can be shifted to avoid the transmission notch at the desired frequency band by properly controlling the bias voltages of an electro-optic modulator [11,12,13,14,15,16] or simply adjusting the phase difference between two orthogonally polarized optical signals through a PC [17,18]. Nevertheless, since the transmission notch has not been removed in those schemes, the immunity to the dispersion-induced power fading is only effective at a specific frequency rather than a large operation bandwidth after a long-distance fiber transmission. Although single-sideband (SSB) modulation is a valid approach to avoid this problem, the employment of an electrical hybrid coupler (HC) [4,19,20] or an optical filter [21,22] limits the operation bandwidth of the fiber link. In addition, the frequency-dependent phase shift induced by the SSB modulation and the GVD introduces an extra signal distortion in the time domain for a broadband signal. The phase diversity technique is able to compensate for the dispersion-induced power fading in a large bandwidth by combining two parallel signals with complementary transfer characteristics [23,24]; this is an attractive method for wideband applications. However, the requirement of two parallel optical links inevitably increases the system’s complexity. To date, a broadband microwave photonic mixer with a tunable phase shift and supporting dispersion-induced power-fading-free fiber transmission based on a compact structure is still absent.

In this paper, a broadband microwave photonic mixer with flexibly tunable phase shift and supporting dispersion-induced power-fading-free fiber transmission is proposed and experimentally demonstrated. In the proposed scheme, two replicas of the IF/RF signals from a power splitter (PS) and two quadrature copies of the LO signals from an electrical 90° HC are used to drive the four sub-MZMs in a DP-DPMZM via carrier-suppressed double-sideband (CS-DSB) modulation. Through the mutual beating between the modulation sidebands of the IF/RF signal and the LO signal in a high-speed photodetector (PD), broadband frequency mixing is achieved. The phase shift of the frequency-converted signal can be continuously tuned within a range of 360° by synchronously adjusting the bias voltages of the parent-MZMs in the two sub-DPMZMs of the DP-DPMZM. In addition, the dispersion-induced power fading is compensated for by properly setting the bias voltage difference between the two parent-MZMs in the two sub-DPMZMs. To the best of the authors’ knowledge, this is the first demonstration of a broadband microwave photonic mixer that is capable of achieving full-range phase tuning and dispersion-induced power-fading-free fiber transmission.

2. Operation Principle

Figure 1 shows the schematic diagram of the proposed microwave photonic mixer. Continuous-wave (CW) light from a distributed-feedback laser diode (DFB-LD) is modulated by the IF/RF signal and the LO signal via a single DP-DPMZM. Specifically, the IF/RF signal and the LO signal are split into two parts by using a PS and an electrical 90° HC, respectively. These replicas are applied to the four sub-MZMs in the DP-DPMZM, i.e., MZM1, MZM3, MZM2 and MZM4, as depicted in Figure 1, where each sub-MZM is biased at its minimum transmission point to achieve CS-DSB modulation. Then the optical signals from the X-DPMZM and the Y-DPMZM are polarization division multiplexed by using a 90° polarization rotator (PR) and a polarization beam combiner (PBC), both of which are integrated in the DP-DPMZM. After propagation through a spool of single-mode fiber (SMF), the optically carried IF/RF and LO signals are distributed to the remote site. At the remote site, the two orthogonally polarized optical signals are detected by using a high-speed PD to generate two groups of signals, where the two groups of signals are superimposed to obtain the frequency-converted signals. An electrical bandpass filter (EBPF) with a specific center frequency is used to pick out the required frequency-converted signal. In the proposed scheme, through properly setting the bias voltage difference between the two parent-MZMs in the two sub-DPMZMs, the frequency response of the fiber link based on the two orthogonally polarized optical signals are complementary with each other, as shown in Figure 1b. Hence, the proposed microwave photonic mixer supports dispersion-induced power-fading-free transmission. In addition, through synchronously setting the bias voltages of the parent-MZMs in the two sub-DPMZMs, the phase of the frequency-converted signal can be continuously tuned within a range of 360°.

Mathematically, the output optical fields polarized in X direction and Y direction from the DP-DPMZM can be written as follows:

$$\left[\begin{array}{l}{E}_{\mathrm{X}}\\ E_{\mathrm{Y}}\end{array}\right]=\frac{\sqrt{2}}{4}{E}_{\mathrm{in}}\left[\begin{array}{l}\cos \left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\cos \left({\omega }_{\mathrm{S}}t\right)+\frac{\pi }{2}\right)+\cos \left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\cos \left({\omega }_{\mathrm{LO}}t\right)+\frac{\pi }{2}\right)\exp \left(-j{\phi }_{\mathrm{X}}\right)\\ \cos \left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\cos \left({\omega }_{\mathrm{S}}t\right)+\frac{\pi }{2}\right)+\cos \left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\sin \left({\omega }_{\mathrm{LO}}t\right)+\frac{\pi }{2}\right)\exp \left(-j{\phi }_{\mathrm{Y}}\right)\end{array}\right]$$

(1)

where E_in = E₀ exp (jω_ct) is the optical filed of the CW light from the DFB-LD; m_s = πV_s/V_π is the modulation index of MZM1 and MZM3; m_LO = πV_LO/V_π is the modulation index of MZM2 and MZM4; V_s and V_LO are the voltage amplitudes of the IF/RF signal and the LO signal, respectively; V_π is the RF half-wave voltage of the sub-MZMs; and ω_s and ω_LO are the angular frequencies of the IF/RF signal and the LO signal, respectively. Under small-signal modulation, the output optical fields from the DP-DPMZM can be simplified to the following:

$$\left[\begin{array}{l}{E}_{\mathrm{X}}\\ E_{\mathrm{Y}}\end{array}\right]\propto \frac{\sqrt{2}}{4}{E}_{\mathrm{in}}\left[\begin{array}{l}\left[\begin{array}{l}-J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\exp \left(j{\omega }_{\mathrm{S}}t\right)-J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\exp \left(-j{\omega }_{\mathrm{S}}t\right)\\ -J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\exp \left(j{\omega }_{\mathrm{LO}}t\right)\exp \left(-j{\phi }_{\mathrm{X}}\right)-J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\exp \left(-j{\omega }_{\mathrm{LO}}t\right)\exp \left(-j{\phi }_{\mathrm{X}}\right)\end{array}\right]\\ \left[\begin{array}{l}-J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\exp \left(j{\omega }_{\mathrm{S}}t\right)-J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\exp \left(-j{\omega }_{\mathrm{S}}t\right)\\ +J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\exp \left[j\left({\omega }_{\mathrm{LO}}t+\frac{\pi }{2}\right)\right]\exp \left(-j{\phi }_{\mathrm{Y}}\right)+J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\exp \left[-j\left({\omega }_{\mathrm{LO}}t+\frac{\pi }{2}\right)\right]\exp \left(-j{\phi }_{\mathrm{Y}}\right)\end{array}\right]\end{array}\right]$$

(2)

where J_n(·) is the nth-order Bessel function of the first kind; φ_X = πV_bX/V_π0 and φ_Y = πV_bY/V_π0 represent the bias-induced phase of the parent-MZMs in X-DPMZM and Y-DPMZM, respectively; V_bX and V_bY are the direct-current (DC) bias voltages applied to the parent-MZMs in X-DPMZM and Y-DPMZM, respectively; and V_π0 is the DC half-wave voltage of the parent-MZMs in X-DPMZM and Y-DPMZM. After propagation through a spool of SMF, the output optical fields can be calculated as follows:

$$\left[\begin{array}{l}{E}_{\mathrm{SMF}}{}_x\\ E_{\mathrm{SMF}}{}_y\end{array}\right]\propto \frac{\sqrt{2}}{2}{E}_{\mathrm{in}}\left[\begin{array}{l}-J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\cos \left({\omega }_{\mathrm{S}}t\right)\exp \left(j{\beta }_{2}L{\omega }_{\mathrm{S}}^2/2\right)-J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\cos \left({\omega }_{\mathrm{LO}}t\right)\exp \left(j{\beta }_{2}L{\omega }_{\mathrm{LO}}^2/2\right)\exp \left(-j{\phi }_{\mathrm{X}}\right)\\ -J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\cos \left({\omega }_{\mathrm{S}}t\right)\exp \left(j{\beta }_{2}L{\omega }_{\mathrm{S}}^2/2\right)-J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\sin \left({\omega }_{\mathrm{LO}}t\right)\exp \left(j{\beta }_{2}L{\omega }_{\mathrm{LO}}^2/2\right)\exp \left(-j{\phi }_{\mathrm{Y}}\right)\end{array}\right]$$

(3)

where β₂ is the GVD coefficient of the SMF with a unit of ps²/km, and L is the length of the SMF. After photodetection, the current from the PD can be calculated as follows:

$$\begin{array}{cc}i\left(t\right)=\hfill & i_x\left(t\right)+i_y\left(t\right)\propto E_{\mathrm{SMF}x}\times E_{\mathrm{SMF}x}^{\ast }+E_{\mathrm{SMF}y}\times E_{\mathrm{SMF}y}^{\ast }\hfill \\ \hfill =& \frac{1}{2}{E}_0^2\left\{\begin{array}{c}\left[\begin{array}{l}\frac{1}{2}\left(J_1^2\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)+J_1^2\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)+J_1^2\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\cos \left(2{\omega }_{\mathrm{S}}t\right)\right)\\ +J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\cos \left[\left({\omega }_{\mathrm{LO}}-{\omega }_{\mathrm{S}}\right)t\right]\cos \left[{\beta }_{2}L\left({\omega }_{\mathrm{LO}}^2-{\omega }_{\mathrm{S}}^2\right)/2-{\phi }_{\mathrm{X}}\right]\\ +J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\cos \left[\left({\omega }_{\mathrm{LO}}+{\omega }_{\mathrm{S}}\right)t\right]\cos \left[{\beta }_{2}L\left({\omega }_{\mathrm{LO}}^2-{\omega }_{\mathrm{S}}^2\right)/2-{\phi }_{\mathrm{X}}\right]\end{array}\right]+\\ \left[\begin{array}{l}\frac{1}{2}\left(J_1^2\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)+J_1^2\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)+J_1^2\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\cos \left(2{\omega }_{\mathrm{S}}t\right)\right)\\ +J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\sin \left[\left({\omega }_{\mathrm{LO}}-{\omega }_{\mathrm{S}}\right)t\right]\cos \left[{\beta }_{2}L\left({\omega }_{\mathrm{LO}}^2-{\omega }_{\mathrm{S}}^2\right)/2-{\phi }_{\mathrm{Y}}\right]\\ +J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\sin \left[\left({\omega }_{\mathrm{LO}}+{\omega }_{\mathrm{S}}\right)t\right]\cos \left[{\beta }_{2}L\left({\omega }_{\mathrm{LO}}^2-{\omega }_{\mathrm{S}}^2\right)/2-{\phi }_{\mathrm{Y}}\right]\end{array}\right]\end{array}\right\}\hfill \end{array}$$

(4)

By setting the DC bias voltages of the parent-MZMs in X-DPMZM and Y-DPMZM to satisfy φ_Y = φ_X − π/2, Equation (4) can be simplified to the following:

$$i\left(t\right)\propto \frac{1}{2}{E}_0^2\left\{\begin{array}{c}{J}_1^2\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)+J_1^2\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)+J_1^2\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)\cos \left(2{\omega }_{\mathrm{S}}t\right)\\ +J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\cos \left[\left({\omega }_{\mathrm{LO}}-{\omega }_{\mathrm{S}}\right)t+{\beta }_{2}L\left({\omega }_{\mathrm{LO}}^2-{\omega }_{\mathrm{S}}^2\right)/2-{\phi }_{\mathrm{X}}\right]\\ +J_1\left(\frac{m_{\mathrm{S}}}{2\sqrt{2}}\right)J_1\left(\frac{m_{\mathrm{LO}}}{2\sqrt{2}}\right)\cos \left[\left({\omega }_{\mathrm{LO}}+{\omega }_{\mathrm{S}}\right)t+{\beta }_{2}L\left({\omega }_{\mathrm{LO}}^2-{\omega }_{\mathrm{S}}^2\right)/2-{\phi }_{\mathrm{X}}\right]\end{array}\right\}$$

(5)

Equation (5) indicates that the input IF/RF signal at ω_s is converted to |ω_LO + ω_s| and |ω_LO − ω_s| through frequency mixing with the LO signal at ω_LO. The phase shift of the frequency-converted signals at |ω_LO + ω_s| and |ω_LO − ω_s| can be tuned through synchronously adjusting the DC bias voltages, V_bX and V_bY. Since both V_bX and V_bY can be adjusted from 0 to 2 V_π0, and the phase shift of the frequency-converted signals at |ω_LO + ω_s| and |ω_LO − ω_s| can be tuned within a range of 360°. In addition, it can be seen from Equation (5) that there is no dispersion-induced power fading of the frequency-converted signals at |ω_LO + ω_s| and |ω_LO − ω_s| if the relationship of φ_Y = φ_X − π/2 is maintained during the phase tuning.

3. Experiment and Discussion

A proof-of-concept experiment was carried out to verify the feasibility of the proposed scheme. In the experiment, a DFB-LD (Innovoton, INNO9303-220M, Chongqing, China) at 1545.43 nm and with an output optical power of 16 dBm was used to generate the optical carrier. A DP-DPMZM (Fujitsu, FTM7977HQ, Kawasaki, Japan) with a 3 dB modulation bandwidth of 23 GHz and an RF half-wave voltage of 3.5 V was employed to load the IF/RF and LO signals onto the optical carrier. Thereinto, the IF/RF signal and the LO signal were equally divided by using an electrical power splitter (Talent Microwave, RS2W10400-K, Suzhou, China) with an operation bandwidth from 1 GHz to 40 GHz and an electrical 90° HC (Marki Microwave, QH-0440, Morgan Hill, CA, USA) with an operation bandwidth from 4 GHz to 40 GHz, respectively. Three DC voltage-stabilized power supplies (RIGOL, DP832A, Suzhou, China) with voltage tuning accuracy of 1 mV were used to precisely control the DC bias voltages of the DP-DPMZM. A spool of SMF with a GVD coefficient of −23 ps²/km at 1550 nm and a length of 20 km was adopted to distribute the optically carried signal. In addition, an erbium-doped optical fiber amplifier (EDFA, Amonics, AEDFA-PA-35-B-FA, Hong Kong, China) and a tunable optical bandpass filter (OBPF, Santec, OTF-350-W-S-FC-A, Komaki, Japan) were used after the fiber transmission to compensate for the fiber link loss and to suppress the out-of-band amplified spontaneous emission (ASE) noise from the EDFA. By tuning the gain of the EDFA, the input optical power of the PD (Discovery, DSC-10H, Marlboro, NJ, USA) was maintained at 5 dBm, where the PD has a 3 dB bandwidth of 40 GHz and a responsivity of 0.6 A/W. The spectra of the optically carried signals and the frequency-converted signals were measured by using an optical spectrum analyzer (OSA, Yokogawa, AQ6370D, Tokyo, Japan) and an electrical spectrum analyzer (ESA, R&S, FSW67, Munich, Germany), respectively.

3.1. Frequency Conversion Efficiency

A single-tone IF signal at 5 GHz and with a power of 7 dBm was up-converted to 22 GHz and 32 GHz by using an LO signal at 27 GHz and with a power of 11 dBm. Figure 2a,b present the optical spectrum from the DP-DPMZM and the electrical spectrum of the output signal after the frequency mixing, where all the sub-MZMs in the DP-DPMZM are biased at their minimum transmission points, and the parent-MZMs in the two sub-DPMZMs are biased at their maximum transmission point and quadrature transmission point, respectively, to guarantee φ_Y = φ_X − π/2. It can be seen from Figure 2a that the carrier suppression ratio of the CS-DSB modulation is about 6.1 dB, which is attributed to the limited extinction ratios (ERs) of the sub-MZMs in the DP-DPMZM. Hence, there is a vestigial LO signal after frequency mixing, as depicted in Figure 2b. The power of the frequency-converted signals at 22 GHz and 32 GHz is measured to be −26.02 dBm and −25.88 dBm, respectively.

Figure 3a shows the conversion efficiency of the microwave photonic mixer working in the down-conversion mode, where the RF signals in the range of 9 GHz to 19 GHz are down-converted to the range of 1 GHz to 11 GHz by using an LO signal at 8 GHz. Figure 3b presents the conversion efficiency of the microwave photonic mixer working in the up-conversion mode, where an IF signal at 5 GHz is up-converted to the range of 11 GHz to 21 GHz by using the LO signals in the range of 6 GHz to 16 GHz. It can be seen from Figure 3 that the conversion efficiency of the microwave photonic mixer is −32.69 ± 0.91 dB and −32.44 ± 0.63 dB for the down-conversion mode and the up-conversion mode, respectively. The frequency conversion efficiency can be further improved by using a high-power DFB-LD, a DP-DPMZM with a low half-wave voltage and a low insertion loss, and a PD with high optical power handling capability.

3.2. Suppression of Dispersion-Induced Power Fading

The microwave photonic mixer worked in the up-conversion mode to verify the ability to eliminate the dispersion-induced power fading in the fiber transmission. A single-tone IF signal at 5 GHz and with a power of 7 dBm was up-converted to the range of 10.5 GHz to 42 GHz by using LO signals in the range of 5.5 GHz to 37 GHz and with an identical power of 11 dBm. Figure 4 exhibits the frequency response of the microwave photonic mixer after the 20 km SMF transmission. Thereinto, the measurement results denoted by the black squares were obtained by biasing the parent-MZMs in the two sub-DPMZMs at their maximum transmission point and quadrature transmission point, respectively, which corresponds to φ_Y = φ_X − π/2. The measurement results denoted by the red circles were obtained by biasing parent-MZMs in the two sub-DPMZMs at their maximum transmission points, which correspond to φ_Y = φ_X. It can be clearly seen from Figure 4 that the proposed scheme successfully eliminated the serious dispersion-induced power fading at the vicinity of 19 GHz, 28.5 GHz, 34.5 GHz and 40 GHz. The vestigial frequency-dependent power fading in the case of φ_Y = φ_X − π/2 is attributed to the parameter deviations and the frequency response of the devices. Thereinto, the parameter deviations include the limited extinction ratio and the bias-voltage drift of the DP-DPMZM, and the power and phase imbalance of the 90° HC and the PS. A detailed analysis of this issue is given as follows.

The extinction ratios of the parent-MZM and the sub-MZMs in the DP-DPMZM were measured to be 22 dB and 20 dB, respectively. The maximum bias-voltage deviations of the four sub-MZMs in the DP-DPMZM were measured to be 7%, 4%, 14% and 4%, respectively. In addition, the frequency response of the PS and the 90° HC were measured by using a vector network analyzer (VNA, R&S, ZNA67, Munich, Germany), where the maximum phase deviation and amplitude imbalance are ±6° and ±0.4 dB for the PS, and ±5° and ±0.4 dB for the HC. Thereinto, the amplitude imbalance of ±0.4 dB corresponds to a power imbalance ratio of 0.45 to 0.55. Figure 5a–f show the frequency response of the microwave photonic mixer obtained by carrying out numerical calculation with one parameter deviation under the worst cases and all other parameters holding at ideal levels, respectively. It can be seen from Figure 5 that, although the frequency-dependent power fading induced by independent parameter deviation is small, the combined effect together with the influence of the frequency response introduced by the DP-DPMZM, the PD and the cables is obvious, as depicted in Figure 4. Hence, it is critically important to select devices with large bandwidths and small parameter deviations to enhance the frequency-response flatness of the microwave photonic mixer with long-distance fiber transmission. In addition, the vestigial frequency-dependent power fading of Figure 4 is repeatable, and it can be compensated by increasing the power of the LO signal.

3.3. Phase-Shift Tunability

The phase-shift tunability of the microwave photonic mixer is evaluated by using a four-port VNA (VNA, R&S, ZNA67, Munich, Germany) based on the vector mixing measurement method, as depicted in Figure 6. The VNA is firstly calibrated by using a set of calibration kits and a calibration mixer (R&S, ZN-ZM292, Munich, Germany) in unknown through mode. In the measurement, Port 1 and Port 4 of the VNA provide the IF/RF signal and the LO signal for the microwave photonic mixer, respectively. The frequency-converted signal is sent to Port 2 of the VNA to achieve measurement. The power of the IF/RF signal and the LO signal are set to be −6 dBm and 15 dBm, respectively. In addition, V_bX and V_bY are synchronously adjusted to tune the phase shift of the frequency-converted signal and simultaneously guarantee φ_Y = φ_X − π/2.

Four phase-shift values with a step of 90° are preset in the measurement. Figure 7a,b present the power response and the phase response of the microwave photonic mixer working in the down-conversion mode, respectively, where the RF signals in the range of 10.1 GHz to 19.9 GHz are down-converted to the range of 0.1 GHz to 9.9 GHz by using an LO signal at 10 GHz. Figure 8a,b exhibit the power response and the phase response of the microwave photonic mixer working in the up-conversion mode, respectively, where the IF signals in the range of 1 GHz to 9.9 GHz are up-converted to the range of 11 GHz to 19.9 GHz by using an LO signal at 10 GHz. Figure 9a,b show the power response and the phase response of the microwave photonic mixer working in the up-conversion mode, respectively, where the IF signals in the range of 1 GHz to 9.9 GHz are up-converted to the range of 21 GHz to 29.9 GHz by using an LO signal at 20 GHz. These results indicate that the proposed microwave photonic mixer can achieve a broadband phase-shift tuning without obvious power variation even after long-distance fiber transmission. Finally, it should be pointed out that the phase-shift-tuning range can be enlarged beyond 360° through synchronously tuning V_bX and V_bY over a range larger than 2 V_π0. In addition, the bandwidth of the continuously adjustable phase shift can be further improved by using a DP-DPMZM with a larger analog bandwidth.

4. Conclusions

In summary, we proposed and demonstrated a broadband microwave photonic mixer based on a DP-DPMZM, which can simultaneously support flexible phase-shift tuning and long-distance fiber transmission. The IF/RF signal and the LO signal are applied to the four sub-MZMs biased at their minimum transmission points via a PS and a 90° HC, respectively. Through mutual beating between the IF/RF and the LO modulation sidebands in a high-speed PD at the remote site, up-conversion and down-conversion are achieved. By setting the bias-induced phase difference between the parent-MZMs in the two sub-DPMZMs of the DP-DPMZM to be π/2, the dispersion-induced power fading over long-distance fiber transmission is eliminated. Through synchronously adjusting the DC bias voltages of the parent-MZMs in the two sub-DPMZMs, the phase shift of the frequency-converted signal can be continuously tuned in a range of 360°. In the experiment, a microwave photonic mixer with a 6 dB operation bandwidth of 40 GHz and supporting dispersion-induced power-fading-free transmission over 20 km SMF was realized. In addition, a continuously tunable phase shift over 360° in the frequency range of 0.1 GHz to 29.9 GHz was also demonstrated, where the power variation during phase tuning was measured to be smaller than 4 dB. The proposed scheme can find applications in the DAA-based wide-range radar systems.