1. Introduction
In recent years, the development of microwave photonic technology, which combines the advantages of flexible access in the microwave field and fine regulation in the optical field, has attracted wide attention in various fields [
1,
2]. Compared with traditional electrical technology, microwave photonic technology features broadband operation ability, electromagnetic immunity and optical fiber transmission compatibility, as well as small size, light weight and low transmission loss [
3,
4]. In a typical example of applying microwave photonic technology to realize a variety of signal processing functions, frequency mixing is an essential function and has great application value [
5]. Specifically, converting high-frequency signals down to intermediate frequency (IF) is particularly helpful in reducing the sampling rate requirements for the subsequent analog-to-digital (ADC) conversion and digital signal processing (DSP) [
6]. In addition, the combination of microwave photonic technology and in-band full-duplex (IBFD) systems can provide full play to the parallel processing capability and wideband multi-dimensional parameter tuning capability of photonics, as well as the high-spectrum utilization efficiency of IBFD technology, thus becoming a typical example to solve the contradiction between limited spectrum resources and high data rates [
7,
8]. Therefore, photonic-enabled in-band full-duplex frequency down-conversion systems have gained great attention in RF transceivers for wireless communication systems, radar frontends and satellite payloads [
9]. However, an urgent problem to be solved in the IBFD frequency down-conversion system is the disturbance of in-band interference containing unwanted image interference and self-interference [
10,
11]. Image interference and self-interference (SI) will inevitably be captured by the antennas of RF transceivers. These two kinds of in-band interference will be spectral-aliased with the desired IF signal after mixing with the local oscillation (LO) signal, which leads to the distortion of the desired IF signal. Therefore, an image rejection mixer (IRM) with a self-interference cancellation (SIC) ability is the key to effectively solving the Gordian knot and achieving efficient signal sending and receiving.
Previously, several photonics-based schemes to eliminate these two kinds of in-band interference have been demonstrated [
12,
13,
14,
15], which are very instructive. However, the SIC function in Refs. [
12,
13] was realized by virtue of electrical components to achieve time delay and amplitude matching between SI and reference signals. Since the cancellation depth is greatly affected by the time delay and amplitude matching, the SIC performance based on electrical components lags behind schemes applying optical devices with high tuning precision due to its limited tuning accuracy [
16]. In Refs. [
14,
15], electrically tuned devices were avoided, thus guaranteeing the wideband deep cancellation of SI and image interference. However, there are many tuning devices and several discrete paths in Ref. [
14]. It has four independent paths that need to be tuned and matched, not only increasing the difficulty of parameter tuning but also increasing the complexity, which is not conducive to practical applications. In Ref. [
15], an incoherent optical system was applied, which inherently limits the compactness and simplicity of the system [
17]. Therefore, an IRM with simultaneous SIC features of ultrawideband operation, deep cancellation and compact configuration is desired for practical applications.
In this paper, a photonic-enabled IRM with simultaneous ultrawideband SIC is proposed. It is constructed in a single optical path structure based on a coherent optical source. The phase reversion and amplitude matching of SIC operation are realized by using optical polarization regulation, and the time delay matching is realized by using an optical high-precision time delay device, making SI directly canceled in the optical domain. A pair of orthogonal LO signals can be constructed by controlling the phase shift introduced by the DC bias without requiring a bandwidth-limited high-frequency electrical 90-degree hybrid coupler (HC) so that the wideband image rejection (IR) operation can be achieved through the typical phase cancellation method. This scheme not only avoids the use of high-frequency electrical devices with limited tuning precision and bandwidth, but also simplifies system complexity. Therefore, the SI and image interference can be simultaneously canceled in a compact configuration. Its application can promote high-efficiency communication under the strong interference of the IBFD system.
2. Principle
Figure 1 shows a schematic diagram of the proposed IRM with simultaneous SIC. A coherent optical structure based on a main polarization-division-multiplexing dual-parallel Mach–Zehnder modulator (PDM-DPMZM) is constructed. An optical carrier is generated from a laser diode (LD), the expression of which is described as
Ec(
t) =
Ecexpj
ωct. Ec and
ωc donate the optical carrier amplitude and angular frequency. After being adjusted by a polarization controller (PC) to be aligned at the main axis of the PDM-DPMZM, the optical carrier is injected into the modulator. There are two dual-parallel Mach–Zehnder modulators (DPMZMs) embedded in the PDM-DPMZM, and the modulated signals of the two DPMZMs are polarization-orthogonal at the output of the PDM-DPMZM due to the integration of a 90-degree polarization rotator (90°PR) and a polarization beam combiner (PBC). Each DPMZM has two single-drive Mach–Zehnder modulators (MZMs) embedded in the upper and lower arms of a Mach–Zehnder interferometer (MZI), having two RF input ports, two sub-modulator DC bias points and a main modulator DC bias point. In the DPMZM
x, the impaired received signal containing the desired signal of interest (SOI) and unwanted in-band interference, namely, image interference and SI, is injected into the upper sub-modulator of DPMZM
x (MZM
x1); meanwhile, the LO signal is injected into the lower sub-modulator (MZM
x2). The reference (REF) signal is divided equally into two paths and injected into the two sub-modulators of the DPMZM
y (MZM
y1 and MZM
y2). Therefore, when the electrical signals injected into the four RF input ports are represented as
Vx1(
t),
Vx2(
t),
Vy1(
t) and
Vy1(
t), the output signals of the four sub-MZMs in PDM-DPMZM can be expressed as:
where
φx1/2 and
φy1/2 equal π
Vbiasx1/2/
Vπ and π
Vbiasy1/2/
Vπ, which are the DC-bias-induced phases of MZM
x1/2 and MZM
y1/2, respectively.
Vbiasx1/2 and
Vbiasy1/2 are the DC bias voltages of MZM
x1/2 and MZM
y1/2, respectively
. EMZMx1(
t),
EMZMx2(
t) and
EMZMy1(
t),
EMZMy2(
t) represent the output signals of four MZMs.
Vπ is the voltage corresponding to the additional phase of π, which is called the half-wave voltage. j represents an imaginary unit.
IL is the insertion loss of the modulator.
γ denotes the power splitting ratio of both Y-branch waveguides (assumed to be symmetrical) in each MZM (four MZMs are assumed to have the same parameters), which is given by:
where
ExtRatio is linked to the parameter extinction ratio (ER) (this is assumed to be the ON–OFF optical extinction ratio of the MZM modulator).
The output of DPMZM is similar in principle to that of MZM. We assumed that two DPMZMs have the same ER and the spectral ratio obtained according to Equation (2) is
γ1. Therefore, the output signals of the two DPMZMs are represented as:
where
φx3 and
φy3 equal π
Vbiasx3/
Vπ and π
Vbiasy3/
Vπ, which are the DC-induced phases of the main modulator of DPMZM
x and DPMZM
y. Vbias3 and
Vbiasy3 are the DC bias voltages of the main modulator of DPMZM
x and DPMZM
y, respectively
. EDPMZMx(
t) and
EDPMZMy(
t) represent the output signals of DPMZM
x and DPMZM
y, respectively.
The modulated signals of the two DPMZMs are polarization-orthogonal at the output of the PDM-DPMZM due to the integration of a 90°PR and a PBC. Assuming that the spectral ratio of the whole PDM-DPMZM is obtained according to Equation (2) as
γ2, the polarization-multiplexed signal output from the PDM-DPMZM can be expressed as:
The SOI, SI, image and LO signals are denoted as
VSOIsin
ωSOI(
t + τSOI),
VSIsin
ωSI(
t + τSI),
Vimagesin
ωimage(
t + τimage) and
VLOsin
ωLO(
t + τLO).
Vi,
ωi and
τi are the amplitude, angular frequency and initial arrival time of the SOI, SI, image interference and LO signals (
i represents SOI, SI, image and LO). The reference (REF) signal is expressed as
VREFsin
ωREF(
t + τREF). Similarly,
VREF,
ωREF and
τREF are the amplitude, angular frequency and initial arrival time of the REF signal. As described above, the electrical signals of the four RF inputs are as follows. The spectrum diagrams of the input signals are shown in
Figure 2a–c.
By substituting Equation (5) into Equation (4) and using the modulation index
βi to replace π
Vi/Vπ (
i can be replaced by SOI, SI, image, LO and REF), the output of PDM-DPMZM can be written as:
The sub-modulators in both the upper and lower DPMZM all operate at the minimum transmission point (MITP) to suppress the optical carrier without loading information, thus increasing the power of modulated sidebands. Therefore, the phases introduced by the DC biases of the sub-modulators are all π. That is,
φx1/2 and
φy1/2 are π.
φx3 and
φy3 are the DC-induced phases of the main modulator of DPMZM
x and DPMZM
y.
φx3 equals 0 to maximize the power of REF-modulated sidebands. By substituting the DC-bias-induced phase into Equation (6) and expanding Equation (6) based on the Bessel function to reserve the first-order sidebands for simplified analysis, Equation (6) is transferred into Equation (7). The Jacobi–Anger expansion (Bessel function) is an expansion of exponentials of trigonometric functions on the basis of their harmonics [
18,
19]. Here, it is also used to describe the spectral amplitudes of harmonics resulting from intensity modulation through electro-optic conversion. J
n(
β) represents the
nth-order Bessel function of the first kind. The spectrum diagrams of the modulated signals from DPMZM
x and DPMZM
y are shown in
Figure 2d,e.
Then, the polarization orthogonal signals output by PDM-DPMZM will enter a polarization beam splitter (PBS) to be polarization-demultiplexed, and an optical time delay line (OTDL) will realize high-precision time delay tuning of the REF signal in the lower path. The polarization-demultiplexed signals will then enter the PBC for secondary polarization beam combination and be combined through an adjustable polarizer (Pol) with an angle,
α. Therefore, the output signal of the Pol can be expressed as:
In Equation (8), the first brace is the residual optical carrier; the second brace contains SOI-, image-, and LO-modulated sidebands; and the third brace contains SI- and REF-modulated sidebands. Therefore, it can be seen from Equation (8) that the SIC conditions of amplitude matching and phase reversion can be achieved by adjusting the angle of Pol, and the time delay matching can be achieved by tuning the optical time delay line (OTDL). The manual OTDL commonly used at present is usually a passive device tuned by changing the length of the propagation path. The basic structure of an OTDL is packed with two fiber optic collimators and a reverse reflector on a movable platform [
20]. The position of the movable reflector can be adjusted with some adjusting knobs so as to change the propagation path of the optical signal in the device and realize the adjustment of the time delay. Specifically, when the conditions shown in Equation (9) are satisfied, the SIC signal can be canceled directly in the optical domain.
The conditions for solving the parameters can be rewritten as:
As a result, through optical parameter regulation, the SI-modulated sidebands and REF-modulated sidebands in the third brace cancel each other out, leaving only the items in the first and second brace. A spectrum diagram output from the Pol is shown in
Figure 2f. As can be seen from
Figure 2f, after the adjustment of the Pol and optical time delay line, the REF-modulated sidebands (blue checked squares) and SI-modulated sidebands (solid green squares) are in reverse phase; then, amplitude and delay matching are satisfied, SI-modulated sidebands can be eliminated and, thus, SIC can be achieved directly. Replacing the residual carrier component with
ξ (
ξ = cos
α(2
γ−1)
γ2[
γ1J
0(
βSOI)J
0(
βSI) J
0(
βimage) + (1−
γ1)exp(j
φ3) J
0(
βLO)] + sin
α(2
γ−1)(1−
γ2)J
0(
βREF)) to simplify the expression and substituting Equation (10) into Equation (8), the output can be obtained as:
Then, the remaining signals are amplified by an erbium-doped fiber amplifier (EDFA) and injected into wavelength division multiplexing (WDM). Through WDM, the positively and negatively modulated sidebands of the SOI, image and LO signals are separated into two paths. Assuming that the passband responsiveness of WDM is ideal and the optical carrier falls on the edge of the passband response, the two signals output by WDM can be expressed as:
where
η1 and
η2 are the amplitude attenuation coefficients of the optical carrier in two paths after filtering by WDM. The process of going through WDM is shown in
Figure 2g–i. The separated signals are converted into photocurrents by two photodetectors (PDs) with the same responsivity. The electrical signals obtained in the two paths are then combined in a 90-degree hybrid coupler (90°HC), the output of which can be expressed as:
Here, the gain of the EDFA is
G, and the responsivity of the PDs is
R. As can be seen from Equation (14), when the phase introduced by the DC bias of the main modulator in DPMZM
x is 45 degrees, namely,
φx3 equals π/4, the image interference can be canceled due to phase reversion and the desired IF signal is, thus, enhanced. Therefore, the useful IF signal, free of self-interference and image interference, can finally be obtained as Equation (15). The output of two PDs and the 90°HC is shown in
Figure 2j–l.
As can be seen from Equation (15), image rejection is performed. However, the non-ideal ER results in the higher power of the residual carrier, which finally leads to a high-power DC component and the leakage of RF and LO signals. Fortunately, the deterioration of ER has no effect on the SIC and IRM performances. The stray component can be filtered by the straighter and the electric filter. The deterioration of ER may cause changes in the spectral ratio, resulting in a change in useful signal power, but the change is within the acceptable range, which can be compensated for by an amplifier. Therefore, the scheme has good resistance to ER deterioration. In addition, MZMs with extremely high ER have been verified, which provides the possibility for further improving system performance [
21].
In addition, the operating bandwidth of the modulator may affect the signal-tuning frequency range. However, most of the current commercially available modulators have bandwidths above 23 GHz to meet the needs of general systems [
22]. With further improvement in manufacturing technology, the bandwidth of modulators is gradually improving [
23]. Superior performance modulators will help further extend the system bandwidth.
The conversion gain of the receiver, defined as the power ratio between the IF and RF signals, can reflect the loss and efficiency of the whole system to some extent, which can be calculated as [
24]:
where
PIF and
PRF are the powers of the IF and RF signals.
Rout and
Rin are the output- and input-matched impedance of the system. In principle, the conversion gain can be improved by applying PD with higher responsivity and a modulator with smaller half-wave voltage, as well as by increasing the gain of the amplifier, improving the optical carrier intensity of the laser output or setting the proper modulation index of the LO signal.