1. Introduction

Microwave frequency shift keying (FSK) signal, conveying the coding information through the hopping of its carrier frequency, has been widely used in wireless communication, radar, electronic countermeasures systems due to the excellent anti-jamming and anti-interception ability [1,2]. Traditionally, the electrical-based solutions to generate microwave FSK signal, constrained by the well-known electronic bottleneck, are faced with a crucial challenge raised by ever-increasing demand for microwave frequency and bandwidth. Alternatively, microwave photonics provides a promising way to overcome this limitation with the merit of improved frequency, expanded bandwidth, and immunity to electromagnetic interference, enabling the easier realization of high frequency and flexible tunable microwave signal generation [3–5]. Various photonic-assisted methods have been proposed to generate microwave FSK signal in recent years, which can be roughly divided into three types: filter-based, bias-switching-based and polarization-switching-based methods.

The filter-based methods use a tunable optical comb filter to select optical frequency tones [6,7], which are then recovered to microwave signals with different frequencies after optical-to-electrical (O/E) conversion. Since the application of optical comb filter, the system tunability is limited. Moreover, the optical frequency tones should be located in the peak, valley or medium place of the filter transfer function, thus a complex adjustment is needed and the sub-carriers of the generated microwave FSK signal have a fixed multiple relationship which cannot be tuned flexibly. For the bias-switching-based method [8,9], a single Mach-Zehnder modulator (MZM) is controlled by a binary coding sequence to switch the bias point between quadrature transmission point (QTP) and minimum transmission point (MITP). As a result, a microwave FSK signal can be obtained by quite simplified structure. However, the two sub-carriers are still unable to be tuned flexibly and have a double relationship. This form of microwave FSK signal generated in [5–8] occupies a large operating bandwidth since the two sub-carriers are far apart enough, making it suitable for military applications such as sophisticated electronic warfare system due to the stronger anti-jamming and anti-interception capabilities. Yet, this feature conversely limiting its applications in commercial communication or some scenarios where more flexible sub-carriers are desired.

For the polarization-switching-based method [10–13], a polarization modulator (PolM), driven by binary coding sequence, is used to switch the polarization direction of output optical signal, which are then inject into a polarization-dependent parallel structure [10–12] or a polarization maintaining fiber Bragg grating [13]. Optical signal at different polarization direction experiences different modulation processes, then a microwave FSK signal can be obtained after O/E conversion. There are some other methods to generate microwave FSK signal like the one in [14] and [15], only a polarization-division-multiplexed MZM is employed to act as an equivalent photonic switch. This form of microwave FSK signal generated in [10–15] have two flexibly tunable sub-carriers. However, compare with the first form, an additional radio frequency (RF) input signal or an additional tunable optical source is required.

Few schemes can realize the generation of these two forms of microwave FSK signal by a single structure. Recently, we have proposed a method to achieve it by a dual-polarization quadrature phase shift-keying modulator (DP-QPSKM) and a tunable filtering switch (TFS) [16]. By adjusting the system parameters, either of the two forms of microwave FSK signal can be obtained, making up for the deficiencies of single forms microwave FSK signal, enabling the scheme to accommodate more flexible and resource-constrained applications. Nevertheless, the use of TFS reduces the frequency tunability and increases the system complexity.

In this paper, we demonstrate a photonic-assisted generation scheme to generate the aforementioned two forms of microwave FSK signals based on a DP-QPSKM and a Sagnac loop structure. The filter-free system features more flexible tunability and better integration condition than the one in [16]. The core idea of the scheme is that the DP-QPSKM (functions as optical sidebands selector) provides two orthogonally-polarized high-order sidebands, and then the interference result of these two polarization states is controlled by the following Sagnac loop structure, so as to realize the generation of two forms of microwave FSK signals with high frequency multiplication factor. However, the benefit of this idea is not only to achieve two forms of microwave FSK signal generation in a single structure. Firstly, since the optical sidebands selector is separated from the Sagnac loop structure, the output of optical sidebands selector can be extracted to provide optical local oscillator for the realization of optical-domain demodulation. Based on this, for the first time to the best of our knowledge, we design a photonic-optimized coherent demodulation structure to perform demodulation of the received signal, which provides new insights into the all-optical processing of FSK-based communication systems. This design can effectively avoid the electronic bottleneck problem faced by traditional electrical methods in demodulating high-frequency and broadband microwave FSK signals. Secondly, the output of the optical sidebands selector can be divided into multiple channels and connected to multiple Sagnac loop structures modulated by different data, which can be applied to the latest MIMO-FSK communication system. In contrast, although other schemes with simpler structure [8,9,14,15] have advantages in terms of cost and system loss, it is difficult to realize the two extended functions mentioned above, limiting the further application at the system level.

2. Theory and principle

The diagram of the proposed multiform microwave FSK signal generator is schematically shown in Fig. 1(a). In this design, an optical carrier given by a laser diode (LD) is sent into a DP-QPSKM and then modulated by the microwave driving signal provided by microwave signal generator (MSG). The output optical signal is injected into a the subsequent Sagnac loop structure for further phase modulation. Finally, multiform microwave FSK signal can be obtained after O/E conversion by carefully adjusting the parameters of DP-QPSKM and the Sagnac loop structure, which will be analyzed in the next section 2.1-2.3.

 

Fig. 1. (a) Schematic diagram of the proposed multiform microwave FSK signal generator. (b) The inside view of QPSK modulator. LD: laser diode; DP-QPSKM: dual-polarization quadrature phase shift keying modulator; PC: polarization controller; OC: optical circular; PBS: polarization beam splitter; PM: phase modulator; 90° PR: 90° polarization rotator; Pol: polarizer; PD: photodetector; MSG: microwave signal generator; PPG: pulse pattern generator; PS: phase shifter; MZM: Mach-Zehnder modulator.

Download Full Size | PDF

2.1 Adjusting of DP-QPSKM

The DP-QPSKM comprises a 3dB coupler, two QPSK modulators, a 90° polarization rotator (PR), and a polarization beam combiner (PBC). In our scheme, QPSK modulator functions as optical sidebands selector, and its inside view, as illustrated in Fig. 1(b), comprises two sub-MZMs (MZM₁ and MZM₂) embedded in a main MZM. A sine waveform with an angular frequency of ω is equally divided into two paths by a power splitter, one path drives the MZM₁ directly, the other path is firstly phase shifted by φ through a phase shifter (PS) and then drives the MZM₂. For simplify, assuming that θ₁ and θ₂ respectively represent the bias phases of the two sub-MZMs and the main MZM, and the expression of the input lightwave is exp(jω_ct), where ω_c is the angular frequency. Then the output signal of QPSK modulator can be written as:

Table 1. the value of n that can meet cos(a-nb) = 0.

View Table | View all tables in this article

As a result, E_QPSKM(t) would present different results by control the two amplitude factors cos(θ₁-nπ/2) and cos(θ₂-nφ/2). For example, we firstly set θ₁= 0, the cos(θ₁-nπ/2) will equal to 0 at n=±1, ±3, ±5···, then we set θ₂= 0 and φ=π/2, the cos(θ₂-nφ/2) will equal to 0 at n=±2, ±6, ±10···. At this time, the E_QPSKM(t) only retains the optical sidebands at n = 0, ±4, ±8···. Considering the quite low power of the optical sidebands at n ≥ ±8, it could be suggested that the output signal only contains optical carrier and ±4th order sidebands. Likewise, when the parameter settings are θ₁= 0, θ₂=π/2, and φ=π/2, the output signal only contains ±2th order sidebands. These two situations will be used in our scheme, which are listed in Eq. (2) and Eq. (3). Apparently, the QPSK modulators function as an optical sidebands selector by control its input parameters.

2.2 Adjusting of Sagnac loop structure

The Sagnac loop structure comprises a polarization controller (PC₁), an optical circular (OC), a polarization beam splitter (PBS), a phase modulator (PM), a 90° PR and a polarizer (Pol). Firstly, the polarization direction of the input signal is rotated by the PC₁ to have an angle of α relative to the principle axis of the PBS. Afterwards, the input signal is polarization separated into two paths at the PBS, and then the two light beams propagate in opposite directions of the loop and interfere when they reach the PBS again. The PM embedded in the loop, driven by a binary coding sequence, is applied to introduce a phase shift β to the forward injected light. The 90° PR is used to ensure that the two optical lights can be output from the PBS. Finally, the Pol is applied to merge the two polarization separated signals into one path, with its angle is adjusted to 45° relative to the principle axis of the PBS. It is well known that the PM is a traveling wave modulator, in which only the forward transmitted light can be modulated and the reverse modulations are ignored because of the velocity mismatch [17]. Based on the above analysis, we can get the transmission matrix of the Sagnac loop structure as follows:

In Eq. (4), the β equals to πs(t)/2V_π, where s(t) denotes the binary coding sequence and V_π is the half-wave voltage of the PM. The employed Sagnac loop structure in our scheme ensures that the equal optical paths of clockwise and counterclockwise light which can enhance the robustness against environmental perturbations, making itself more stable than the schemes in [10–12].

2.3 Generation of multiform microwave FSK signal

When it comes to the generation of microwave FSK signal whose two sub-carriers have a fixed double relationship, two microwave signal with a phase difference of ϕ, namely sin(ωt) and sin(ωt+ϕ), are used to drive the two QPSK modulators respectively, and the rotated angle α is adjusted to 0°. Then the output optical signal after the Pol can be expressed as:

As can be seen from Eq. (5), a third amplitude factor cos(β/2-nϕ/2) is obtained compared with Eq. (1). Based on the conclusion in Table. 1, when the parameter settings are θ₁= 0, θ₂= 0, φ=π/2, ϕ=2π/3 and β=π, the E_out(t) only remains the ±4th-order sidebands, while if β changes to π/3 and other parameters stay the same, the E_out(t) only remains the +4th-order sideband and the optical carrier. Therefore, under the premise that θ₁= 0, θ₂= 0, φ=π/2, ϕ=2π/3, when the amplitudes of bit ‘1’ and bit ‘0’ in the s(t) are respectively set to V_π and V_π/3, that is, β=π for bit ‘1’ and β=π/3 for bit ‘0’, the Eq. (5) can be simplified as:

As can be seen, a microwave FSK signal is obtained with frequency multiplication factors of 4 for bit ‘0’ and 8 for bit ‘1’. It is noticeable that, in order to keep the uniform amplitude for the two subcarriers, the modulation index m should be adjusted to 2.30 or 2.56 to satisfy |J₀(m)|=|J₄(m)|.

When it comes to the generation of microwave FSK signal with flexibly tunable subcarriers, two microwave signal with different frequency, namely sin(ω₁t) and sin(ω₂t), are respectively used to drive the two QPSK modulators, and the rotated angle α is adjusted to 45°. Then the output optical signal after the Pol can be expressed as:

As can be seen from Eq. (8), if β is adjusted to 0, only the component E_x(t) is retained and the component E_y(t) is eliminated, while if β is adjusted to π, it has contrary results. Therefore, when the amplitudes of bit ‘1’ and bit ‘0’ in the s(t) are respectively set to V_π and 0, that is, β=π for bit ‘1’ and β=0 for bit ‘0’, the Eq. (8) can be simplified as:

Consequently, if we set θ₁= 0, θ₂= 0, and φ=π/2 for both the two QPSK modulators, then only the ±2nd-order sidebands are retained in E_x(t) and E_y(t) as expressed in Eq. (3), the Eq. (9) can be specifically written as:

After O/E conversion by the PD, the generated electrical signal ignoring DC components can be represented as:

From Eq. (11), we can see that a microwave FSK signal is obtained with its subcarrier jumps between four times ω₁ or four times ω₂, and the two subcarriers can be tuned flexibly and independently.

2.4 Photonic-optimized transceiver for wireless communication

In wireless communication, coherent demodulation is the commonly used and effective method for recovering the data carried on microwave FSK signal [18], whose diagram is given in Fig. 2(a). In this figure, the received signal is firstly divided to two paths and filtered by two parallel-connected bandpass filters (BPF), then each path is multipled by the local coherent carrier. After filtered by lowpass filters (LPF), the two paths are combined together and decided by digital signal processing (DSP) to recover the data. Assuming that f₀ and f₁ are the frequencies of the two sub-carriers of microwave FSK signal, then the center frequencies of two BPFs should follow f₀ and f₁ and the two local coherent carriers should be of the same frequency and in-phase with the signal. In the schemes of Ref. [8,9,11,12,16], photonic frequency multiplication can be realized, which is attractive and cost-effective to obtain high frequency microwave FSK signal and reduce the bandwidth requirement for electrical devices at transmitter. However, at the receiver, high frequency microwave FSK signal will significantly increase the bandwidth requirement of the BPF and mixer, and cause great trouble to the traditional extraction of local coherent carrier. These problems are not taken into account in the above schemes. In the proposed scheme, the output of DP-QPSKM can be extracted to demodulate the received signal in optical domain. As a result, the alleviated electronic bottlenecks, simplified receiving structure and real time processing can be achieved simultaneously. Figure 2(b) shows the configuration of the photonic-optimized transceiver.

 

Fig. 2. (a) Schematic diagram of electrical coherent demodulation. (b) Configuration of the proposed transceiver. EDFA: Erbium-doped optical fiber amplifier; OTDL: optical tunable delay line; LPF: lowpass filter; DSP: digital signal processing; EA: electrical amplifier; BPF: bandpass filter; PBC: polarization beam combiner.

Download Full Size | PDF

In the system, the DP-QPSKM is configured to generate two orthogonally polarized optical signals with specific sidebands. The output lightwave is firstly amplified by an Erbium-doped optical fiber amplifier (EDFA) and then equally divided to two parts by an optical splitter. One part is fed into transmitter after a further phase modulation in the Sagnac loop structure. In the transmitter, a PD performs O/E conversion to obtain the microwave FSK signal, whose power is then boosted by an electrical amplifier (EA1) and emitted for wireless transmission. The other part is fed into the receiver. In the receiver, an optical tunable delay line (OTDL) is used to introduce a time delay τ to ensure the phase synchronization with the received signal. Meanwhile, the communication signal received by an antenna is injected into the RF ports of dual-polarization MZM (DPol-MZM) after amplified. The output of DPol-MZM is sent to a polarization separated structure containing two parallel-connected PD. Both the PC₂ and PC₃ are adjusted to ensure the polarization alignment between the lightwave and their following PBS. As a result, the lightwave in x-axis transmit along the upper MZM and the upper PD, contrarily, the lightwave in y-axis transmit along the lower MZM and the lower PD. Both the two MZMs are biased at quadrature transmission point. Finally, the output signal from PD is sampled and decided by DSP after filtered by an LPF.

Suppose two identical communication terminals, with an aligned direction of the antennas maximum gain, are employed to transmit and receive microwave FSK signal. When the communication signal chooses the microwave FSK signal whose two sub-carriers have a fixed double relationship, just as the signal in Eq. (7), the received signal output from EA₂ can be given by:

If the received signal is the frequency component at 4ω, then the output of upper MZM is:

Note that, small signal modulation is assumed and the sidebands higher than 1st-order are ignored. Then the output optical signal is converted to electrical signal by the upper PD and filtered by an LPF. Consequently, the signal output from the upper LPF can be expressed as:

In addition, if the received signal is not the frequency component at 4ω or 8ω, the output signal only remains η(3J₀²(m) + 6J₁²(m)). In order to simplify, we use I_D to denote 3J₀²(m) + 6J₁²(m), and assume that the time delay τ is adjusted to equal Δ for synchronize, then the expression of i_{LPF_upper}(t) can be revised as:

As can be seen in Eq. (17), the output signal at upper LPF has an amplitude difference step of 4J₀(m)J₁(m) for these three cases, which can be processed by DSP to recover the data carried by the microwave FSK signal. Therefore, the coherent demodulation can be complete by only use the upper path. In fact, the lower path will give the same demodulation result. It is worth mentioned that if the time delay τ and Δ are out of synchronize, namely, τ≠Δ and cos(4ω(Δ-τ)) < 1, the amplitude difference step will be reduced, causing trouble to DSP in decision and increasing the bit error rate.

When the communication signal chooses the microwave FSK signal with flexibly tunable subcarriers, just as the signal in Eq. (11), we can also get the output of LPF in line with the same derivation method, and the results is shown in Eq. (18). Here, the I_D′ denotes 2J₀²(m) + 4J₁²(m). As can be seen, the upper LPF and the lower LPF gives two complementary waveforms, which can be decided by the DSP to recover the data.

3. Simulation and discussion

To investigate its mechanism, A simulation system based on Fig. 2(b) is performed with the aid of OptiSystem 15.0. Table. 2 lists the basic parameters settings adopted in the system according to actual devices. A BPF with an ideal rectangular response and a passband of 1-40 GHz is applied to simulate the bandpass characteristics of the antenna. The cut-off frequencies of the LPFs used in demodulation is set to 3 GHz.

Table 2. the parameters settings in simulation

View Table | View all tables in this article

First, we demonstrate the generation of microwave FSK signal with the frequency multiplication factors of four and eight. The microwave driving signal is set to 2 GHz with an amplitude of 2.93 V to satisfy |J₀(m)|=|J₄(m)|. The rotation angle of PC₁ is set to 0°. In view of the almost identical output spectrum of the two QPSK modulators, only one of them is shown in Fig. 3(a). As can be observed, the optical carrier and ±4th-order sidebands are retained with equal amplitude. Besides, the optical spurious suppression ratio (OSSR) is 23.93 dB so that other unwanted sidebands with much lower power can be ignored. Figure 3(b) shows the optical spectrum at the output of the Pol when the PM is driven by a DC voltage of 3 V (red plot) or 1 V (blue plot). As can be seen, the spectrum in red plot contains the ±4th-order sidebands with an OSSR of 23.22 dB. Conversely, the spectrum in blue plot contains the optical carrier and the +4th-order sideband with an OSSR of 22.57 dB. The results are consistent with the theoretical analysis in Eq. (6).

 

Fig. 3. (a) The optical spectrum at the output of the QPSK modulator. (b) The optical spectrum at the output of the Pol when the DC driven voltage of PM is 3 V (red plot) or 0 V (blue plot). (c) the electrical spectrum and (d) the waveform of the generated 1Gbit/s 8/16 GHz microwave FSK signal.

Download Full Size | PDF

Subsequently, the DC driven of the PM is changed to a 1Gbit/s binary coding sequence with a pattern of ‘‘01100101’’, and the amplitudes of bit ‘0’ and bit ‘1’ are respectively set to 1 V and 3 V. Then a microwave FSK signal can be obtained at the output of PD, whose electrical spectrum and waveform are shown in Fig. 3(c) and (d). As can be seen, there are two subcarrier frequencies at 8 GHz and 16 GHz, indicating that the frequency multiplication of 4/8 is successfully implemented. In addition, the RF spurious suppression ratio (RFSSR) is 21 dB, which means the generated signal features well electrical spectrum purity. From the waveform diagram, we can easily observe the frequency jumps, and the zoomed-in views from 1.8 to 2.3 ns and 8.2 to 8.7 ns show that a single cycle has a time of 1/16 ns and 1/8 ns, corresponding to the frequency of 16 GHz and 8 GHz.

Data recovery can be achieved by the photonic-optimized coherent demodulation scheme proposed in Section 2.4. As illustrated in Fig. 4, the curve after the upper LPF (blue line) and the recovered data (red line) after DSP is consistent with the coding pattern of ‘‘01100101’’.

 

Fig. 4. The curve after the upper LPF (blue line) and the recovered data (red line) after DSP.

Download Full Size | PDF

On the other hand, we demonstrate the generation of microwave FSK signal with flexibly tunable subcarriers. At this time, the two QPSK modulators are driven by two 1.5 V microwave signals at 4 GHz and 3 GHz respectively, and the rotation angle of PC₁ is adjusted to 45°. Figure 5(a) shows the output spectrum of the upper QPSK modulator (red line) and the lower QPSK modulator (blue line). Both of them are contain ±2nd-order sidebands with an OSSR of more than 33 dB. Figure 5(b) shows the optical spectrum at the output of the Pol when the PM is driven by a DC voltage of 3 V (red plot) or 0 V (blue plot). As can be seen, only the output signal of the upper QPSK modulator is retained in the red plot, on the contrary, there are only the signal output from the lower QPSK modulator is retained in the blue plot. The results are consistent with the theoretical analysis in Eq. (9) and Eq. (10). Note that, compared with Fig. 5(a), the optical spectrums in Fig. 5(b) features little change, which indicating that the two signals are ideally polarization separated. However, due to the impact of polarization state drift existed in practice, part of their energy will leak to the other side, resulting in the quality deterioration of the generated microwave FSK signal. The corresponding discussion will be given later.

 

Fig. 5. (a) The optical spectrum at the output of the upper QPSK modulator (red line) and the lower QPSK modulator (blue line). (b) The optical spectrum at the output of the Pol when the DC driven voltage of PM is 3 V (red plot) or 1 V (blue plot). (c) the electrical spectrum and (d) the waveform of the generated 1Gbit/s 12/16 GHz microwave FSK signal.

Download Full Size | PDF

Similarly, the DC driven of the PM is changed to the binary coding sequence as described above, except that its amplitude of bit ‘0’ is adjusted to 0 V. Then a microwave FSK signal can be obtained at the output of PD, whose electrical spectrum and waveform are respectively shown in Fig. 5(a) and (b). As can be seen, the electrical spectrum features high purity and the two subcarrier frequencies at 12 GHz and 16 GHz implies that the frequency multiplication of 4 is successfully implemented. Obvious frequency jumps can be observed from the waveform, and the zoomed-in views from 1.8 to 2.3 ns and 8.2 to 8.7 ns show that a single cycle has a time of 1/16 ns and 1/12 ns, corresponding to the frequency of 16 GHz and 12 GHz. Figure 6 shows the results of the photonic-optimized coherent demodulation, the curve after the upper LPF (green line) and the lower LPF (blue line) present two complementary waveforms, and after DSP processing, the recovered data (red line) is consistent with the coding pattern of ‘‘01100101’’.

 

Fig. 6. The curve after the upper LPF (green line) and the lower LPF (blue line), the red line exhibits the recovered data after DSP.

Download Full Size | PDF

Benefit from its rich RF and DC ports, the DP-QPSKM can be skillfully adjusted to generate several kinds of high order optical sidebands such as the 0, ±4th order sidebands or ±2nd order sidebands in the scheme, providing the basis for the generation of multiform microwave FSK signal with high multiplication factor. As a cost, the DC bias drift is an important non-ideal factor affecting the signal quality. Although well-researched and commercialized feed-back bias control technologies can be applied to mitigate the impact [19,20], it is still necessary to investigate the influence of the slight DC bias drift since there are many DC biases need to be controlled in the scheme.

According to the Eq. (2,3) and the relationship θ_i=πV_i/V_π (i = 1, 2), where V_π is the half-wave voltage of the MZM and equals to 4 V in the simulation work, we can get the corresponding ideal DC bias, namely, V₁= 0 V and V₂= 0 V when generating 0, ±4th order sidebands, V₁= 0 V and V₂= 2 V when generating ±2nd order sidebands. Here, we discuss the impact of DC bias drift within 0.1 V on the OSSR of the output spectrum of QPSK modulator, and the results of the above two situations are respectively shown Fig. 7(a) and (b). As can be seen from Fig. 7(a), the OSSR decreases as the DC bias deviate from ideal value, and when V₁ and V₂ are deviate to ±0.1 V, the OSSR decreases from the maximum value of 23.93 dB to minimum value 13.05 dB. Besides, the contour map shows that the V₁ drift and V₂ drift have almost the same impact on the OSSR and the larger the bias drift, the faster the OSSR decreases. Compared with Fig. 7(a), Fig. 7(b) shows a different trend. When V₁ and V₂ are deviate to ±0.1 V, the OSSR decreases from the maximum value of 33.6 dB to minimum value 21 dB. The contour map shows the impact of V₂ drift is greater than that of V₁ drift, and the impact of V₁ drift is masked if the V₂ drift is higher than 0.05 V. Fortunately to both the two situations, if the bias drift can be controlled below 0.05 V, the OSSR is still higher than 20 dB, thereby the optical signal quality can be guaranteed. This requirement can be achieved by the existing bias control technologies. In fact, at the expense of halving multiplication factor, the DP-QPSKM in this scheme can be replaced by DP-BPSKM, which will significantly reduce the influence of DC bias drift.

 

Fig. 7. The impact of DC bias drift within 0.1 V on the OSSR of the output spectrum of QPSK modulator. (a) when generating optical carrier and ±4th order sidebands. (b) when generating ±2nd order sidebands.

Download Full Size | PDF

Due to limited device’s accuracy and environment perturbation, the non-ideal phase shift of the PS and non-ideal polarization angle of the PC₁ would be another two main factors lead to the deterioration of the signal quality. Firstly, we discuss the impact of phase shift drift within 3° on the OSSR of the output spectrum of QPSK modulator in two situations, namely, generating 0, ±4th order sidebands and generating ±2nd order sidebands, and the results are respectively shown as the red line and blue line in Fig. 8(a). As can be seen, when the QPSK modulator is adjusted to generate 0, ±4th order sidebands, the OSSR drops greatly from the maximum value of 23.93 dB to minimum value 10.2 dB within 3° drift, the reason is that the power of unwanted ±2nd order sidebands increase quickly as the phase shift deviate from the ideal value. While the phase shift drift has little effect on the OSSR when the QPSK modulator generate ±2nd order sidebands. The results show that the realization of high frequency multiplication factor comes at the cost of strict requirements on the system stability and device performance. Specially, if the phase deviation is limited within 1°, the OSSR can still higher than 19 dB to guarantee the signal quality, and this can be easily achieved by current commercially available phase shifters. Then, the impact of polarization angle drift within 3° on the OSSR of the output spectrum of Pol is investigated. Figure 8(b) shows the results of the two situations, in which the solid line depicts the OSSR when the binary coding sequence is all bit ‘1’ and the dotted line depicts the OSSR when the binary coding sequence is all bit ‘0’. It can be seen that the non-ideal polarization angle has little effect on the first situation while has obvious influence on the second situation, yet the OSSR can still remain above 22 dB. Actually, the polarization of optical signal is sensitive to the environment perturbation, which will cause trouble to the polarization alignment. As for this problem, polarization maintaining fiber is usually employed to stabilize the polarization state of the lightwave, and a polarization tracker can also be utilized to further improve the system stability.

 

Fig. 8. (a) The impact of phase shift drift within 3° on the OSSR of the output spectrum of QPSK modulator. (b) The impact of polarization angle drift within 3° on the OSSR of the output spectrum of Pol. Solid line: the OSSR for all bit ‘1’, Dotted line: the OSSR for all bit ‘0’.

Download Full Size | PDF

Given that the non-ideal factors may deviate simultaneously in the spectrally cluttered environment, we set all the DC bias drift to 0.05 V, all the phase shift drift and polarization angle drift to 1°, to investigate the joint influence on the quality of the generated signal. Figure 9 shows the electrical spectrum, the waveform and the recovered data of the generated 1Gbit/s 8/16 GHz microwave FSK signal. It can be seen that there is a decrease from 21 dB to 15.2 dB on the RFSSR due to the increased power of the spurious components, and the deteriorated waveform can also be observed. However, the frequency jump is still clear and the recovered binary data is also consistent with the binary coding sequence. Similarly, Fig. 10 shows the same contents of generated 1Gbit/s 12/16 GHz microwave FSK signal, and the binary data can still be correctly demodulated despite the deterioration of signal quality.

 

Fig. 9. (a) the electrical spectrum, (b) the waveform and (c) the recovered data of the generated 1Gbit/s 8/16 GHz microwave FSK signal under the joint influence of non-ideal factors. The red line in (b) is the curve after the upper LPF.

Download Full Size | PDF

 

Fig. 10. (a) the electrical spectrum, (b) the waveform and (c) the recovered data of the generated 1Gbit/s 12/16 GHz microwave FSK signal under the joint influence of non-ideal factors. The red line in (b) is the curve after the upper LPF.

Download Full Size | PDF

Last but not least, compared to the schemes based on a single modulator [8,9,14,15], the serial structure in this scheme will increase the insertion loss but ensures more effective extension to realize multi-paths microwave FSK signal generation, as is often required in the transmitter of MIMO-FSK system [21]. The topology of the multi-paths microwave FSK signal generator is exhibited in Fig. 11.

 

Fig. 11. Topology of the multi-paths microwave FSK signal generator.

Download Full Size | PDF

4. Experimental verification

To verify its feasibility, a proof-of-concept experiment is also carried out as shown in Fig. 1(a). A laser source (Oppu Times Inc.) is used to provide a 1550.68 nm optical carrier at a power of 16 dBm, which then injected into a DP-QPSKM (Fujitsu, FTM7977HQA) with a half-wave voltage of 3.5 V. A microwave signal generator (Rohde & Schwarz SMW 200A) offer the sine waveform to drive DP-QPSKM. and a 2G Sa/s arbitrary waveform generator (AWG, Tektonix, AFG3252C) offer the coding sequence to drive the PM (EOSPACE). The modulated optical signal is detected by a PD (Conquer, KG-PD-50G) with a responsivity of 0.6 A/W. An optical spectrum analyzer (OSA, THORLABS 202C) and an oscilloscope (Tektonix, MSO64B) are employed to monitor the optical spectrum and waveforms, respectively. In addition, a BPF with a Chebyshev type II response and a passband of 1-5 GHz was created through an offline program to simulate the bandpass characteristics of the antenna.

By carefully adjusting the amplitudes of microwave driven signal, the PC₁ and the Pol, the generation of optical signal with 0 and ±4th-order sidebands are verified as shown in Fig. 12(a-c). Here, the frequency of the microwave driving signal is set to 5 GHz, the amplitude of binary coding sequence is set to 0 V for bit ‘0’ and 1.8 V for bit ‘1’. Similarly, Fig. 12(d-f) shows the spectrum of the generated optical signal with ±2th-order sidebands when two microwave driving signals are applied with frequencies of 9 GHz and 12 GHz, and the amplitude of binary coding sequence is set to 0 V for bit ‘0’ and 3.5 V for bit ‘1’. Compared with the simulation results in Fig. 3(a-b) and Fig. 5(a-b), the OSSR of the optical spectrum has dropped a lot, which is caused by the lower extinction ratio of the DP-QPSKM (<30 dB) and the combined influence of non-ideal factors. Due to the limited bandwidth (4 GHz) of the oscilloscope, the frequency of the microwave driving signal was limited below 1 GHz in the next experiment, then the corresponding optical spectrum cannot be observed since the wavelength resolution of the OSA is ∼7.5 GHz.

 

Fig. 12. Spectrum after the PBS and Pol of the optical signal with 0 and ±4th-order sidebands.

Download Full Size | PDF

As for the generation of microwave FSK signal whose frequency multiplication factors are respectively four and eight, we set the frequency of the microwave driving signal to 500 MHz, and the coding sequence is set to 100Mbit/s with a pattern of ‘‘01100101’’. Since the oscilloscope (Tektonix, MSO64B) comes with an FFT digital processing program to calculate the electrical spectrum, both the electrical spectrum and waveform of the generated 100Mbit/s 2/4 GHz microwave FSK signal are recorded by the oscilloscope, as shown in Fig. 13(a) and (b) respectively. Compared with Fig. 3(c), the RFSSR in Fig. 13(a) deteriorated from 21.0 dB to 10.1 dB. Two main reasons are as follows: 1) The generation and application of high-order optical sidebands (0 and ±4th order sidebands, or ±2th order sidebands) are easily disturbed by non-ideal factors such as low extinction ratio of the DP-QPSKM, phase shift drift, and polarization angle drift, resulting in the incomplete suppression of the unwanted optical harmonic components. 2) Since the scheme is constructed by discrete devices, the system stability is easily affected by environmental disturbances, for example, temperature variation, mechanical vibration and air fluctuation. Fortunately, the different subcarrier frequencies that vary according to the coding sequence can still be observed from the waveform in Fig. 13(b), and the recovered data in Fig. 3(c) agrees with the binary coding sequence. It is worth mentioned that since the lack of some key devices, we cannot complete the coherent demodulation in experiment, but instead import the waveform data into OptiSystem to obtain the recovered data by simulation, where the cut-off frequencies of LPFs are set to 300 MHz. As for the generation of microwave FSK signal with flexibly tunable subcarriers, two microwave driving signals with frequencies of 550/950, 600/900, 650/850, 700/800 MHz are applied. The generated waveforms are calculated by Hilbert transform to obtain their instantaneous frequencies, which are shown in Fig. 14. The results also agree with the binary coding sequence.

 

Fig. 13. (a) the electrical spectrum, (b) the waveform and (c) the recovered data of the generated 100Mbit/s 2/4 GHz microwave FSK signal. The red line in (b) is the curve after the upper LPF.

Download Full Size | PDF

 

Fig. 14. the calculated instantaneous frequencies of the generated 100Mbit/s 2.2/3.8, 2.4/3.6, 2.6/3.4, 2.8/3.2 GHz microwave FSK signal.

Download Full Size | PDF

The experiment results in Fig. 14 can also verify the frequency tunability of the scheme, although the frequency tuning range is controlled within 4 GHz due to the bandwidth limitation of the oscilloscope. Fortunately, by means of mathematical analysis, we can obtain the theoretical frequency tuning range of this scheme [22]. Assuming that the operation bandwidths of the devices in the scheme are [f_min-PS, f_max-PS] for the phase shifters, [f_min-s, f_max-s] for the power splitters, and [f_min-D, f_max-D] for the DP-QPSKM. f_PD represents the 3 dB bandwidth of the PD. Then the minimum and maximum operation frequency of the system is:

5. Conclusion

In conclusion, a photonic-assisted generation scheme for multiform microwave FSK signal is proposed based on a DP-QPSKM and a Sagnac loop structure. By adjusting the inputs of the modulators, microwave FSK signal with fixed double relationship (at frequency multiplication factors of 4/8) or flexibly tunable (at a frequency multiplication factor of 4) subcarrier frequencies can be obtained. Besides, a photonic-optimized transceiver scheme is also proposed to achieve the demodulation of high-frequency and wide-bandwidth microwave FSK signal without electronic bottleneck. Theoretical analysis and simulation work have been performed to investigate the mechanism. Discussions about the non-ideal parameters are given in Section III. It is found that the signal quality could be guaranteed when the DC bias drift is within 0.05 V, as well as the phase shift drift and polarization angle drift are within 1°.

A proof-of-concept experiment has been carried out to verify the system feasibility although the limited conditions. Microwave FSK signals with subcarrier frequencies of 2/4, 2.2/3.8, 2.4/3.6, 2.6/3.4, 2.8/3.2 GHz are successfully generated and the frequency hopping laws are in line with the binary coding sequence. The signal quality in experiment is obviously inferior to the one in simulation, which is mainly due to the superposition of various non-ideal factors and the system instability caused by the discrete structure. Potential photonic integration could enhance the system performance and further reduce the complexity, making it attractive for the future practical implementation.