1. Introduction

Optoelectronic oscillator (OEO) has attracted lots of attention due to its superior advantage in the generation of the high-frequency micro/millimeter-wave signals with low phase noises since it was proposed by Yao and Maleki in 1996 [1–3]. During last two decades, it has been widely used in many application scenarios such as radar and sonar, electronic warfare, wireless communication and so on [4]. Traditionally, one standard OEO configuration is equipped with one laser source, a high-speed electro-optic modulator (EOM), a long optical fiber, one photodetector, as well as an opto-electrical feedback loop, in which a wideband electrical amplifier (EA) with high gain and an electrical bandpass filter (EBPF) with narrow 3-dB bandwidth are always indispensable as the Fig. 1(a) shows [5]. However, once the EBPF is fixed, the frequency of the generated microwave signal is simultaneously fixed. Thus, the EBPF with a tunable central frequency is urgently demanding. But for high frequencies, it is very hard to find such an EBPF equipped with sufficiently narrow 3-dB bandwidth, huge central frequency, as well as wide tuning range at the same time [5]. To overcome the tunability limitation, in recent years, several OEO structures based on tunable microwave photonic filters (MPFs) to replace the function of the traditional EBPF have been reported as shown in Figs. 1(b)–1(c) [6–10]. Figure 1(b) mainly utilizes the wavelength-selective optical gain of a single-mode DFB laser under optical injection to realize a wideband tunable MPF while the schematic diagram in Fig. 1(c) is the further study of Fig. 1(b), where one monolithically integrated dual-mode semiconductor laser is utilized to replace the discrete optical injection systems. The integrated laser itself functions as an active tunable MPF. Based on this method, stable microwave signals continuously tunable from 17.3 GHz to 21.7 GHz were successfully allocated in [7] just by tuning the injection currents of the laser sections to change the central frequency of the active tunable MPF. Nevertheless, it can be easily found that in this scheme, the built-in MPF realized can be tuned from 16.9 GHz to 34.7 GHz, corresponding to about 20-GHz tuning bandwidth. Obviously, even if the tuning characteristic of the MPF is rather wide, the tuning range of the generated microwave signals remains limited within several GHz in some extent. This phenomenon would attribute to the narrow bandwidth of the EAs. Actually, subject to the bottleneck problem universally lying in conventional electrical instruments, EAs are enslaved to the high cost and complicated fabrication process when it comes to the high-frequency region, leading to a finite amplification region and thus the tuning frequency of the generated microwave signal is often confined to a dozen GHz. Contrarily, the common frequencies in the aforementioned application scenarios basically cover the full band of the microwave and millimeter-wave signals with the frequencies from 300 MHz to 300 GHz [2]. Thus, the usage of the EAs would be against the original intention of the OEO to generate the high-frequency and low-phase-noise microwave signals and heavily limit the development of the OEO. Moreover, in principle, only when the gain is greater than the total losses, the oscillation can be sustainably self-starting, and as a result, multistage EAs are usually required to provide enough gain (up to 60 dB) to compensate for the large insert loss and the electro-optical conversion loss derived from the external modulation, leading to complex systems and a large power consumption [10]. Last but not least, once multi-stage microwave amplifiers are used in the feedback links, the phase noises of the generated microwave signals can be heavily constrained by the excess thermal noise of the EAs, which is one of the main noise sources in the OEO cavity [11].

 

Fig. 1. The schematic of (a) traditionally standard OEO, (b) OEO using optical injection semiconductor lasers, (c) OEO using the integration optical source device, and (d) a novel OEO configuration using all-optical gain.

Download Full Size | PDF

Recently, a novel kind of scheme, employing the all-photonic gain to replace the microwave amplifiers to compensate the losses was proposed [12–14]. In [13], an all-optical gain OEO operating at 10 GHz was investigated by utilizing an erbium-doped fiber amplifier (EDFA) to provide the necessary RF gain for oscillation. Compared to the EA configuration, the improvement in the phase noise at least 10 dB was verified due to the lower RF noise figure in the all-optical gain process. However, the tuning range of the generated microwave signals is restricted due to the utilization of the EBPF in this system. While in [14], the function of the EBPF was substituted with one equivalent MPF by making the most of the optical sideband injection locking effect. Hence, the tuning range can be enhanced largely. In this system, the necessary gain was mainly provided by the period-one (P1) state of the optical injection mechanism. It should be explained that the beating of the two main oscillating optical modes under P1 state will greatly promote the initial oscillation of the OEO while in the traditional OEO mechanism just like in [13], the oscillation is built up from the lower transient noise in the loop constantly. Thus, in this system, no additional EA was required owing to the undamped relaxation oscillations in the P1 state. Yet, due to the usage of the phase modulator, the maximum frequency of the generated microwave signals is also limited by the modulation bandwidth of the external modulator. Moreover, the complexity and the cost of the system are also increased heavily because of the applications of the external modulator as well as the discrete optical injection systems.

In this paper, one novel photonic method based on the monolithically integrated dual-frequency semiconductor laser (MI-DFSL) with all-optical gain to realize the OEO is proposed and experimentally demonstrated. On one hand, the tunable MPF with narrow 3-dB bandwidth as well as the high-efficiency modulation in high frequency can be simultaneously allocated by making full use of the photon-photon resonance (PPR) effect induced modulation response enhancement. On the other hand, the conventionally narrow-band EAs are replaced with one wide-band EDFA to provide the necessary gain for achieving oscillation. By combining both advantages of the above two aspects, the generated microwave signals would largely break the limitations in bandwidth in traditional OEO configuration as shown in Figs. 1(a)-(c) because of no utilization of any electrical instrument such as EOMs, EAs and EBPFs. To the best of our knowledge, it is the first time to utilize the monolithic integrated device with all-optical gain method to generate the microwave signals as shown in Fig. 1(d). In principle, the maximum frequency of the generated microwave signal can be infinitely close to the upper bound of the P1 frequency up to several hundred GHz. In the proof-of-concept experiment, the microwave signals with wide tuning range from 14.2 GHz to 25.2 GHz are realized. The single sideband (SSB) phase noises in all tuning range are below -103.77 dBc/Hz at10 kHz and the best signal of the -106.363 dBc/Hz at10 kHz is achieved at the frequency of 17.2 GHz, indicating the effectivity of the all-optical gain OEO. In addition, the successful application of the MI-DFSL, which highly inherits the numerous merits of the discrete optical injection systems, provides a simple and cost-effective method by taking full advantage of the photonic integration.

2. Principle

Figure 2(a) illustrates the experimental schematic of the proposed all-optical gain OEO. As the core device in the system, the MI-DFSL can be seen as an extraordinary free-emitting laser structure, in which two Bragg gratings share the same active substrate as Fig. 2(b) shows. Thus, the MI-DFSL is born with natural dual-wavelength mechanism in one single laser package due to the integration of the two DFB laser onto one chip. When the appropriate injection currents are separately applied to control the two DFB lasers to oscillate themselves, the two DFB lasers would be mutually injected simultaneously and drastically interact with each other due to the shortage of the optical isolator/circulator. As a result of the short distance of the phase section between the two DFB laser sections, the MI-DFSL would go into the so-called short-coupling regime and in this case, abundant nonlinear phenomena can be observed from two irrelevant separate wavelengths, mutual injection locking, P1 state, period-two (P2) oscillation, quasi-periodic (QP) pulsation, chaos (C) and so on. Additionally, more prominent improvements in the whole electro-optical conversion performances than that of the single-section semiconductor lasers can be also allocated due to the enhanced light–matter interactions in small space such as modulation response enhancement, chirp suppression, nonlinear reduction and so on.

 

Fig. 2. (a) The experimental setup of the all-optical gain OEO with the help of the laser module, (b) the schematic diagram of the laser structure.

Download Full Size | PDF

In this paper, the operation principle is mainly based on the modulation response enhancement in the MI-DFSL by making full use of the PPR effect. When the MI-DFSL is tuned to work under dual-mode state (sidemode suppression ratio, SMSR > 30 dB), the direct modulation response can have a striking improvement and two salient features would be captured. At first, the relaxation resonance frequency (RRF) can be enhanced. In this case of the MI-DFSL, the RRF is defined as the frequency difference between the two main oscillating optical lights. Thus, with the variation of the frequency interval between the two main oscillating modes in the MI-DFSL when different injection parameters are given, the RRF can be correspondingly changed from several GHz to several hundred GHz. In theory, the RRF in the modulation response curve can be up to the maximum of the P1 frequency, which will largely eliminate the urge requirement of the high-speed external modulator. The limitation in the maximum of the CPR induced RRF in traditional directly modulated semiconductor lasers would be also broken. Second, the modulation response around the RRF would be enhanced. Compared to other frequencies, an obviously sharp response peak with narrow 3-dB bandwidth at the frequency near RRF can be intuitively got, leading to a high modulation efficiency. Moreover, the modulation response peak would be also changed with the RRF when the currents applied onto the two DFB lasers are changed. Thus, one equivalently tunable MPF with a relatively narrow 3-dB bandwidth of MHz level can be obtained, whose amplitude response shape strictly follows the modulation response curve and the homologous central frequency is dependent on the RRF. Because the RRF can be tuned in an ultra-widely range, the central frequency of the MPF can be also arbitrarily changed in a wide range. On the basis of these two premises, once the opto-electrical feedback with enough gain is provided, different OEOs with various oscillating frequencies would be generated.

Further, once the opto-electrical feedback loop is closed, the sideband amplification injection locking (SAIL) effect would make effort to ulteriorly amend the equivalent MPF in advancing the long stability of the central frequency as well as compressing the 3-dB bandwidth as Fig. 3 shows. The procedure can be divided into two steps. Firstly, the beating frequency is fed back to modulate the front laser (FL) section and the +1^st optical modulation sideband would get amplified due to the ultra-narrow gain spectrum around the cavity mode of the rear laser (RL) section. Secondly, the enhanced modulation sideband would lock the cavity mode of the RL section because the optical modulated signal exactly right falls into the range of the injection locking and the pulling bandwidth. Thus, once the signal cycles many times in the opto-electrical loop, the equivalent MPF with a narrower 3-dB bandwidth and a more stable central frequency would be realized.

 

Fig. 3. The principle diagram of the equivalent microwave photonic filter based on the combined impact of the PPR effect and the SAIL effect: (a) the red-shifted cavity modes in the twin DFB laser sections with a gain spectrum, (b) the optical spectrum in the FL section under modulation, (c) the +1^st modulation sideband is amplified by the gain spectrum and locks the cavity mode of the RL section, the other spurious mode is filtered equivalently, (d) the frequency relative response of the FL section under injection and without mutual injection as well as the RRF point based on CPR and PPR effect.

Download Full Size | PDF

Gain is also an essential part and worthy of being considered in the discussion of OEOs. Firstly, it needs to be emphasized that different from traditional OEOs, in which the beating frequencies are derived from the noise, in our systems, the two main oscillating modes under undamped relaxation oscillation owing relatively high power can greatly reduce the self-starting requirement. Besides, the SAIL effect can also provide partial gain applied for the link oscillation, which would alleviate the requirement of the extra gain. The remaining gain to oscillate is mainly completed in optical domain through one small-signal EDFA. Compared with EAs only amplified in several decade GHz, the EDFA can realize an ultra-broadband amplification about 30 nm with a great flatness and low noise figure. Therefore, on the basis of these aforementioned factors, the bandwidth limitation in the frequency of the generated microwave signal would be broken and the advantage of the OEOs to generate high-frequency signals with low phase noises can be emerged.

Based on the natural dual-wavelength mechanism, the optical signal can be expressed as follow:

After modulation,

While the ${\omega _1} - {\omega _2}\textrm{ = }{\omega _m}$,

So,

From the formula, there still exists the constant photocurrent even if ${I_m} = m\sin ({\omega _m}t) \to 0$,

The amplitude at the frequency component of ${\omega _m}$ is $2\rho G{E_1}{E_2}$.

Compared to the OEO based on the traditional external EOM,

3. Experiment

3.1 Experimental setup

A proof-of-concept experiment is performed following the structure in the Fig. 2(a). The MI-DFSL chip is packaged in butterfly housing as a module for ease of use and the phase section is set unconnected to simplify the system. The control currents of the twin DFB laser sections are provided separately by two laser diode controllers (Thorlabs LDC205C) and the working temperature is guaranteed at 25°C through a temperature controller (Thorlabs 200C) to ensure a stable working environment during the entire process. The output light from the laser module is firstly divided by an optical splitter. 10% is extracted to the optical spectrum analyzer (Finisar Wave-Analyzer 1500s) with a resolution of 1.2 pm to monitor the working state of the MI-DFSL. The rest 90% is transmitted to the optical link. To suppress the sidemode and achieve a high SMSR, i.e., a stable single mode oscillation, the vernier caliper effect is also adopted in this system. One polarization controller (PC) is firstly applied to match the polarization of the signal light with the polarization beam splitter (PBS). Subsequently, the two orthogonal polarized lights separated by PBS travel two SSMFs with different fiber lengths, one is 2.9 km and the other is 10.1 km. Finally, the two polarized lights combined at the position of the polarization beam combiner (PBC) are amplified by one small-signal EDFA with a 4.5-dB noise figure to provide the suitable gain about 26 dB. Due to the limitation of the maximum input power of the photodetector (PD: U2T XPDV2120R), one optical attenuator is optional to reduce the output optical power from the EDFA. An extra optical splitter is also given to separate the optical light, and 50% part is transmitted to the electrical spectrum analyzer (ROHDE&SCHWARZ FSWSIGNAL& ANALYZER) to monitor the electrical spectrum, while the other 50% is electrically fed back to the laser module to form the OEO loop by modulating the feedback signal on the FL. Obviously, it is no difficult to find that in this system, none of the additional high-speed external EOM, high-central-frequency EBPF or multistage EAs are needed in this system, leading to a low-cost configuration with a high level of agility.

3.2 Experimental results

In the experiment, the appropriate control currents are applied to make sure the MI-DFSL to operate in the dual-mode state and when the control currents are changed within a certain range, the MI-DFSL would operate with diverse frequency intervals. The concrete principle is similar to that in [15] and the static characteristics of this MI-DFSL in operation are measured when the opto-electrical feedback loop is open. As Fig. 4 shows, the optical spectrum as well as the corresponding electrical spectrum with I_DC1 varied from 54 mA to 88 mA and I_DC2 fixed at 81.47 mA are demonstrated. Both of the cavity modes in the RL and FL sections are red-shifted with the increasing current applied on the RL section. They move different wavelength distances, leading to the different beating frequencies after optical-to-electrical conversion. Due to the carrier injection induced refractive index change, the oscillation wavelength of the RL section moves toward the longer wavelength side obviously while the wavelength of FL section changes slightly owing to the indirect anti-guidance effect. As a result, the frequency difference between the two DFB laser sections gets gradually larger. However, though the generated microwave signals have certain tuning property, the signal qualities in the linewidth and the phase noises are supposed to be considered and optimized further. In addition, one thing to be clarified here is that limited to the bandwidth of the measuring instrument, the maximum of the generated microwave signals is only 25.322 GHz.

 

Fig. 4. (a) Measured optical spectrum with I_DC1 varied from 54 to 88 mA while I_DC1 being fixed at 81.47 mA, (b) measured electrical spectrum with I_DC1 varied from 54 to 88 mA while I_DC1 being fixed at 81.47 mA.

Download Full Size | PDF

The enhanced modulation responses due to the PPR effect in different control currents without the EDFA are also evaluated. As the Fig. 5(a) shows, when giving the two DFB laser sections different combinations of the control currents, the interaction between the photons to photons would drastically change and as a result, the modulation response characteristics can be enhanced. The black curve is the corresponding modulation response when no current is applied to the RL section while the red curve represents the enhanced modulation response when RL is modulated by direct current. It can be seen one evident RRF peak with a narrow 3-dB bandwidth is recorded when comparing the two curves. The RRF induced by PPR effect is about 22.8 GHz, while the RRF induced by CPR effect is only 10 GHz, indicating a twice enhancement is realized. In addition, the modulation response value of the RRF peak is -2.5 dB while in many previous architectures, the modulation response peaks are just around the -20 dB, indicating a significantly high modulation efficiency. Thus, the demand of the amplitude of the feedback signal would be reduced highly. In other words, the high modulation efficiency also can provide partial gain for oscillation. Further, by constantly changing the control currents, different RRF points with sharp modulation response would be recorded as Fig. 5(b) shows. From the curves in Fig. 5(b), the peaks of the frequency responses can experience from 13.8 GHz to 35.06 GHz. Moreover, basically all peaks in different curves are close to 0 dB and even some are more than 0 dB. This phenomenon indicates the generality of the high modulation efficiency in a wide frequency range, guaranteeing the feasibility of the proposed scheme. Once adding the EDFA to this loop to provide the rest gain, the all-optical gain OEO would be generated. However, the responses in different peaks are uneven and it mainly imputes to the mode competition within the two wavelengths [16] and when concerning to the high frequency, the limited transmission bandwidth of the microwave cable is also a main influence factor. For each curve, the steep peak around the RRF as well as the rather small response in other frequencies can lead to the MI-DFSL a new functionality with a built-in MPF, in which the frequency around RRF would go through and the superfluous signals outside the region would be filtered. The 3-dB bandwidth of the equivalent MPF is about 30 MHz due to the wide sideband injection locking range. However, it is convinced that the wide sideband injection locking range would not make effect in the signal quality of the generated microwave signal in the condition of the closed opto-electrical feedback loop.

 

Fig. 5. (a) The modulation response comparison diagram with/without mutual injection, (b) the enhanced modulation response curves under different combinations of control currents.

Download Full Size | PDF

To better understand the effect of the opto-electrical feedback loop, the output signals separately in the case of the closed/opened feedback loop are compared. As Fig. 6(a) shows, the optical spectrums are nearly similar without any variation. However, in Fig. 6(b), the beating signals are changed evidently. The curve in the blue is the generated microwave signal in the case of the closed loop while the red curve represents the generated microwave signal when the feedback loop is not closed. It can be seen when the loop is not closed, the linewidth is comparatively wide and the central frequency is rather unstable. After closing the feedback loop, the linewidth can be reduced from MHz level to kHz level, indicating more than three order of magnitude reduction compared with the opto-electrical feedback OFF state. In addition, by using dual-loop effect, the SMSR of the obtained signal is enhanced to be 50 dB. Moreover, the performance of the long-term frequency stability of the output signal is investigated by continuously recording the frequency drifts. As Fig. 6(c) shows, within 5 minutes, the center frequency of the microwave signal only drifts 9 kHz without any hopping mode. The SSB phase noise is also evaluated to verify the signal quality of the generated microwave signal. As Fig. 6(d) shows, the SSB phase noise of the generated signal at the frequency of 17.2 GHz is -106.363 dBc/Hz at10 kHz.

 

Fig. 6. The comparison diagram of the optical spectrum with/without the closed loop, (b) the comparison diagram of the RF spectrum with/without the closed loop, (c) the measured electrical spectrum of the generated 17.2 GHz beat signal, the inset is the corresponding spectrum within 100-kHz span, (d) the phase noise of the 17.2 GHz signal.

Download Full Size | PDF

The tunability has been convinced above and in the following concept, for making a better understanding of the improvement of the all-optical gain OEO, the assessed microwave signals are given at a fixed frequency interval by setting different combinations of control currents to the integrated laser module. The SSB phase noise characteristic is considered to further investigate the improvement in the signal quality of the generated microwave signal with opto-electrical feedback. As Fig. 7 shows, the SSB phase noises of the generated microwave signals at a 10-kHz frequency offset from the carrier in the closed loop are all below -103.77 dBc/Hz, indicating an evident performance optimization. The values of the phase noises at the frequency offset of 1kHz are also measured and all of them are lower than -83.61 dBc/Hz in the whole observation points.

 

Fig. 7. (a) The measured phase noises of the various OEOs at different frequencies, (b) the phase noise values at 1-kHz and 10-kHz frequency shift under different frequencies.

Download Full Size | PDF

4. Conclusion

What we need to pay attention to is that, the tuning range of the generated microwave signals can be further improved. It has been reported that the minimum frequency interval between the two main oscillating modes in P1 state is limited by the intrinsic laser parameters such as linewidth enhancement factor. Consequently, adopting an appropriate linewidth enhancement factor parameter can be beneficial to allocate one smaller beating frequency, i.e. achieving a bigger tuning range. The maximum microwave frequency is limited by the bandwidth of the electrical spectrum analyzer and thus can be not monitored correspondingly. It can be forecasted that the widely-tunable microwave signals with frequencies up to several hundred GHz can be realized with this method in principle if not considering the bandwidth of the photodetector. However, when concerning the high-frequency region, the extrinsic response of the parasitic network in laser packaging technology would be a challenging problem, which would deteriorate the modulation response of the packaged MI-DFSL. Specially, one point that needs a significant emphasis here is that the dual-mode oscillation in the MI-DFSL may be not limited to the P1 state, and the two lasing wavelengths without any coherence can be also included here. Different from the traditional OEO, the two uncoherent lights in the free-running state can be undamped oscillated continuously. The SAIL effect would function as a tunable MPF though the bandwidth of the original PPR induced equivalent MPF may be not narrower as that in P1 state. Thus, in theory, the central frequency of the MPF can be increased infinitely due to the infinite frequency interval induced modulation response enhancement. It will be the next research topic, which will significantly increase the flexibility of this approach.

Besides the frequency tuning range, the phase noise of the generated microwave signal is also worthy of being discussed. Compared to the discrete optical injection system with all-optical gain [14], the phase noise based on the monolithic integrated optoelectrical device can be fairly good. However, compared to the traditional OEO schemes based on the multistage EAs, the phase noise is not impressive. It mainly attributes to the limitation of the maximum of the input power of the photodetector, which highly corresponding to the link gain. Thus, it can be forecasted that a lower phase noise would be realized once this characteristic of the adopted wideband photodetector is improved.

Additionally, this system can be further optimized and towards the direction of the integration. In the next step, the amplification effect of the EDFA can be replaced by one semiconductor optical amplifier (SOA) and the SOA can be integrated into the MI-DFSL by the active waveguide. In this case, the integrated device can produce multiple effects and simultaneously work as one optical source carrier, external modulation, EBPF and the EAs. In addition, the bulky fiber loop, which is not easily integrated can be replaced with other miniaturized optical energy storages to provide high performance Q-factors, such as whispering gallery mode resonators (WGMRs) [17]. Thus, the system can be more simplified.

To conclude, in this paper, we have proposed and experimentally demonstrated a novel low-cost photonic solution to generate wideband microwave signals with all-optical gain OEO. By making full use of the PPR effect induced modulation response enhancement, the high-speed EOM as well as the tunable EBPF can be eliminated. By taking advantage of the EDFA, the necessary gain in generating ultra-high-frequency microwave signals would be superiorly settled. In this case, the generated microwave signals can be tuned in a wide range from the lower bound to upper bound of the P1 frequency by applying different combinations of control currents in principle. The SSB phase noises of the generated microwave signals at the frequency of 17.2 GHz is -106.363 dBc/Hz at 10 kHz and -83.61 dBc/Hz at 1 kHz, confirming the effectiveness of this scheme. Overall, the factors in bandwidth limitation in traditional OEO configuration are broken out, leading to the OEO possible to develop to the direction of a higher frequency.