3.1. Fundamental RF Signal Injection Locking
The baseline performance of the tunable OEO prior to injection locking was first evaluated.
Figure 3a,b show the RF spectrum for different spans and
Figure 3c shows the phase noise of the 10 GHz output. The RF signal exhibits high spectral purity with a spurious suppression ratio of approximately 75.9 dB. The side mode spacing of
corresponds to a short loop length of 300 m. The phase noise reaches −133.83 dBc/Hz at a 10 kHz offset from the carrier frequency. By adjusting the current of the YIG filter, the free-running OEO can be tuned within the range of 2–22 GHz.
Figure 4a presents the RF spectrum of the tunable OEO with a 1 GHz frequency step, while
Figure 4b shows the corresponding optical spectra at several selected frequencies. The overlay of single-sideband phase noise curves across the entire tuning range is shown in
Figure 4c. Throughout this range, the phase noise at a 10 kHz offset from the carrier frequency remains below −130.02 dBc/Hz.
Despite its superior phase noise performance, the free-running OEO suffers from unsatisfactory low-frequency stability. As indicated by the black dash-dotted line in
Figure 5a, it is difficult to identify a clear center frequency of the signal within a measurement span of 1 kHz. The phase noise at a 10 Hz offset, as depicted in
Figure 5b, approaches 0 dBc/Hz; however, its reliability is compromised due to frequency drifting, making it difficult to discern accurately.
Subsequently, the OEO was injection-locked by its fundamental frequency, with the injection signal’s phase noise measuring −75.87 dBc/Hz @ 10 Hz offset and −125.55 dBc/Hz @ 10 kHz offset, respectively. Following injection locking, a well-defined and stable RF tone (blue solid line) emerges within the identical measurement range. The phase noise decreases to −43.57 dBc/Hz @ 10 Hz (blue solid line), demonstrating a significant improvement of nearly 40 dB, as illustrated in
Figure 5b. At frequencies far from the carrier, the phase noise remains close to that of the free-running OEO (e.g., −133.84 dBc/Hz @10 kHz), better than that of the injected fundamental RF signal. The comparison of Allan variance before and after injection locking is shown in
Figure 5c. The Allan variance of the OEO decreases from 3.40 × 10
−17 at 0.1 s in the free-running state to 3.99 × 10
−22 at 0.1 s, representing an improvement of five orders of magnitude.
Next, the phase noise and Allan variance of the OEO under frequency tuning are evaluated.
Figure 6 summarizes and compares the phase noise at a 10 kHz offset from the carrier frequency and the Allan variance at 0.1 s for the wideband-tunable OEO in both the free-running and fundamental injection-locking states. The phase noise remains nearly identical to that of the free-running OEO under RF injection, staying below −130 dBc/Hz, while the Allan variance improves by approximately five orders of magnitude across the entire tuning range compared to the free-running state.
All the Allan variance measurements in this article were constrained to a maximum duration of 0.1 s due to the measurement limitations of the FSWP. To evaluate long-term frequency stability beyond this constraint, the output frequency drift of the OEO was characterized by recording its peak position over a 5 min period using the spectrum analyzer’s MaxHold function, as shown in
Figure 7. The resulting spectrum demonstrates a stable peak frequency and almost no frequency drift is observed within 5 min, which confirms the OEO system’s long-term stability.
The influence of the injection power on the performance of the OEO is also investigated.
Figure 8a,c depict the relationship between the phase noise and the Allan variance of the injection-locked OEO under different injection power levels. As the injection power increases from −24 dBm to 15 dBm, the phase noise @ 10 Hz decreases from −36.06 dBc/Hz to −58.40 dBc/Hz, showing a minimum 36 dB improvement compared to that of the free-running OEO. The phase noise far from the carrier frequency gradually converges to that of the free-running OEO.
Figure 8b summarizes variation in the phase noise at a 10 Hz and 10 kHz offset from the carrier frequency with the injected RF signal power. Similarly, the improvement in Allan variance also increases with higher injection power, particularly for time intervals longer than 1 μs. Notably, for a time interval of 0.1 s, a minimum improvement of four orders of magnitude has been achieved, even with very weak injection.
Figure 8d summarizes the variation in the Allan variance at 0.1 s under different injection powers. Investigations of the 15 GHz and 20 GHz cases also demonstrate similar performance.
Finally, the minimum RF power required for the fundamental injection locking of the OEO was experimentally investigated. Measurements were taken at 2, 5, 10, 15, and 20 GHz, as shown in
Figure 9a. Across the entire tuning range of the OEO, the minimum power required for fundamental injection locking gradually rises from −50 dBm @ 2 GHz to −18 dBm @ 20 GHz due to degradation of the system response as frequency increases.
Figure 9b,c show the phase noise and Allan variance of the OEO system in the minimum-RF-power injection-locking state, respectively. Notably, across the entire tuning range, the phase noise of the OEO system remains the same as that in the free-running state, while the Allan variance is significantly improved, consistent with the earlier experimental findings.
3.2. Subharmonic RF Signal Injection Locking
Fundamental injection locking is effective in reducing close-to-carrier phase noise and improving the Allan variance of the OEO. To further reduce system costs associated with the wideband synthesizer, a subharmonic injection-locking scheme is explored. This approach allows a synthesizer operating at the fundamental frequency of the OEO to achieve the desired frequency tuning range. For the OEO system, the n-th harmonic of the injected RF signal can be directly obtained from the Mach-Zehnder modulator, using its inherent nonlinear response. Based on the previous investigation, which demonstrated that the minimum injection power for fundamental injection locking can be reduced to −20 dBm, it is feasible to obtain the n-th harmonic of the subharmonic injection signal with sufficient power to effectively lock the OEO.
The subharmonic injection-locking range can be expressed as [
14]:
The locking range is also related to the
Q factor and the injection power of the injected signal.
The OEO was first investigated using the
,
,
, and
subharmonics of 10 GHz, each with a power of −7 dBm.
Figure 10 presents the phase noise and Allan variance of the signal generated by the OEO under different subharmonic injection-locking conditions. As shown in
Figure 10a, regardless of the injected subharmonic frequency, the phase noise at a 10 Hz offset from the carrier frequency is reduced by approximately 40 dB, and the phase noise curves of the subharmonic injection-locked OEO system remain highly consistent within a 1 kHz offset from the carrier frequency. Beyond 1 kHz, however, the phase noise of the OEO system increases after subharmonic injection locking. As
. increases, the phase noise curve at frequencies farther from the carrier also rises.
Figure 10b summarizes the phase noise at a 10 kHz offset from the carrier frequency under different injection orders. The phase noise gradually deteriorates, increasing from −133.83 dBc/Hz in the free-running state to −131.49 dBc/Hz @ 5 GHz injection, −129.77 dBc/Hz @ 3.3 GHz injection, −123.43 dBc/Hz @ 2.5 GHz injection, and −113.87 dBc/Hz @ 2 GHz injection, respectively. This degradation occurs because, as
increases, the subharmonic signal becomes weaker under the same modulation depth. The practical subharmonic order depends on the extent to which the phase noise degrades to outperform that of the fundamental RF signal. For the 10 GHz case, our RF synthesizer can generate a signal with a phase noise of −125.55 dBc/Hz at 10 GHz. Based on this, the effective subharmonic order is determined to be 3, corresponding to a phase noise of −129.77 dBc/Hz at 3.3 GHz. For an RF synthesizer with poorer phase noise performance, the usable subharmonic order can be further increased if the goal is to obtain an RF signal with better phase noise characteristics than the synthesizer itself.
Figure 10c illustrates the Allan variance of the OEO within the 0.1 s measurement range before and after subharmonic injection locking. In the free-running state, the Allan variance of the OEO system at 0.1 s is 3.4 × 10
−17. After subharmonic injection, the Allan variance for all cases at 0.1 s decreases to approximately 5 × 10
−22, representing a reduction of nearly five orders of magnitude, similar to the improvement achieved with fundamental injection locking.
Next, the influence of subharmonic power on phase noise and Allan variance is investigated using
subharmonic injection. As shown in
Figure 11a, when the power of the injected subharmonic signal increases from −7 dBm to 10 dBm, the phase noise at a 10 Hz offset from the carrier frequency improves from −42.05 dBc/Hz @ −7 dBm to −59.63 dBc/Hz @ 10 dBm, achieving a maximum reduction of approximately 40 dB compared to the free-running case. At a 10 kHz offset from the carrier frequency, the phase noise remains relatively stable, ranging from −126.68 dBc/Hz to −129.77 dBc/Hz, which represents a 5~7 dB decrease compared to that of the fundamental injection locking.
Figure 11b summarizes the phase noise at 10 Hz and 10 kHz offsets for different injection power levels.
Figure 11c,d illustrate the influence of subharmonic power on the Allan variance. As the injection power increases, the Allan variance at 0.1 s improves from 5.87 × 10
−22 at −7 dBm to 1.65 × 10
−23 at 10 dBm, demonstrating a dependence on subharmonic injection power. However, for all injection power levels, the Allan variance can be optimized to five orders of magnitude.
The minimum RF power required to lock the OEO at different subharmonic orders is investigated. We define the critical injection-locking state as the point where the Allan variance of the injected OEO becomes comparable to that of the free-running OEO. The results are presented in
Figure 12. For a specific frequency, when injecting
subharmonics of different frequencies, the minimum RF power required for the OEO to reach the locked state increases as
increases. Additionally, when
is constant, the minimum RF power required for locking rises as the free-running frequency of the OEO increases. This trend is consistent with that observed in the fundamental injection-locked OEO system.