1. Introduction

Microwave Photonics (MWP) brings together the field of Radio Frequency (RF) engineering and optoelectronics by manipulating the interaction of RF waves with optical waves. The combination brings mutual benefit to both microwave systems and optical communication systems [1,2]. MWP not only enables complex signal processing that is inefficient or unachievable in the microwave domain but also creates new opportunities for optical communication systems and networks like sensors and interrogators [3–7]. One of the essential applications of MWP is Radio-over-Fiber (RoF) as it facilitates efficient and cost-effective transmission of RF signals from a central base station to remote antenna locations over long distances [8, 9].

The signal in RF domain is transferred to optical domain by using various electro-optic modulation techniques, such as phase and intensity modulation. Existing methods generally use intensity modulation of an optical carrier at the transmitter end and recover back the signal in RF domain by direct detection using a photodetector at the receiver end, also known as Intensity Modulation with Direct Detection (IM/DD) method. An optical carrier, at λ_c, acquires sidebands, at λ_c ± n Λ_rf, after modulated with an RF source at Λ_rf where n can be even or odd depending upon the DC bias of the IM. Double Sideband with Carrier (DSB+C) modulated output is one such configuration where the sidebands are located equidistantly on either side of the optical carrier [10]. Since optical fiber dispersion scales with wavelength and fiber length, both the sidebands of DSB+C signal will experience different values of dispersion during transmission and hence will acquire different phase.

The chromatic dispersion in the fiber could distort the phase relation between the sidebands resulting in a complete loss of signal if the phase shift between the sidebands is an odd multiple of π. The detection at photodetector produces two RF signals, each sideband beating with the carrier, which will be π out of phase and hence cancel out each other. Such dispersion-induced power penalty limits the use of DSB+C signal transmission over the desired distance as well as modulation frequency [11, 12]. This problem is mitigated by suppressing one of the sidebands resulting in single sideband with carrier (SSB+C) format [13–16]. In an intensity modulated carrier, sideband suppression can be achieved by choosing appropriate modulator bias. However, sideband suppression is unattainable in phase modulation scheme. In such a scheme, post-modulation suppression is essential. In SSB+C modulation format, either of the sidebands of DSB+C is used for transmitting the signal and hence the destructive beating of sidebands at the detector is alleviated. The SSB+C format also offers better spectral efficiency compared to DSB+C that requires double the spectral space to carry the same information. Moreover, SSB+C scheme can also be utilized in applications like phased array antennas, phase shifters, vector network analyzer systems and its performance is only limited by the phase decorrelation in long-haul communication networks [17–19].

One of the most common techniques used for generating SSB+C signals have been based on driving a dual-drive Mach–Zehnder modulator with a 90° hybrid RF coupler. The dual-drive requires accurate phase differential between the drives which is a challenge to achieve for broadband and also adds complexity to the transmitter architecture. The SSB+C generated using such technique is restricted to fixed suppression ratio [20]. The other common technique is a dual-parallel MZM that has two MZMs integrated into the two arms of a conventional MZM. Though the configuration offers SSB generation, the RF drive and phase should be carefully controlled over a wide frequency band for broadband operation [21]. Recently, integrated photonics based MPW has gained considerable interest, and various microwave photonics signal processing units have been demonstrated [22]. In particular, Silicon photonics is a versatile photonic integrated circuit platform that offers compact, lightweight and low-cost solutions to achieve complex functionalities on a chip. Multiple efforts have been made to generate a SSB+C signal in Silicon platform. Such attempts mostly include generation of a DSB+C signal and then suppressing one of its sidebands. It has been demonstrated using Micro Ring Resonator (MRR), whispering gallery mode resonators, phase-shifter Bragg grating and using stimulated Brillouin scattering as well [23–27]. However, these schemes do not provide reliable, tunable, and broadband operation. For example, MRR bases solution in [23] requires two coupled rings with a length difference of 4 nm whereas Bragg grating based solution in [25] has a strict requirement of accurate phase shift of Bragg gratings during the fabrication process. In addition to direct modulator tuning, SSB-C signal could also be generated by using band-pass filters with high shape factor or steep roll-off. Since such filters are crucial in many wavelength selective application, some architectures were proposed and demonstrated in silicon and other material platform [28–30]. All the reported demonstrations use higher order filters configuration that requires individual rings or stages spectrally aligned to each other. Additional resonance tuning is required to compensate for fabrication non-uniformity. Thus, it is desirable to employ a single stage filter with high roll-off. Furthermore, some of the schemes require challenging sub-nanometer fabrication accuracy.

In this letter, we demonstrate a widely tunable suppression and broadband optical single sideband generation using a simple Self-Coupled MRR (SCMRR) [31]. Using degenerate resonance mode split in a self-coupled cavity, carrier and one sideband is preferentially selected while suppressing the other sideband. We report tunable sideband suppression ratio from 6 dB to 45 dB in a frequency range of 7 GHz-13.5 GHz, dynamic range performance for the analog signal as high as 101 dB.Hz^2/3 at 500MHz is achieved. We also report digital data transmission over a 43 km single mode fiber and present bit error rate performance and signal quality with DSB and OSSB signalling.

2. Working principle

The basic principle of the sideband suppression is shown in Fig. 1(a). A continuous wave laser source is modulated with a RF signal via MZM that results in a DSB+C output. A programmable filter has been placed to suppress one of the sidebands with a rejection ratio of 50 dB. The SSB+C output is then transmitted over an optical fiber with a dispersion of 17 ps/nm/km and detected with a high-speed photodetector. To illustrate the effect of dispersion on the transmission of both DSB+C and SSB+C signals, both the formats were transmitted over a 43 km single mode fiber. The generated DSB+C signal after MZM, driven by a RF source and DC bias, is shown in Fig. 1(b). We define Suppression Ratio (SR) as the insertion loss difference between the carrier and sidebands as labelled in Fig. 1(b). The modulated signal is then passed through a programmable optical filter that selects/rejects certain wavelength band.

 

Fig. 1: (a) [i] A simple RF-over-fiber transmission block diagram with optical-SSB generating filter, a[ii] and a[iii] principle of sideband suppression using waveshping filter, LSB: Lower SideBand, USB: Upper SideBand, (b) experimentally obtained typical optical spectrum of a DSB+C signal generated from a MZM, (c) demonstration of power penalty and (d) phase response associated with DSB+C and SSB+C after passing the signal through a 43 km SMF.

Download Full Size | PDF

For DSB+C signal, the filter was kept in all pass mode whereas for SSB+C format the lower sideband was suppressed. Figure 1c and 1(d) shows the amplitude and phase response of the two signalling schemes. It has been shown that for a fixed optical fiber transmission length, power fading primarily depend on the operating frequency and dispersion. The frequency (f_c) at which complete suppression of the signal is given as [11],

The optical filter to suppress the sideband designed in this work has the transfer characteristic as shown in Fig. 1a(ii) and a(iii), and has been realized using SCMRR. SCMRR can selectively keep optical carrier and one sideband while at the same time suppresses the other sideband. It is worth mentioning that the operating range of this filter, in terms of the modulation frequency, depends upon the 3 dB linewidth of the filter transfer function peaks. A baseband RF signal with carrier RF more than the FWHM falls outside the resonance peaks and will get attenuated by the filter’s out-of-band roll-off.

The schematic of a SCMRR is shown in Fig. 2(a). It is constructed in an hourglass fashion with a Self-Coupling Region (SCR) at the center. The SCMRR has three output ports; Through Port (TP), Drop Port (DP) and Back-Drop Port (BDP). The SCR is a directional coupler with a coupling coefficient κ₂ that enables the generation of Counter-Clock-Wise(CCW) mode in addition to the Clock-Wise(CW) cavity mode. The interaction between the two modes leads to resonance splitting due to the lifting of mode-degeneracy [32].

 

Fig. 2: (a) Schematic of SCMRR. The self-coupling region is constructed using a directional coupler of length L_sc and gap g₂ that inturn determines the self-coupling co-efficient κ₂, (b)simulated response of SCMRR for 7% κ₂. The resonance split is characterized by split-ER, ER and split λ₂ – λ₁, (c) and (d) variation of the splitting at BDP and DP as a function of κ₂, (e) BDP resonance split variation for full range of κ₂, and (f) distribution of a transverse component of E-field in the cavity with symmetric/anti-symmetric coupling at λ₁ and λ₂ in the SCR.

Download Full Size | PDF

The simulated response at all the three ports with resonance splitting is shown in Fig. 2(b). The response has been characterized by three spectral characteristics (Fig. 2(b)); resonance split, split-Extinction Ratio (split-ER) and Extinction Ratio (ER). The ER refers the strength of resonance and is measured as the difference between the maxima and minima of a resonance, split ER refers to extinction of the minima in between the split wavelengths that establishes the distinction between two split resonances (λ₁ and λ₂), and resonance split is λ₂ − λ₁ specifying the extent of splitting. Mode lifting is more prominent for higher power in counter-propagating mode and as a consequence split ER_BDP < split ER_DP whereas the ER of BDP is higher than the ER of DP; ER_BDP > ER_DP. The variation of transmission as a function of κ₂ at BDP and DP is shown in Fig. 2(c) and (d). Split as well as Split-ER increases at both the ports with an increase in κ₂, whereas ER decreases in BDP response. Field distribution at the split wavelengths in the cavity is depicted in Fig. 2(e) showing the symmetric and anti-symmetric coupling in the SCR confirms the splitting due to the interaction between CW and CCW mode.

Self-coupling co-efficient κ₂ is determined by the directional coupler parameters L_sc and g_sc. SCR excites another resonant mode at the same wavelength with same decay rate in the cavity as both the modes traverse same optical length. Such modes are called degenerate modes and two counter-propagating degenerate modes in a single cavity can interact to produce spectral response completely different than a single cavity [33]. Since the amount of resonance splitting depends upon the optical power in the counter-propagating mode, a careful engineering of SCR parameters will result in different values of splitting. Variation of splitting as a function of κ₂ at BDP is shown in Fig. 2(e) for a ring with FSR of 195 GHz. No optical power appears at BDP when κ₂ = 0 and with increasing coupling the amount of optical power appearing at BDP increases and becomes maximum at κ₂ = 1. At κ₂ = 1, the cavity behaves as a single MRR where all the power is now appearing at BDP instead of DP. Since О2 = 1 will impart a phase shift of π/2 at SCR, the resonance wavelength of the cavity will also shift by FSR/2.

3. Experiment and results

The devices were fabricated on a standard Silicon-on-Insulator substrate with 220 nm Silicon layer and 2µm buried oxide with waveguide dimension of 220 nm × 470 nm. The waveguide patterns were defined using e-Beam lithography and dry etch process. A 2µm buffer layer of PECVD oxide was subsequently deposited, and thin film metal heaters (90nm/10nm Ti/Pt) were defined for local tuning of SCMRR resonance.

Figure 3(a) shows the microscope image of fabricated device with SEM image of SCR in the inset. Fig. 3(b) shows the spectral response of SCMRR for all the three ports with no self-coupling (κ₂ = 0). Due to the absence of any CCW mode in the cavity, it behaves like a MRR with transmission dips at TP and peaks at DP. A negligible amount of optical power couples to BDP due to parasitic coupling at bus-ring coupling section. However, with non-zero self-coupling (κ₂ ≠ 0), the excited CCW mode splits the resonances resulting in optical power drop at BDP as shown in Fig. 3(c). There is no extra insertion loss associated with Self-Coupled MRR as compared to MRR. However, since the cavity allows controlled excitation of counter-propagating mode, there will be optical power redistribution between DP and BDP that will reflect as reduced optical power at DP and increased optical power at BDP. The effect of self-coupling κ₂ on splitting is shown in Fig. 3(d) for SCMRR with a Free Spectral Range (FSR) of 1.51 nm. The resonance splitting increases with κ₂ from 0.47 nm at 3% coupling to 1.8 nm at 9% of coupling. Since FSR of the cavity is 1.51 nm, for split more than 1.51 nm the symmetric and anti-symmetric resonances cross-over each other. Due to the cross-over, the change in the Split-ER_BDP and Split-ER_DP should be carefully followed, since the flip in the extinction would happen due to the cross-over. Fig. 3(e) shows change in response split as a function of applied electrical power to the micor-heaters. With the local thermo-optic tuning we achieve tuning of 47MHz/mW.

 

Fig. 3: (a) Microscope image along with measured spectral response of a SCMRR when (b) κ₂ = 0, (c) when response at κ₂ ≠ 0, (d) resonance splitting as a function of κ₂ and (e) thermo-optically tuned κ₂ and splitting using micro-heater.

Download Full Size | PDF

Figure 4 shows the experimental setup used to demonstrate OSSB generation. A DSB+C signal is generated by modulating a laser source using a bulk LiNbO₃ Mach–Zehnder Modulator (MZM). The optical carrier is modulated by a pseudo-random bit stream at various bit rates (1–12.5 Gbps). The modulated signal is amplified using an EDFA followed by an Band Pass Filter (BPF) to suppress out-of-band EDFA spontaneous emission before it is transmitted through an SCMRR filter. Light is coupled in and out of the device through a grating coupler. The side-band suppressed signal output from the BDP is then transmitted through a 43 km SMF28 fiber spool. The transmitted signal is then fed to a high-speed Photodetector. The electrical and optical spectrum of the transmitted signal is characterized using an electrical and optical spectrum analyser. The bit error rate is measured using a Bit Error Rate Tester (BERT) to analyse the effect signalling scheme on various RF carrier frequency and error-rate. The same architecture is also used to characterize Spurious Free Dynamic Range (SFDR) of the system by modulating the optical carrier with two low-frequency RF signal instead of a data stream.

 

Fig. 4: Experimental setup for side band suppression. TLS: Tunable laser source. MZM: Mach–Zehnder modulator. EDFA: Erbium doped fiber amplifier. BPF: Band pass filter. PD: Photo detector. DCA: Digital component analyzer. BERT: Bit error rate tester. ESA: Electrical spectrum analyzer. OSA: Optical spectrum analyzer. PRBS: Patterned random bit sequence.

Download Full Size | PDF

Since the resonance split of SCMRR can be tuned by changing κ₂, the optical power in the CCW will change that results in different value of split-ER and ER at both the ports. Hence, it becomes important to study the effect of resonance split on the bandwidth of the device. The experimental setup for bandwidth calculation is shown in Fig. 5(a) where a light source is modulated using LCA (43.5 GHz) and given to the TP of SCMRR. S₂₁ parameter at BDP is then measured and is shown in Fig. 5(b) and 5(c) for two values of resonance split. For a resonance split of 0.38 nm and split-ER of 1 dB at BDP, the 3 dB bandwidth will be 47.5 GHz (43.5 GHz shown here is limited by the capability of the network analyser used). The DP shows lower bandwidth (10 GHz) than the BDP due to larger split-ER. As we increase the splitting to 0.55 nm, the split-ER at both the ports increases that reduces the 3 dB bandwidth of the system. Both DP and BDP now have a bandwidth of 10 GHz. It is clear from the observation that resonance split effects the bandwidth of the system and a device having small-ER should be used to provide large bandwidth.

 

Fig. 5: Bandwidth measurement at DP and BDP for two different values of resonance split where the resonance split has been controlled by varying the self-coupling, (a) experimental setup for calculating S₂₁ parameter, LCA: Lightwave Component Analyzer, (b) and (c) S21 parameter for different resonance splitting of SCMRR and hence different split-ER at DP and BDP.

Download Full Size | PDF

Figure 6(a) shows the Optical Sideband Suppression Ratio (OSSR) of a DSB+C signal after passing through the SCMRR filter. The inset shows the suppression of either of the sidebands depending on the carrier alignment. The OSSR depicted in Fig. 6 is measured from an SCMRR whose spectral response is fixed. Since the spectral response is fixed, lower frequencies yield lower suppression ratio that steadily increases with frequency and saturates. The suppression depends on the position of the carrier and the sideband relative to the spectral character of the SCMRR. At lower frequencies, the to-be-suppressed sideband falls very close to the split peaks but at the slope of the resonance. The saturation occurs when the BDP extinction has already suppressed the entire sideband and increasing the modulation frequency will not further affect the suppression. The OSSR would reduce for frequencies beyond the Δsplit, which could be as high as >100 GHz (Fig.3(c)).

 

Fig. 6: (a) OSSR as a function of modulation frequency. Carrier and sideband (USB and LSB) position in (b) uncompensated SCMRR and (c) – (e) compensated SCMRR. The resonance splitting is varied in compensated case whereas it remains static in uncompensated case.

Download Full Size | PDF

Figure 6b shows a schematic of the working principle of OSSB generation with static SCMRR spectral response referred to as uncompensated scheme. With the carrier aligned to one of the resonance, increase in the modulation frequency would move the upper and lower sidebands away from the carrier. Since the carrier is aligned at λ₀ and LSB at λ₀ − Δλ, the insertion loss at the ER of BDP suppresses the USB as shown in Fig. 6(b). As the frequency increases the IL of the USB increases further resulting in better suppression, however, due to the static nature of the device, LSB would also be suppressed resulting in deteriorate Signal-to-Noise (SNR) ratio at the receiver. The method to mitigate this issue is to tune the SCMRR split with the modulation frequency. Figure 6(c) – 6(e) shows a scenario where the optical carrier and USB is always aligned to the SCMRR response while the LSB is suppressed. Such a configuration would ensure better SNR that can be achieved by thermo-optically tuning the Δsplit using thin film micro-heaters.

Figure 7a–7d shows the effect of uncompensated and compensated SCMRR on the unmodulated RF signal quality at the receiver. The received signal remains distortion less for frequencies below 5 GHz and starts to deteriorate and becomes undetectable beyond 13 GHz in an uncompensated SCMRR. The deterioration of the signal depends upon the 3 dB bandwidth and roll-off at the BDP. A roll-off of 90 dB/nm is measured from the spectral response of the SCMRR that translates to a complete suppression of the to-be-suppressed sideband and 20 dB suppression of the desired sideband at 20 GHz. In addition, propagation loss and additional noise further deteriorate the signal. Figure 7b shows the IL and ER of the electrical spectrum that clearly indicates that that the IL increase while the ER reduces with increasing frequency. However, with a compensated SCMRR, the ER and IL loss is maintained relatively stable(±3 dB) over 1–20 GHz (Fig. 7(d)) range by thermo-optically tuning the Δsplit to match the incoming frequency. Figure 7c shows the recovered RF signal using a tunable SCMRR where the signal quality remains relatively constant for the entire modulation frequency range 1–20 GHz. This demonstrates the tunable nature of the proposed OSSB generation scheme. It has to be noted that the maximum frequency range used for the demonstration is limited by the availability of test instrument. As mentioned earlier, the configuration could be used for higher frequencies as well.

 

Fig. 7: Received RF signal in time domain and spectral characteristics of uncompensated (a and b) and compensated SCMRR (c and d), (b) inset shows received signal at 3 GHz with ER and IL variation for uncompensated case, (e) SFDR measurement at 500 MHz and (f) variation of SFDR as function of modulation frequency.

Download Full Size | PDF

A crucial feature of a microwave link is identification of weak signals in presence of other large RF signals. It is determined by calculating SFDR of the system. The major limiting factor of SFDR is the third order intermodulation produce (3IMD) that creates spurs in the frequency band and cannot easily be filtered. Higher forms of IMD product are generally filtered using various filters. However, since 3IMD product is close to the frequency of interest, minimizing the power level of 3IMD signal is essential to improve SNR. The SFDR performance is measured using two-tone measurement that are 75 MHz apart. The power distributed between the IMD product and the fundamental frequency is used as a metric to calculate the SFDR performance. The experimental setup is shown in Fig. 4. The two tones are combined using a 50:50 RF combiner and fed to the optical modulator that generates intermodulation products spaced at 2 f₁ − f₂ = 425 and 2 f₂ − f₁ = 650, where f₁ and f₂ are RF tone frequencies; 500 MHz and 575 MHz. Figure 7e shows the power in the 3IMD with varying fundamental tone power and SFDR measurement as well. We measure an SFDR of 100.3 dB.Hz^2/3 after extrapolating the 3IMD power and identifying the crossover with the phase noise floor at the fundamental frequency; 500MHz. The noise floor was measured to be −146 dBm. Figure 7f shows the stability of SFDR over 1-5 GHz frequency range. We observe a stable SFDR of 99.9±1.05 dB.Hz^2/3 over the measured frequency range.

The data transmission performance of SSB+C against DSB+C is characterised using the test setup shown in Fig. 4. Chromatic dispersion induced power penalty equally affects high data rate signals as much as it affects analog links and hence we have analyzed SCMRR generate OSSB performance for varying data rates. The quality of the received signal is analysed with Bit Error Rate Tester (BERT) and eye diagram measurements. A 2¹⁵ – 1 pseudo-random bit stream (PRBS) signal is generated at various rates (1–12 Gbps) and used to modulate the laser source. The transmitted of the modulated signal without (DSB+C) and with (SSB+C) the SCMRR filter is characterized. The signal after transmitted through a 43 km fiber is detected using a high-speed photodetector and analysed for BER using a BERT. A part of the signal is also analysed using a scope as well.

Figure 8a shows BER of a 500 Mbps DSB+C and SSB+C signal transmission through a 43 km fiber. The BER shows better performance for DSB+C than SSB+C. The received power for error-free operation in SSB+C is 2.1 dBm more than DSB+C. It can be attributed to the reduced power in SSB+C due to the suppression of one of the sidebands and hence the requirement of more power for the same level of error-free operation. However, BER performance of DSB+C in the vicinity (8.5 Gbps) of high power penalty frequencies suffers severely as shown in Fig. 8(c). At 8.5 GHz, the penalty suffered by DSB+C signal is 7 dB as shown in Fig. 8(b) compared to SSB+C signal and hence power requirement escalates for DSB+C. The BER performance of the SSB+C generated using an SCMRR at variation modulation frequencies is shown in Fig. 8(d). An error-free transmission occurs at almost same optical power for varying frequencies with a power variation of <1 dB. Eye diagram variation for the compensated case is shown in Fig. 9. It remains open for a frequency variation of 1 GHz to 12.5 GHz. As mentioned earlier the data rates shown here is limited by the characterisation facility limitation. The proposed SSB+C generation scheme could be extended to applications higher frequencies as well.

 

Fig. 8: (a) BER of a 500 MHz SSB+C and DSB+C signalling after transmission over a 43 km fiber, (b) power penalty of SSB+C and DSB+C, penalty at 8.5 GHz (c) BER at 8.5 GHz for both the formats, and (d) BER at various frequencies for SSB+C modulation format.

Download Full Size | PDF

 

Fig. 9: Eye diagram at different frequencies for compensated case. The Signal to noise ration is constant as the eye is wide open for 1Gbps-12.5 Gbps.

Download Full Size | PDF

4. Conclusion

In conclusion, we have demonstrated an optical single sideband generation scheme using a Self-Coupled Micro-Ring Resonator. The resonance split of the micro ring is used to selectively transmit carrier and sideband to circumvent power fading while transmitting RF signal over fiber. We have demonstrated transmission of RF signal between 1-20 GHz without power fading over a single mode optical fiber length of 43 km. We demonstrate, the tunable nature of the proposed device that could be potentially applied for higher frequency as well. In addition, we presented detailed signal quality analysis with Optical spectrum, SFDR, BER and eye diagram measurements. Through this analysis, we have demonstrated a broadly tunable versatile optical single sideband generation scheme for many RF-over-fiber applications.