1. Introduction

The cost-effective passive optical network (PON) is persistently considered to meet the demand of broadband and high-speed transmission with extreme data throughput nowadays. For the further fusion and extension between wired optical and wireless microwave access services based on microwave photonics [1, 2], the millimeter-wave (MMW) carrier generation for wireless access networks and satellite communications (46 to 56 GHz) is also considered to be integrated into the wavelength division multiplexed PON (WDM-PON) [3, 4]. Architecturally, the optical data from the central office (CO) is firstly transmitted to the optical network unit (ONU), and is then up-converted by an electrical mixer combined with a local MMW oscillator for wireless communication. However, the high expense for constructing and maintaining such system makes this method difficult to be realized, especially when the demand central frequency of the required MMW carrier for satellite communication is relatively high [5–7]. Recently, the optical heterodyne detection is comprehensively investigated as the promising solution to connect the wireless LAN with the PONs [8], which enables the simplified generation and delivery of high-frequency MMW carrier by using the dual-longitudinal-mode light source. Such a dual-longitudinal-mode light source can be obtained by using a continuous-wave laser diode (LD) combined with a sinusoidal-wave modulated Mach-Zehnder modulator (MZM) for MMW carrier generation in optical domain [9].

As early as 1998, G. H. Smith et al. have already used a 19-GHz sinusoidal-wave modulated MZM to implement a 38 GHz MMW carrier generation for transmitting a 155-Mb/s on-off-keying (OOK) data over a 5-m in free space [9]. In 2005, A. Wiberg et al. used the same approach to demonstrate 2.5-Gb/s OOK downstream transmission over a 25-km single-mode fiber (SMF) and to deliver 40-GHz MMW carrier at base station (BS) [10]. In 2006, Yu et al. demonstrated the optical carrier suppression modulation scheme to simultaneously perform 2.5-Gb/s OOK data transmission for 16 DWDM channel and 40-GHz MMW carrier generation [11]. The LD-MZM based transmitter delivered MMW carrier takes the advantages of low phase noise and outstanding frequency stability, which would be beneficial from implementing the satellite communication.

In addition, some groups have employed the injection-locking technique for the LD-MZM based transmitter setup to enhance its frequency response, suppress its relative intensity noise (RIN), enlarge its output power and narrow its modal linewidth which represents the laser coherence [12, 13]. In 2008, Hong et al. employed the LD-MZM generated dual-mode optical carrier to injection-lock two FPLDs respectively [14]. Therein, the 60-GHz MMW carrier with a single sideband (SSB) phase noise of −97.5 dB/Hz is successfully delivered to transmit the 1.25-Gb/s OOK data over a 23-km single-mode fiber (SMF). Although the LD-MZM transmitter can generate the stabilized high-frequency MMW carrier, its purity relies strictly on the LO with high aging stability and the MZM with sufficiently high frequency bandwidth, which inevitably increase the system cost and complexity.

Alternatively, a few groups proposed the use of two single-mode lasers to directly beat the high-frequency MMW carrier [15]; however, the beated carrier with huge frequency instability of 200 MHz induced by the phase [16] and intensity noises of independent laser diode characteristics [17] were also observed. Therefore, using the dual-mode injection-locking technique to suppress these noises is straightforward. In 2011, Wu et al. proposed the dual DFBLD injection-locked FPLD to generate a 170-GHz MMW carrier and for transmitting 2.5-Gb/s OOK data over a 20-km long SMF [18]. Although this method is relatively cost-effective with employing the MZM, the delivered MMW carrier is quite unstable because of the independent phase and wavelength perturbation between two DFBLDs. It requires an additional heterodyne optical phase lock loop (OPLL) to stabilize the generated MMW carrier [19]. In addition, note that no matter which dual-mode master is used to injection-lock the slave laser diode, the four-wave-mixing (FWM) effect would be induced to degrade the data quality during the fiber transmission because of the severe modal dispersion.

In this work, a dual-mode master-to-slave injection-locked colorless FPLD pair with a cost-effective compact feature is demonstrated for 18-Gb/s orthogonal frequency-division multiplexing (OFDM-PON) transmission and stable 47.7-GHz MMW carrier generation at remote node for the wireless local access network (LAN). In the proposed OFDM-PON, both the single-mode and dual-mode master-to-slave injection-locked colorless FPLD pairs are preliminarily investigated to discuss the effect of amplified spontaneous emission added behind the maser colorless FPLD on their modal characteristics and modulation performances. The trade-off between the noise suppression and throughput attenuation with the optical bandpass filter added after the optically amplified master is observed. The functionality of the AWG-based DWDM multiplexer at CO is also characterized to optimize the 18-Gb/s 64-QAM OFDM-PON. The discussion on the resonant FWM effect is also emphasized. The spectral stability of the 47.7-GHz MMW carrier beat from the dual-mode master-to-slave injection-locked colorless FPLD pair is determined and compared with that synthesized from the two DFBLDs injection-locked colorless FPLD.

2. Experimental setup

The experimental setup of a dual-mode master-to-slave injection-locked colorless FPLD pair for simultaneously transmitting the 64-QAM OFDM data and generating the millimeter-wave carrier is illustrated in Fig. 1. After single- or dual-mode filtering, at optical line terminal (OLT), the master colorless FPLD with a higher front-facet reflectance of 2.5% was employed to injection-lock the slave colorless FPLD with a lower front-facet reflectance of 1%. The higher front-facet reflectance reveals stronger cavity effect to provide higher coherence and peak power of longitudinal modes for master-to-slave injection-locking operation [20]. With such a master, the slave colorless FPLD would benefit from a better injection-locking performance. In contrast to the master, the lower front-facet reflectance not only makes the slave colorless FPLD a wider gain spectrum with flexible mode selectivity, but also provides the slave colorless FPLD a higher injection-locking efficiency. Both the master and slave colorless FPLDs have the same cavity length of 900 μm, they were operated at a temperature as low as 21°C so as to enhance the external quantum efficiency. An Erbium doped fiber amplifier (EDFA) was added behind the master colorless FPLD to increase the injection power. In addition, a polarization controller was employed to achieve the efficient polarization matching between master and slave colorless FPLDs.

 

Fig. 1 The dual-mode master-to-slave injection-locked colorless FPLD pair for OFDM-PON transmission and MMW carrier generation simultaneously.

Download Full Size | PDF

To enable the maximal data transmission capability, the electrical baseband 64-QAM OFDM data generated by an arbitrary waveform generator (70001A) was used to directly encode the slave colorless FPLD through a bias-tee (HP 33150A). After 25-km single-mode fiber (SMF) transmission, the optical 64-QAM OFDM data was received by a conventional photodiode (PP-10G) to discuss its baseband performance with a digital phosphor oscilloscope (DPO, Tektronix 71604C). Besides, the 47.7-GHz MMW carrier with carried QAM OFDM data was optoelectronically beat by a high-speed photodiode (XPDV2020R) and amplified by a RF amplifier (Quinstar QLW-33505540). To frequency down-convert the QAM OFDM data, a local oscillator (Anritsu MG3692C) with a central frequency of 4.77 GHz was sent to a waveguide harmonic mixer (Keysight 11970Q) with its up-scaled frequency ranged from 33 to 50 GHz at a bandwidth of 1.3 GHz and a harmonic number of 10. Subsequently, the MMW-to-baseband down-converted 4-QAM OFDM data was passed through a post-amplifier (Coaxial ZKL-1R5 + ) and an electrical filter (Mini-Circuits 15542) to amplify its gain and suppress its noises, respectively, which is finally received by the DPO and decoded in a homemade MATLAB program.

3. Results and discussions

The multi-mode master colorless FPLD exhibits inherently strong intensity noise because of its weak resonant feature; however, it may affect the injection-locking performance. Therefore, the 1st stage optical bandpass filter (1st OBPF) with a 3-dB passband of 0.46 nm is tentatively used to avoid such an intensity noise for the master colorless FPLD, as shown in Fig. 2. To independently discuss the functionality of the AWG-based DWDM multiplexer shown in Fig. 1, it can equivalently be functioned as two OBPFs (2nd and 3rd OBPFs), as architectured in Fig. 2. Since the filtered master colorless FPLD may conditionally require a high-gain EDFA to guarantee the sufficient injection power, it could induce amplified spontaneous emission (ASE) related intensity noise to degrade the injection-locking performance. The AWG-based 2nd OBPF can be effectively minimized the residual intensity noise induced by EDFA. Moreover, the AWG-based 3rd OBPF helps to filter out the other side-modes of the injection-locked slave colorless FPLD for further suppressing the modal dispersion during long distant fiber transmission.

 

Fig. 2 Experimental setup of the dual-mode master-to-slave injection-locked colorless FPLD pair with different-stage OBPFs for 64-QAM OFDM transmission.

Download Full Size | PDF

3.1 Typical single-mode master-to-slave injection-locked colorless FPLD for 64-QAM OFDM transmission

To compare the performance between the single-mode and the dual-mode master-to-slave injection-locking, the master colorless FPLD firstly needs to be single-mode or dual-mode filtered; however, the dual-mode peak power suffers from the limited 3-dB passband of a typical 50-GHz DWDM multiplexer which is around 0.46 nm. Therefore, an EDFA is conditionally added to guarantee the sufficient injection power. To further avoid the inherent intensity noise of the master colorless FPLD from affecting the amplified injection-locking performance, the 1st OBPF is added in front of the EDFA. For comparison, the effects of the added EDFA on the output dynamics of the injection-locked slave colorless FPLD, the single-mode master-to-slave injection-locking is experimentally investigated.

The power-to-current (P-I) curve of the slave colorless FPLD injection-locked by the 1st OBPF filtered master colorless FPLD with and without EDFA amplification is shown in Fig. 3(a). Originally, the single-mode master colorless FPLD output obtained after 1st OBPF filtering can only achieve the highest power of −3 dBm for supporting the injection-locking. The injection power can easily be increased up to 3 dBm with the aid of EDFA, although it inevitably adds numerous spontaneous emitting photons into the slave colorless FPLD cavity. Whether the EDFA is employed or not, the continuously increased injection power makes the injection-locked colorless FPLD decreasing its threshold current and enlarging its output power, which helps to increase the power budget of whole system. In addition, the frequency response of the injection-locked colorless FPLD shows that its power-to-frequency slope rolls off as the injection power increases no matter the injection power is amplified or not, as shown in Fig. 3(b). The modulation bandwidth of the injection-locked colorless FPLD is extended when enlarging the injection power from −9 dBm to −3 dBm without EDFA amplification; however, the frequency response of the injection-locked colorless FPLD only enlarges its throughput intensity but maintains its relaxation oscillation frequency at 7.2 GHz by further enlarging the injection power from −3 dBm to 3 dBm under amplification.

 

Fig. 3 (a) The power-to-current curve and (b) the frequency response of the slave colorless FPLD injection-locked by the 1st OBPF filtered master colorless FPLD with or without EDFA amplification.

Download Full Size | PDF

To discuss the EDFA induced relative intensity noise (RIN) during injection-locking, the RIN spectrum of the injection-locked slave colorless FPLD under different injection levels are shown in Fig. 4. Note that the relaxation oscillation frequency related RIN peak of the injection-locked slave colorless FPLD is up-shifted to 8.4 GHz and the RIN level is also suppressed to −104.5 dBc/Hz by increasing the injection power of the 1st OBPF filtered master to −3 dBm, as shown in Fig. 4(a). Abnormally, the Fig. 4(b) shows that the RIN level of the slave colorless FPLD injection-locked by the amplified single-mode master is inversely increased to −101.5 dBc/Hz by further enlarging the injection power (corresponding to enlarge the EDFA gain) to 3 dBm, as the spontaneous emitting photons induced the additional intensity noise whereas the frequency and power of the RIN peak nearly unchanged. Especially, the RIN level under 3-dBm injection-locking is even higher than that at free-running condition. To summarize, a high RIN peak results in a deteriorated transmission performance; therefore, the disadvantage of using EDFA becomes distinct, even though although the injection power and related system power budget can be increased concurrently. At the same injection power of −3 dBm, the 1st OBPF filtered master without EDFA amplification provides the slave colorless FPLD a lower RIN level and an upper shifted RIN peak when comparing with that obtained under the EDFA amplified case.

 

Fig. 4 The RINs of the slave colorless FPLD injection-locked by the 1st OBPF filtered master (a) without and (b) with EDFA amplification.

Download Full Size | PDF

In detail, the Fig. 5(a) shows the optical spectra of the slave colorless FPLD injection-locked by the single-mode master without and with EDFA amplification. Without using the EDFA, the larger injection power effectively leads to higher the side-mode suppression ratio (SMSR) for the injection-locked slave colorless FPLD. However, the inversely decreasing trend on the SMSR response is also observed when the EDFA is added to increase the injection power because the ASE noise hardly supports the single-mode master a purified low noise output. For quantitative discussion, the Fig. 5(b) shows that the peak power of the injection-locked mode is slightly increased from −4 dBm to −3.6 dBm when increasing the injection power from −9 dBm to −3 dBm, and the highest side-mode power is also suppressed from −42.8 dBm to −45 dBm, which improves the SMSR from 38.8 dB to 41.4 dB. In contrast, by using the EDFA to increase the injection power of the single-mode master from −3 to 3 dBm, the injection-locked slave colorless FPLD inversely decreases its peak power from −5.7 to −6.2 dBm and increases its highest side-mode power from −42.7 to −40.8 dBm such that its SMSR is inevitably deteriorated from 37 dB to 34.6 dB.

 

Fig. 5 (a) The optical spectra of slave colorless FPLD injection-locked by the filtered master colorless FPLD without and with EDFA amplification; (b) The slave SMSR and the peak power of injection-locked and highest side-mode.

Download Full Size | PDF

The slave colorless FPLD injection-locked by the single-mode master without or with EDF amplification is directly encoded by the 24-Gb/s 64-QAM OFDM data with a bandwidth of 4 GHz. To enlarge the transmitted data quality, the SNR of 24 Gb/s 64-QAM OFDM data delivered by the single-mode master-to-slave injection-locked colorless FPLD pair is shown in Fig. 6. The declined SNR slope mainly results from the frequency response of the injection-locked slave colorless FPLD. As shown in Fig. 6(a), the average SNR is obviously enhanced from 22.5 dB to 24 dB when enlarging the injection power from −9 to −3 dBm, indicating that the sufficient injection power accompanied with low intensity noise is important. Unfortunately, the EDFA assisted amplification of the injection power from −3 to 3 dBm not only increases the optical gain but also induces the ASE noise to deteriorate the SNR from 24 dB to 23.8 dB still higher than the FEC criterion of 21.5 dB. In addition, the SNR is dependent on the biased current of the slave colorless FPLD, as shown in Fig. 6(b). When increasing the biased current from 42 mA to 51 mA, the average SNR is finitely enhanced from 21.5 dB to 24 dB due to the limited enhancement on because of the extended modulation bandwidth and SMSR. On the other hand, the average SNR inversely deteriorates from 24 dB to 21 dB when the biased current is overly increased from 51 mA to 60 mA, as the severe gain competition occurs in the slave colorless FPLD cavity to degrade the injection performance.

 

Fig. 6 The SNRs of the slave colorless FPLD injection-locked by the single-mode master colorless FPLD (a) at different injection powers and (b) biased currents.

Download Full Size | PDF

Note that the decreased SNR of the 64-QAM OFDM data at low frequencies mainly results from the bias-tee (Agilent 33150A) which sets a lower cut-off at frequencies below 100 kHz. Although the SNR degradation can be minimized by up-shifting the central frequency of the 64-QAM OFDM data, which would face another problem on the insufficient throughput intensity of the colorless FPLD at high frequencies. As evidence, the constellation plots and related BERs for the transmitted 24 Gb/s 64-QAM OFDM data are shown in Fig. 7. When the injection power decreases to −9 dBm, the insufficient SMSR and modulation bandwidth blurs the constellation plot to result in an EVM of 8.2% and a BER of 2.2 × 10⁻³. At optimized injection power of −3 dBm without using the EDFA, the lowest BER of 1.5 × 10⁻⁴ is observed, which is better than that using the EDFA to achieve the same injection power. The clearest constellation can also be observed with an EVM of 6.3% in this case. As the injection power overly increases to 3 dBm by EDFA, the power-to-frequency slope of the injection-locked slave colorless FPLD dramatically rolls off as the EDFA induces strong intensity noise, which seriously degrade the transmitted data quaintly to cause an EVM of 7.5% and a BER of 1.1 × 10⁻³ even though it still can pass the FEC criterion.

 

Fig. 7 The constellation plots and related BERs of the 64-QAM OFDM data generated from the slave colorless FPLD injection-locked by the single-mode master at different injection powers and biased currents.

Download Full Size | PDF

In addition, increasing the DC bias of the injection-locked colorless FPLD from 42 to 51 mA eliminates the waveform clipping effect of the transmitted 64‐QAM OFDM data to improve its BER from 2.8 × 10⁻³ to 1.5 × 10⁻⁴. However, the BER is inversely degraded from 1.5 × 10⁻⁴ to 3.9 × 10⁻³ by further enlarging the DC bias from 51 to 60 mA as the extinction ratio of the 64-QAM OFDM data decreases by the finite modulation depth of the injection-locked colorless FPLD. Note that the large tolerances on biased current range of about 18 mA facilitates the flexible application of such an injection-locked slave colorless FPLD at a FEC required BER criterion of 3.8 × 10⁻³. In brief, without EDFA amplification, the 51-mA biased slave colorless FPLD under −3 dBm injection-locking has already performed the optimized output dynamics with an increased SMSR of 41 dB and a suppressed RIN level of −104.5 dBc/Hz at 8.4 GHz, which successfully delivers the 64-QAM OFDM data with an EVM of 6.3%, an SNR of 24 dB and a BER of 1.5 × 10⁻⁴.

When considering the optimizing condition of the 51-mA based slave colorless FPLD which is injection-locked by the master colorless FPLD without EDFA amplification at an injection power of −3 dBm, the transmission performances of the delivered 24-Gbit/s 64-QAM OFDM data are shown in Fig. 8. Note that an average SNR of 21.8 dB and an EVM of 8.2% are observed at BtB case, and are slightly deteriorated to 21.5 dB and 8.5%, respectively, after 25-km SMF transmission, as shown in Fig. 8(a). Even after transmitting through a 25-km SMF, the obtained SNR and EVM are still fit the FEC criterion of 21.5 dB and 8.5%, respectively. At a receiving power of −3 dBm for the BtB case, the best BER of 1.5 × 10⁻⁴ can be observed. To meet the FEC required BER of 3.8 × 10⁻³, the optical receiving powers for BtB and 25-km SMF transmission cases can be reduced to −10 and −9.6 dBm, which exhibits a power penalty of only 0.4 dB because of the low fiber dispersion.

 

Fig. 8 (a) SNR, (b) BER and (c) constellation plots of the 64-QAM OFDM data generated from the dual-mode master-to-slave injection-locked colorless FPLD pair with and without the 3rd OBPF.

Download Full Size | PDF

3.2 Dual-mode master-to-slave injection-locked colorless FPLD for 64-QAM OFDM transmission

To characterize the dual-mode injection-locking performance, the Fig. 9 displays the optical spectra of the master and the slave colorless FPLDs under versatile filtering and injection-locking conditions. For the master colorless FPLD after EDFA amplification, the power gain of selected dual-mode could be suppressed when the peak power of other side-modes is significantly high. Therefore, the unwanted side-modes should be filtered out to reduce the gain competition during amplification process, as shown in Fig. 9(a). Ideally, the unwanted side-modes of the master colorless FPLD can totally be filtered by simply adding the 1st OBPF in front of the EDFA. The peak power difference on the dual-mode results from the homogeneous gain broadening of the colorless FPLD. After 2nd stage optical filtering, a noise background with its power level of −45 dBm due to the insufficient input power before the EDFA such that the residual spontaneous emitting photons left with the amplified dual-mode carrier, as shown in Fig. 9(c). To maintain the sufficiently large input, it is necessary to remove the 1st OBPF to enhance the EDFA input power although the inherent intensity noise of the master colorless FPLD may affect the injection-locking performance, as shown in Fig. 9(b). Since the gain competition between the selected dual-mode and unwanted side-modes in the master colorless FPLD cavity, the EDFA is still necessary provide a high power gain to guarantee the sufficient injection power.

 

Fig. 9 Optical Spectra of the master and the slave colorless FPLDs at different filtering and injection-locking conditions.

Download Full Size | PDF

Most important, the selected dual-mode wins the most of stimulated emitting photons in EDFA with sufficient input power, thus suppressing the ASE noise to decrease the noise background to −55 dBm after 2nd optical filtering, as shown in Fig. 9(d). To achieve the identical peak power of −3 dBm for each mode of the dual-mode injection master, the most appropriate EDFA pumping current of 120 mA is adjusted for amplifying the pre-dual-mode filtered master which is lower than that of 190 mA for the master without 1st optical filtering. Despite the fact that the EDFA with lower pumping current exhibits less ASE noise. After dual-mode injection-locking, the optical spectra of the slave colorless FPLD are shown in Fig. 9(c). The master colorless FPLD with the 1st OBPF before EDFA amplification has insufficient input power and induces residual ASE component to deteriorate the injection-locking performance of the injection-locked slave colorless FPLD. Note that the FWM signals also appear due to the dual-mode injection-locking, and are enhanced by the resonant amplification in the slave colorless FPLD cavity, as shown in Fig. 9(e). In contrast, when removing the 1st OBPF and maintaining the same injection power, the residual ASE component can be fully suppressed to improve the injection-locking performance of the slave colorless FPLD, as shown in Fig. 9(f); however, the better dual-mode injection-locking inevitably causes the stronger peak power of FWM signals inevitably increases, which eventually results in the decreased SMSR and enlarged modal dispersion to degrade the transmission performance.

Whether the filtering block is necessary or not can also be justified from the RIN of the slave colorless FPLD injection-locked by the master colorless FPLD at different operating conditions, as shown in Fig. 10. If the master colorless FPLD is amplified without 1st but with 2nd optical filter, the injection-locked slave colorless FPLD without 3rd optical filter exhibits a RIN level as low as −104.2 dBc/Hz at 10 GHz because the residual ASE component in DEFA can be minimized with sufficient input power, which is even lower than that of −99.5 dBc/Hz obtained for the master with both 1st and 2nd optical filters. Because of the multi-mode feature of the master colorless FPLD, its mode partition noise (MPN) at low frequencies becomes significant after optical filtering to affect the slave colorless FPLD. After 3rd optical filtering, the slave MPN at 10 MHz with and without 1st optical filtering deteriorates from −95.6 dBc/Hz to −86.3 dBc/Hz and from −95.8 dBc/Hz to −89.9 dBc/Hz, respectively. In principle, the injection-locking helps to suppress the MPN of the slave FPLD [21], which implies that the injection-locked slave colorless FPLD with higher SMSR could lead to lower MPN. As a whole, the slave colorless FPLD injection-locked by the dual-mode master with EDFA amplification can perform higher RIN than that at free-running condition.

 

Fig. 10 The RIN of the dual-mode master-to-slave injection-locked colorless FPLD pair with different-stage OBPFs.

Download Full Size | PDF

Figure 11 discusses the effect of 1st OBPF on the transmission performance of 18 Gb/s 64-QAM OFDM data carried by the dual-mode master-to-slave injection-locked colorless FPLD pair. With 1st optical filtering, the output SNR from the dual-mode master injection-locked slave colorless FPLD after 25-km SMF transmission degrades from 22.6 dB to 14.6 dB when comparing with the back-to-back (BTB) case. Without 1st optical filtering, the SNR degradation is only 2.8 dB between BTB (24 dB) and 25-km SMF (21.2 dB) transmissions. The different SNR degradation between two cases results mainly from the residual ASE noise and the injection-locking performance. Without 1st optical filtering, the master performs better injection-locking to allocate most power of transmitted data mainly encoded onto the selected dual-mode of the slave colorless FPLD without suffering from the FWM induced data degradation. The BER of the dual-mode master-to-slave injection-locked colorless FPLD pair carried 64-QAM OFDM data is shown in Fig. 11(b). The EDFA amplified and 1st stage filtered master helps the injection-locked slave colorless FPLD to fit the FEC required BER of 3.8 × 10⁻³ at a received power sensitivity of −8 dBm, which passes the BTB but fails after 25-km SMF as the transmitted BER is dramatically increased over the FEC criterion due to the residual ASE noise. By removing the 1st stage OBPF, the improved BER reveals a received power sensitivity of −11 dBm at BTB case and can pass the FEC criterion after 25-km SMF transmission with a power penalty of 6 dB. The Fig. 11(c) shows the related constellation plots at the same received power of −5 dBm. With 1st stage OBPF, the master-to-slave injection-locked colorless FPLD pair exhibits a blurred constellation plot with an EVM of 7.4% at BTB case and an unrecognizable constellation plot with an EVM of 18.7% after propagating through 25-km SMF. Without adding the 1st stage OBPF, the master-to-slave injection-locked colorless FPLD pair delivered 64-QAM OFDM data improves its constellation plot with a suppressed EVM of 6.3% and 8.7% at BTB and 25-km SMF transmitted cases. Obviously, the aid of 1st stage OBPF decreases the peak power of dual-mode master to cause insufficient and colorless FPLD is decreased, input to leave strong ASE noise after amplification. Removing the 1st stage OBPF successfully suppresses the residual ASE noise and efficiently favors the dual modes during gain competition. As a result, there is no need for the 1st OBPF between the master colorless FPLD and the EDFA.

 

Fig. 11 (a) SNR, (b) BER and (c) constellation plots of the dual-mode master-to-slave injection-locked colorless FPLD pair delivered 64-QAM OFDM data with and without 1st optical filtering.

Download Full Size | PDF

To suppress the FWM induced modal dispersion during 25-km SMF transmission, the 3rd stage OBPF is employed at a cost of slightly decreasing the dual-mode peak power. The dual-mode transmission performances of the 18-Gb/s 64-QAM OFDM data with or without the 3rd OBPF are compared in Fig. 12. As shown in Fig. 12(a), the BTB and 25-km SMF transmitted SNRs without 3rd stage filtering degrades from 22 dB to 19.2 dB. Adding the 3rd OBPF suppresses the modal dispersion to reduce the SNR degradation between BTB (23.5 dB) and 25-km SMF (21.7 dB) transmissions by only 1.8 dB. As shown in Fig. 12(b), the received power sensitivities of −10.8 dBm and −5 dBm at BTB and 25-km SMF transmissions without the 3rd OBPF are respectively observed, with a power penalty of 5.8 dB. The aid of 3rd stage OBPF minimizes the FMW induced modal dispersion to enhancing the receiving power sensitivities to −12.4 dBm and −10 dBm at BTB and 25-km SMF transmissions, respectively, with the decreased power penalty of 2.4 dB. As a supporting evidence, the Fig. 12(c) illustrates the constellation plots of the dual-mode master-to-slave injection-locked colorless FPLD pair delivered 18 Gb/s 64-QAM-OFDM data at the same receiving power of −9 dBm. Without the 3rd stage OBPF, the transmitted 64-QAM-OFDM data exhibits a blurred constellation plot with an EVM of 7.9% at BTB case, and is further degraded with an enlarged EVM of 10.9% after the 25-km SMF propagation due to the FWM induced modal dispersion. To improve, the 3rd stage OBPF effectively eliminates unwanted FWM side-modes so as to suppress the EVMs from 7.9% to 6.7% for BTB transmission and from 10.9% to 8.2% for 25-km SMF transmission. For single-mode master-to-slave injection-locked colorless FPLD pair, the AWG implements wavelength division multiplexed operation. For dual-mode master-to-slave injection-locked colorless FPLD pair, the AWG is employed not only for wavelength division multiplexed operation but also for band-pass filtering the unwanted side-modes output from the dual-mode injection-locking induced FWM signals.

 

Fig. 12 (a) SNR, (b) BER and (c) constellation plots of the 64-QAM OFDM data generated from the dual-mode master-to-slave injection-locked colorless FPLD pair with and without the 3rd OBPF.

Download Full Size | PDF

With optimized operation, the dual-mode colorless FPLD pair with amplified master, without the 1st stage filter and through with the DWDM multiplexer can successfully transmit 18 Gb/s 64-QAM OFDM data over 25-km SMF to exhibit an EVM of 8.2%, an average SNR of 21.7 dB and a BER of 2.2 × 10⁻³. In contrast, the single-mode master-to-slave injection-locked colorless FPLD pair without amplified master enables 24-Gb/s 64-QAM OFDM data with an EVM of 8.5%, an average SNR of 21.5 dB and a BER of 2.9 × 10⁻³ after 25-km SMF transmission. Note that the single-mode master-to-slave injection-locked colorless FPLD pair can effectively employ whole 4-GHz bandwidth to carry the 64-QAM OFDM data, which is 1-GHz broader than the dual-mode master-to-slave injection-locked colorless FPLD pair because the dual-mode carrier induces severe modal dispersion during 25-km SMF transmission. Even though, to optically generate the MMW carrier for implementing the fusion of MMW wireless LAN and DWDM-PON, the dual-mode master-to-slave injection-locked colorless FPLD pair is still a competitive and cost-effective method.

To compare, The MMW carriers beat by the dual-mode master-to-slave colorless FPLD pair and by the dual master DFBLDs injection-locked slave colorless FPLD are shown in Fig. 13(a). The dual-mode master-to-slave colorless FPLD pair delivered 47.7-GHz MMW carrier exhibits a maximum power of −61.5 dBm and a noise floor at −82.9 dBm, thus providing a carrier-to-noise ratio (CNR) of 21.4 dB at a resolution bandwidth (RBW) of 10 kHz. Note that the slightly high noise floor results from the residual ASE noise in this case. In comparison, the 47.7-GHz MMW carrier beat from the same device injection-locked by two individual DFBLD can provide a maximal power of up to −55.8 dBm and a noise floor at −84.6 dBm, which apparently increases the CNR up to 28.8 dB at same RBW of 10 kHz. Because two individual DFBLDs have already sufficient injection power without amplification, the RF spectrum performs a 1.7-dB lower noise floor.

 

Fig. 13 (a) Beat MMW carrier and (b) SSB spectra from dual-mode master-to-slave injection-locked colorless FPLD pair and dual-DFBLD injection-locked colorless FPLD.

Download Full Size | PDF

When comparing the peak power of generated MMW carrier, the MMW carrier performance of the dual-DFBLD injection-locked colorless FPLD is slightly better than that of the dual-mode master-to-slave colorless FPLD pair, because the DFBLD exhibits higher coherence than the dual-mode master colorless FPLD. Conversely, the stability of the MMW carrier generated from the dual-mode master-to-slave injection-locked colorless FPLD pair behaves an excellent half power linewidth (HPBW) of only 0.48 MHz, which is significantly lower than that obtained from the dual-DFBLD injection-locked colorless FPLD (6.8 MHz). By further enhancing the coherence of the master colorless FPLD based on bias and reflectance enhancement, the 47.7-GHz MMW carrier carried 4-QAM OFDM data at a raw data rate of 2-Gb/s after down-conversion is preliminarily shown in Fig. 14. Therein, the used bandwidth of the 4-QAM OFDM data is only 1 GHz by limitation of the harmonic mixer, and the average SNR of 9.51 dB and received BER of 1.4 × 10⁻³ has already meet the FEC criterion set at 9 dB and 3.8 × 10⁻³, respectively, accompanying with a slightly blurred constellation plot at an EVM of 33.5%.

 

Fig. 14 (a) The 2D constellation plot, related SNR of subcarriers for the down-converted 2-Gb/s 4-QAM OFDM data and (b) the 3D constellation plot and corresponding SNR of each symbol.

Download Full Size | PDF

4. Conclusion

A remote heterodyne MMW carrier over fiber based OFDM-PON with the master-to-slave injected dual-mode colorless FPLD pair is proposed for connecting wired fiber to wireless MMW networks in the near future. To meet this demand, the cost-effective and compact master-to-slave injection-locked colorless FPLD pair with dual-mode feature is demonstrated for 18-Gb/s optical 64-QAM OFDM in PON and stabilized 2-Gb/s 4-QAM OFDM with MMW carrier at 47.7 GHz in remote node. In the proposed OFDM-PON, both the single-mode and dual-mode master-to-slave injection-locked colorless FPLD pairs are compared, and the effect of amplified spontaneous emission added behind the maser colorless FPLD on their modal characteristics and modulation performances is discussed. The functionality of the AWG-based DWDM multiplexer at central office is also characterized to optimize the 64-QAM OFDM-PON. For the single-mode master without amplification, the slave colorless FPLD biased at 51 mA and injection-locked at −3 dBm has already performed the optimized output dynamics with an increased SMSR of 41 dB and a suppressed RIN level of −104.5 dBc/Hz at 8.4 GHz, which successfully delivers the 64-QAM OFDM data at 24 Gb/s over 25-km SMF with an EVM of 8.46%, an SNR of 21.5 dB and a BER of 2.9 × 10⁻³. In contrast, the dual-mode colorless FPLD pair with amplified and unfiltered master can successfully transmit 18 Gb/s 64-QAM OFDM data over 25-km SMF through the DWDM multiplexer to provide EVM, SNR and BER of 8.2%, 21.7 dB and 2.2 × 10⁻³, respectively. When transferring the salve colorless laser diode from single- to dual-mode operation, the encoding bandwidth is only sacrificed by 1 GHz due to the severe modal dispersion occurred during 25-km SMF transmission. Even though, the dual-mode master-to-slave injection-locked colorless FPLD pair is still a competitive and cost-effective method to optically generate the MMW carrier for implementing the fusion of MMW wireless LAN and DWDM-PON. The 47.7-GHz MMW carrier remotely beat from the dual-mode master-to-slave injection-locked colorless FPLD pair exhibits an extremely narrow bandwidth of only 0.48 MHz, which is significantly lower than that obtained from the same colorless FPLD injection-locked by two individually free-running DFBLDs (6.8 MHz). Finally, the 47.7-GHz MMW carrier successfully delivers 4-QAM OFDM data up to 2 Gb/s with EVM, SNR and BER of 33.5%, 9.51 dB and 1.4 × 10⁻³, respectively, after frequency down-conversion to baseband.