End-to-end real-time experimental demonstrations are reported, for the first time, of aggregated 11.25Gb/s over 26.4km standard SMF, optical orthogonal frequency division multiple access (OOFDMA) PONs with adaptive dynamic bandwidth allocation (DBA). The demonstrated intensity-modulation and direct-detection (IMDD) OOFDMA PON system consists of two optical network units (ONUs), each of which employs a DFB-based directly modulated laser (DML) or a VCSEL-based DML for modulating upstream signals. Extensive experimental explorations of dynamic OOFDMA PON system properties are undertaken utilizing identified optimum DML operating conditions. It is shown that, for simultaneously achieving acceptable BERs for all upstream signals, the OOFDMA PON system has a >3dB dynamic ONU launch power variation range, and the BER performance of the system is insusceptible to any upstream symbol offsets slightly smaller than the adopted cyclic prefix. In addition, experimental results also indicate that, in addition to maximizing the aggregated system transmission capacity, adaptive DBA can also effectively reduce imperfections in transmission channel properties without affecting signal bit rates offered to individual ONUs.
©2011 Optical Society of America
Driven by the exponentially growing demand for bandwidth-intensive multimedia services such as video on demand, video conferencing, high-definition TV (HDTV) and interactive games, recent years have seen a significant increase in transmission capacity of access networks [1, 2]. The number of internet users worldwide has increased by ~445% from 2000 to 2010 . It is also forecasted  that advanced internet video (3D and HDTV) demand will increase ~23-fold between 2009 and 2014, and that video-centric traffic will account for >91% of total global consumer traffic by 2014. Clearly, the current dominant copper-wire- and wireless-based access techniques namely very-high-speed digital subscriber line (VDSL) and WiMax  cannot satisfy the ever-increasing bandwidth requirement in such cost-sensitive application scenarios, passive optical networks (PONs) are, therefore, being widely adopted and practically implemented as a promising cost-effective and “future-proof” high-speed technical strategy for broadband access.
Generally speaking, PONs have been evolving from traditional time division multiplexing (TDM) PONs , wavelength division multiplexing (WDM) PONs , hybrid TDM/WDM PONs , and the most recently emerging optical orthogonal frequency division multiple access (OOFDMA) PONs [5–7]. A TDM PON provides a point-to-multipoint access between one optical line terminal (OLT) and a number of optical network units (ONUs) through the use of TDM to convey downstream signals from the OLT to the ONUs, and employs time division multiple access (TDMA) to multiplex upstream signals of a separate wavelength from the ONUs to the OLT. As a direct result, future high-speed TDM PONs impose serious constrains on transceiver designs in terms of ultra-fast burst mode operation function, complex scheduling algorithms and framing technologies, as well as pronounced sensitivity to packet latency. In WDM PONs, data information is transmitted over a pair of dedicated wavelengths assigned to a particular ONU for both downstream and upstream transmission directions. For widespread massive global deployment of WDM PONs, the most critical challenges are cost-effectiveness and flexibility . The first challenging issue arises due to the achievements of some daunting tasks including, for example, colorless network operation and the fundamental alteration to legacy PON optical distribution networks. Whilst the restricted flexibility of WDM PONs is mainly due to the fact that dynamic bandwidth allocation (DBA) cannot be statistically performed at a sub-wavelength granularity. In hybrid TDM/WDM PONs, a number of wavelengths are utilized in each direction to link the OLT to several ONUs, and each individual wavelength is shared among a number of ONUs rather than being dedicated to a single ONU, therefore, it is envisaged that hybrid TDM/WDM PONs inherit, to some extent, the majority of the abovementioned technical challenges associated with TDM PONs and WDM PONs, despite the fact that future TDM/WDM PONs will gradually evolve from a fixed wavelength configuration into a tunable wavelength configuration.
In OOFDMA PONs, for both downstream and upstream transmission directions, an overall channel bandwidth can be divided into a large number of orthogonal subcarriers of different frequencies, one or more of which can be assigned statistically to a specific ONU, and a subcarrier can also be further shared among different applications via TDM. Therefore, DBA can be easily achieved using the following three means: 1) assigning a certain number of information-bearing subcarriers to a single high-speed ONU; 2) allocating a single subcarrier to several low-speed ONUs at different time slots previously assigned, and 3) varying independently signal modulation formats and/or powers taken on subcarriers to perform, in an extra dimension, DBA with an extremely fine granularity . Throughout this paper, DBA with the extra dimension is referred to as adaptive DBA, which cannot only maximize the aggregated system transmission capacity, but also effectively reduce imperfections in transmission channel properties without affecting signal bit rates offered to individual ONUs. In addition, owing to the nature of “noise-like” time-domain OOFDM waveforms, OOFDMA PONs are also free from burst mode operation.
In comparison with other PONs mentioned above, OOFDMA PONs have demonstrated a number of salient advantages capable of satisfying the network carriers’ major targets for future high-speed PONs. The advantages are summarized below:
- • Capability of providing an increased number of subscribers with high signal bit rates over extended reach ;
- • Considerably improved cost-effectiveness. This is because of the full exploitation of rapid advances in modern digital signal processing (DSP) technology, and the considerable reduction in PON system complexity owing to OOFDM’s unique adaptability, excellent resistance to linear component/system impairments and efficient utilization of channel spectral characteristics [5–13];
- • Full-scale DSP-based adaptive DBA with a fine bandwidth granularity. As mentioned above, the available channel bandwidth is shared statistically between various ONUs using dimensions of frequency, time and signal modulation format. This feature can use limited channel spectral bandwidths to provide end-users with required services.
- • Backward compatibility. OOFDMA PONs can potentially support multiple TDM PON standards. This coexistence with different standards offers a seamlessly upgrading of installed legacy PONs ;
- • Excellent flexibility. OOFDM transceivers with adaptive bit and/or power loading  offer the PON systems great adaptability. This feature may not only facilitate the convergence of various access networks but also enable the accommodation of traditional heterogeneous services and newly emerging services over a common platform.
For practical deployment of cost-effective OOFDMA PONs, intensity-modulation and direct-detection (IMDD) using directly modulated lasers (DMLs) is greatly advantageous, which, however, causes the strong optical beat interference (OBI) effect upon direct detection of different ONU upstream signals of similar wavelengths . OBI produces unwanted frequency products that fall into the useful signal spectral region. To mitigate the OBI effect, use can be made of two technical solutions including, 1) the utilization of DMLs operating at different wavelengths for upstream transmission in different ONUs , and 2) the use of coherent detection instead of direct detection to convert upstream signals to the electrical domain in OLTs . As the second solution considerably increases the complexity of transceiver/PON system architectures including DSP algorithms, thus the overall PON system cost, only is the first solution considered in the present paper.
Over recent years, experimental investigations have been reported of OOFDMA PONs using off-line DSP approaches in the transmitters and/or receivers [5–7,9], which, however, do not consider either the limitations imposed by the precision and speed of practical DSP hardware, or the implementation of adaptive DBA. Experimental demonstrations of end-to-end real-time OOFDMA PONs with adaptive DBA are extremely vital for exploring the feasibility of the OOFDM technique for practical deployment in cost-sensitive and high-speed future PON systems. Unfortunately, as far as we are aware, such works have not been reported.
By making use of electrical baseband OFDM signals to modulate various commercially available, low-cost DMLs such as DFB lasers  and vertical cavity surface-emitting lasers (VCSELs) , we have successfully demonstrated experimentally end-to-end real-time 11.25Gb/s OOFDM transmission over 25km standard single-mode fibre (SSMF)-based point-to-point PON systems employing IMDD. In those real-time OOFDM systems, real product-like functions namely on-line performance monitoring and live parameter optimization have also been included. More recently, further improvements in the transceiver design have also been made by incorporating other crucial functionalities including adaptive bit and/or power loading , as well as asynchronous and synchronous symbol synchronization [10,13].
Based on modified real-time OOFDM transceivers, the thrust of the present paper is to experimentally demonstrate, for the first time, end-to-end real-time IMDD OOFDMA PONs with adaptive DBA at 11.25Gb/s over 26.4km SSMFs. The demonstrated OOFDMA PON system consists of two ONUs, each of which uses a DFB-based DML and a VCSEL-based DML. As the physical layer transmission performance is the main focus of the paper and the downstream transmission performance of the PON system is very similar to that published previously in point-to-point systems [8, 10–13], in this paper, special attention is, therefore, given to upstream transmission performances and adaptive DBA, which are the key factors for characterizing the PON system performance. In addition, extensive experimental investigations are also undertaken of the influence of practically encountered major ONU differences upon the performances of all upstream signals. These ONU differences include relative symbol offset, optical launch power and DML modulation characteristics. In addition, detailed discussions are also made of how to fully exploit the adaptability feature in DBA for a given OOFDMA PON system.
2. OOFDMA PON systems
2.1 OOFDMA PON system setup
Figure 1 illustrates schematically the experimental OOFDMA PON system considered here, which consists of two ONUs, a single OLT and a 26.4km SSMF linking a 3-dB optical coupler and a PIN in the OLT. The simple IMDD system is free from both chromatic dispersion compensation and in-line optical amplification. As our numerical simulations indicate that the incorporation of two ONUs in the OOFDMA PON system is sufficiently accurate for evaluating all the dynamic system characteristics of interest of the present paper, for simplicity but without losing generality, two ONUs are, therefore, considered in Fig. 1.
In Fig. 1, each individual ONU is composed of a field programmable gate array (FPGA)-based real-time OFDM transmitter having various DSP processes described in Section 2.2, a DML, an optical filter and an erbium doped fibre amplifier (EDFA) for fixing the optical launch power from each ONU at a required value. In the electrical OFDM transmitter, prior to performing the inverse fast Fourier transform (IFFT), encoded and adaptive power loaded user data is taken on preselected OFDM subcarriers with all the remaining subcarrier powers being set to zero. Combined with an appropriate bias current, the generated real-valued OFDM signal directly drives the DML to convert the upstream electrical signal to the optical domain. After the DML, an EDFA followed by a 0.8nm optical filter is used to set the optical launch power from each ONU at a fixed power level of 6dBm. These two upstream OOFDM signals from different ONUs are combined using a 3-dB optical coupler, and the combined optical signal propagates through the 26.4km SSMF link.
In the OLT, an optical attenuator is utilized to vary the combined optical signal power before injecting into a 12.4GHz linear PIN with a sensitivity of −19dBm (corresponding to a bit error rate (BER) of 10−10, PRBS 231-1. NRZ @10Gb/s). To precisely monitor the received optical power, a 90:10 optical splitter is inserted between the optical attenuator and the PIN. Finally, the received combined upstream optical signal is converted to an electrical signal by the PIN detector. The converted electrical signal is then fed into the real-time OFDM receiver in the OLT for data recovery.
The above-mentioned two DMLs included in the ONUs operate at different wavelengths at a wavelength spacing of 4nm. According to our numerical simulations, such a wavelength spacing is sufficiently large to completely eliminate the OBI effect associated with direct detection of the combined upstream signals in the OLT. Furthermore, to rigorously evaluate the system flexibility, performance robustness and adaptive DBA of the OOFDMA PON system, a 3.6GHz, 1547nm un-cooled VCSEL (a threshold current of 2mA) and a 10GHz, 1551nm DFB laser (a threshold current of 29mA) are used in ONU1 and ONU2, respectively, to perform E/O conversion, as seen in Fig. 1. The operating conditions of these DMLs are presented in Section 3.
2.2 Real-time OFDM transceivers
The real-time OFDM transmitter adopted in each individual ONU and the corresponding real-time OFDM receiver implemented in the OLT are shown in Fig. 2 , whose configurations are similar to those reported previously in [8, 10–13]. In the transceivers, real-time DSPs are employed to perform key transceiver functionalities including, for example, IFFT/FFT, pilot tone-based channel estimation/equalization, adaptive power loading, automatic asynchronous synchronization , live parameter optimization, as well as online monitoring of system performance such as subcarrier/total channel BER and system frequency response. It should be pointed out, in particular, that, in the transmitter FPGA shown in Fig. 2, a newly developed symbol timing offset (STO) compensation block is added prior to the ‘Signed to Unsigned’ function block. For a given upstream OOFDMA PON system, use can be made of the STO compensation block to adjust the signal time delay at a granularity of a sample time duration to synchronize different upstream signals in the OLT, as detailed in Section 2.4.
As the aforementioned transceiver functionalities and the implementation of real-time OFDM signal generation/detection have been presented in detail in [8, 10–13], thus an outline of the transceiver design together with corresponding parameters are briefed below:
In each transmitter FPGA clocked at 100MHz, pseudo random data sequences generate a stream of 84-bit (70 bits for 32-QAM) parallel words representing the data payload. Based on the approach detailed in , each of these parallel words is combined with a fixed 6-bit (5 bits for 32-QAM) pilot word for channel estimation/equalization. The combined 90-bit (75 bits for 32-QAM) words are mapped onto 15 parallel 64-QAM encoders. As discussed in Section 2.3, if Ni, i = 1, 2, represents the total number of information-bearing subcarriers assigned to the ONU, the Ni encoded parallel complex numbers are arranged appropriately, together with 15-Ni zero values, to distribute among these 15 subcarriers, according to a specific adaptive DBA requirement. Adaptive power loading is then performed to the information-bearing subcarriers to compensate the effect of the system frequency response roll-off associated with the entire transmission system from the transmitter IFFT input to the receiver FFT output . These 15 subcarriers and one zero-frequency subcarrier with a zero power are input, with their Hermitian-symmetry counterparts, to the transmitter 32-point IFFT to produce 32 real-valued samples. It should be noted that, in practice, a buffer can be used prior to pilot insertion to produce the required number of parallel words loaded on the information-bearing subcarriers.
Having completed the above-mentioned DSP procedures, a cyclic prefix of 8 samples is added to each OFDM symbol. After performing signal clipping at an optimum clipping ratio of 12.0dB  and 8-bit quantization, the digital OFDM samples generated by the real-time DSP in the transmitter are transferred as 4 parallel samples by a 32-bit wide bus running at 1GHz to an 8-bit, 4GS/s digital-to-analogue converter (DAC) for conversion to an analogue electrical OFDM signal. In combination with a suitable bias current, the real-valued, unsigned electrical OFDM signal emerging from the DAC output port is then adjusted by a variable electrical attenuator/amplifier to have a desired amplitude to directly drive the DML mentioned in Section 2.1. The above transceiver design gives the ONU a raw signal bit rate of 0.75×Ni Gb/s (0.625×Ni Gb/s for 32-QAM). The aggregated upstream raw signal bit rate is 11.25Gb/s (9.375Gb/s for 32-QAM).
In the OLT receiver, to minimize quantization noise and clipping-induced signal distortions in an 8-bit, 4GS/s analogue-to-digital converter (ADC), the baseband electrical signal converted by the PIN is amplified to an optimum level to provide a suitable amplitude prior to digitization by the ADC. All real-time DSP procedures in the receiver FPGA are identical to those reported in [8, 10–13]. With the adaptive DBA-initiated data-bearing subcarrier distribution information for different ONUs, data for each individual ONU is finally recovered. In addition, BER measurements of both individual subcarriers and total transmission channel follow the procedures reported in .
Here it is worthy addressing the following two points:
- • To effectively optimize the transmission performance of each upstream signal and simultaneously improve the DBA capability and the system insusceptibility to variations in transceiver components, adaptive power loading is used independently for each ONU transmitter. In comparison with the most sophisticated algorithms involving adaptive bit loading, the simplest adaptive power loading technique has been experimentally confirmed to be sufficiently effective in escalating the OOFDM system performance to its maximum potential .
As experimentally demonstrated in , in a transmitter with adaptive bit loading, to maintain the input data interface at a fixed bit width regardless of the selected signal modulation formats, for each subcarrier, an adaptive modulator consisting of a number of parallel modulators is employed. At the input of each of these parallel modulators, a bus bit converter is also implemented, which pads a certain number of extra ‘0’ bits into the original user bits allocated to the subcarrier to construct a data interface at a constant bit width, and before signal encoding, extracts only the user bits by removing the padded extra ‘0’ bits. In such a transceiver design, for a specific subcarrier, the exact number of extra ‘0’ bits padded by the corresponding bus bit converter is determined by the modulator selected by the feedback information, which is generated via negotiations between the transmitter and the receiver. On the other hand, in the receiver, a similar adaptive demodulator for each subcarrier is implemented, which performs an inverse of the above-mentioned transmitter DSP according to the same feedback information.
- • To prevent symbol alignment drift and ensure subcarrier orthogonality for data recovery in the OLT, the whole OOFDMA PON system should be synchronized to a common clock source. Therefore, in Fig. 2 a clock synthesizer with an internal common reference clock generates 2GHz clock signals for the 4GS/s, 8-bit DAC/ADC and 100MHz clock signals for all the transmitter/receiver FPGAs (100MHz is the OFDM symbol rate).
2.3 Adaptive DBA
Given the unique OOFDM features of subcarrier orthogonality and adaptive bit and/or power loading, adaptive DBA can be easily implemented in practice using the three approaches mentioned in Section 1. It should be noted that, based on online performance monitoring and live parameter optimization, the present OOFDM transceivers are capable of performing real-time adaptive DBA, according to both the channel quality and users’ bandwidth requirement.
As an example of experimentally demonstrating adaptive DBA in the OOFDMA PON system, here 1st – 7th subcarriers are assigned to ONU1 using the VCSEL-based DML, whilst 8th-15th subcarriers are assigned to ONU2 using the DFB-based DML, as shown in Fig. 3 , where the profiles of both data-encoding bit and adaptively loaded subcarrier amplitude are also presented for two ONUs. Because the modulation bandwidth of the VCSEL is much lower than that corresponding to the DFB laser, the allocation of low (high) frequency subcarriers to VCSEL-based ONU1 (DFB-based ONU2) not only improves the transmission performance of the entire OOFDMA PON system, but also allows a better utilization of the available channel characteristics. Throughout the paper unless mentioned explicitly in the text when necessary, this DBA example is considered for exploring in detail OOFDAM PON performance characteristics presented in Section 3.
Considering the transceiver design parameters mentioned in Section 2.2, for the aforementioned adaptive DBA scheme, the raw signal bit rates for ONU1 and ONU2 are 5.25Gb/s and 6Gb/s, respectively. It is also easy to understand that, for the present 64-QAM-encoded OOFDMA PON system, the bandwidth provided to each ONU can vary from 750Mb/s to 11.25Gb/s with a granularity of 750Mb/s. Of course, the DBA granularity can be further improved if use is made of adaptive bit-and-power loading in the OOFDM transceivers .
2.4 OOFDMA PON synchronization
It is well known that accurate synchronization of all downstream and upstream signals in OOFDMA PONs is vital for the practical realization of the systems with desired transmission performances. Given the fact that an effective asynchronous synchronization technique has been proposed and experimentally demonstrated in point-to-point real-time OOFDM systems , and that the technique can also be employed for downstream transmission in the present OOFDMA PON system, here special attention is, therefore, given to further extending the asynchronous synchronization technique for applications in OOFDMA PONs to enable accurate synchronization of different upstream signals.
A highly precise estimation of the STO associated with a upstream signal in the receiver plays a dominant role in achieving high-quality signal synchronization in multiple access networks. In conventional wireless OFDMA networks, an electrical filter is often used in the receiver to extract a wanted upstream signal from the mixed signals received. Based on the extracted signal and the synchronization technique adopted for downstream transmission, an estimated STO value is obtained, which is then sent back to the corresponding transmitter for conducting STO compensation . For an access network shared by M users, M different filters are required to simultaneously synchronize all upstream signals. Clearly, the filter-based synchronization technique is not suitable for the present OOFDMA PON systems, as an extremely large number of end-users possibly supported by the OOFDMA PONs require a huge bank of filters to be implemented. Undoubtedly, the traditional approach significantly increases the transceiver complexity and the system cost.
Making use of the previously published asynchronous synchronization technique , and considering the fact that the STO of a signal in PONs does not vary significantly over time, the following “one-by-one” approach that is experimentally approved to be very effective, is thus adopted, throughout the paper, for use in the OOFDMA PON system. The procedures of implementing the approach are listed as following:
- • In the initial phase of establishing an OOFDMA PON system, only one ONU is switched on and all other ONUs are switched off. In the OLT, use is first made of the asynchronous synchronization technique to estimate the STO of the upstream signal from the active ONU. The optimum position of the FFT window measured in the OLT is also recorded and regarded as a fixed reference for synchronizing all other ONUs;
- • The estimated STO is fed to the active ONU via the joint test action group (JTAG) interface, as illustrated in Fig. 2;
- • Upon receiving the STO information, the active ONU undertakes STO compensation in the STO compensation block by adjusting symbol offset at a resolution of a sample time duration. This leads to the accurate synchronization of the active ONU with the OLT;
- • Having synchronized the first ONU, the second ONU is switched on. Taking the recorded FFT window position in the OLT as a reference, a repetition of the aforementioned steps results in the synchronization of two upstream signals from both ONUs with the OLT. Such procedures continue until the upstream signals from all the ONUs are simultaneously synchronized successfully.
It should be addressed that, apart from synchronizing the OOFDMA PON system, the above-mentioned synchronization approach can also automatically compensate the chromatic dispersion-induced time delay between different ONUs operating at different wavelengths.
3. Experimental results
Having described the experimental OOFDMA PON system setup and the modified real-time OOFDM transceiver design in Section 2, this section is dedicated to extensive experimental explorations of dynamic OOFDMA PON system properties including, impact of DML operating conditions, ONU launch power variation range, performance susceptibility to symbol offset between different upstream signals, as well as impact of adaptive DBA on OOFDMA PON performance. These issues are of great importance for practical system design.
3.1 Impact of DML operating conditions
To gain an in-depth understanding of the influence of DML operating conditions on the OOFDMA PON performance and, more importantly, to identify optimum operating conditions for various DMLs adopted in the system, experimental measurements are first undertaken of the BER performance as a function of bias current for different DMLs in the OOFDMA PON system subject to a fixed received optical power of −8dBm in the OLT and with only one ONU being activated at a time. The measured results are shown in Fig. 4 , where bias current dependent optimum root mean square (RMS) driving currents are also presented for different DMLs.
As seen in Fig. 4, the observed BER and RMS driving current developing trends are similar between different DMLs, and the dynamic bias current ranges, over which the measured BERs less than the forward error correction (FEC) limit of 2.3×10−3  are obtainable, are from 2.5mA to 10.5mA for the VCSEL-based DML and from 34.5mA to 42mA for the DFB-based DML. For bias currents in the vicinity of a DML threshold, the BER increases sharply with decreasing bias current because of small bias current-induced strong effects of intensity modulation nonlinearity and signal clipping . On the other hand, as shown in Fig. 4(a), for excessively high bias currents, the BER performance degrades slightly due to a large bias current-induced reduction in OOFDM signal extinction ratio . In addition, considering a typical L-I curve associated with a DML, it is easy to understand that the optimum RMS driving currents roughly linearly grow with bias current, as observed in Fig. 4.
In the OOFDMA PON system with two ONUs simultaneously sending their upstream signals to the OLT, the BER of each ONU versus bias current is plotted in Fig. 5 for the VCSEL-based DML and the DFB-based DML operating at fixed RMS driving currents of 0.72mA and 1.2mA, respectively. In measuring the figure, a constant DFB bias current of 37mA and a constant VCSEL bias current of 4.5mA are adopted in Fig. 5(a) and Fig. 5(b), respectively, and for all the cases the received optical launch powers in the OLT are fixed at −5dBm.
As expected, Fig. 5 shows that a variation in DML bias current in an ONU (referred to as varied ONU) affects not only the ONU’s own BER performance but also the performance of the other ONU having a fixed DML bias current (referred to as fixed ONU). In particular, BER degradations suffered by the fixed ONUs become severe when the varied ONUs operate at bias currents close to their DML thresholds, as seen in Fig. 5. This is mainly due to the fact that, over such a region, the output optical power of the DML in the varied ONU is very low, which causes strong EDFA-generated ASE noise to be imposed onto the upstream signal from the fixed ONU. Furthermore, a DML bias current well above the threshold in the varied ONU produces a high optical signal power, which weakens the ASE noise effect, thus flattens the BER curves for the fixed ONU, as presented in Fig. 5.
From Fig. 5 it is also clear that, for the varied ONU, there exists an optimum DML bias current, corresponding to which a minimum BER is observed. The physical mechanisms underpinning the occurrence of the optimum bias current are very similar to those discussed in Fig. 4, except that, for bias currents larger than the optimum value, an extra DC component associated with the fixed ONU leads to a further reduction in signal extinction ratio of the upstream signal from the varied ONU. As a direct result, in comparison with Fig. 4, a much steep BER curve occurs for the varied ONU in Fig. 5.
From Fig. 5, the optimum DML bias currents can be identified easily, which are 4.5mA for the VCSEL-based ONU (ONU1) and 37mA for the DFB-based ONU (ONU2). Also taking into account Fig. 4, these two optimum bias currents correspond to the optimum RMS driving currents of 0.72mA for the VCSEL-based DML and 1.2mA for the DFB-based DML. These optimum DML operating conditions are adopted in all the experimental measurements presented below. It is worth addressing that the optimum RMS driving currents hold well for the present experimental system, as a variation in DML driving current does not considerably affect the system performance when the DML operates above its threshold .
3.2 ONU launch power variation range
In the OOFDMA PON system, the ONU launch power variation range is defined as, for a fixed optical power received in the OLT, the maximum allowable variation range of the optical launch power of a specific ONU, over which the BER of any individual upstream signal simultaneously transmitting in the system is still less than the FEC limit of 2.3 × 10−3. Clearly, a large ONU launch power variation range improves not only the performance robustness but also the system flexibility because of the pronounced ability of accommodating a large diversity of low-cost optical components.
To examine the ONU launch power variation range for the present OOFDMA PON system with all the DMLs operating at their optimum conditions identified in Section 3.1, Fig. 6 is presented, where BER performances of all upstream signals against the optical launch power from an individual ONU are plotted for different received optical powers in the OLT. In obtaining Fig. 6, the output power of the EDFA is adjusted to provide various required optical launch powers from the varied ONU, whilst the optical launch power from the fixed ONU is kept at a constant value of 6dBm. The optical attenuator allocated in the front of the PIN shown in Fig. 1 sets the received optical powers at different values such as −5dBm and −7dBm.
It is observed in Fig. 6 that, for a given received optical power in the OLT, an increase in optical launch power from the varied ONU improves its own BER performance and simultaneously degrades the BER performance of the fixed ONU, mainly resulting from the changes in effective optical signal to noise ratio (OSNR) of the corresponding upstream optical signals. It is very interesting to note that the ONU launch power variation ranges as large as >3dB are practically feasible, which can be further extended if a relatively high received optical power is taken in the OLT. For example, for a received optical power of −5dBm in the OLT, the observed ONU launch power variation range are approximately 3.6dB for ONU1 and 3.2dB for ONU2. The upper limit of the dynamic power variation range is determined by the minimum OSNR allowed by the fixed ONU, whilst the lower limit of the dynamic power variation range is determined by the minimum OSNR allowed by the varied ONU. This implies that an increase in extinction ratio of DML-modulated signals from both ONUs enhances the ONU launch power variation range.
3.3 Performance susceptibility to symbol offset between different upstream signals
In the OOFDMA PON system, to maintain the orthogonality among subcarriers of the combined upstream OOFDM signals for data recovery in the OLT, it is envisaged that all upstream signals from different ONUs should arrive at the OLT at the same time so that OFDM symbols associated with different ONUs are aligned with each other. However, in real network deployments, symbol offsets between various synchronized ONUs sharing a single OLT may still be encountered from time to time due to a number of unexpected practical factors such as component perturbations, laser wavelength drift and extreme environmental conditions. Therefore, from a system designer’s point of view, it is also of great importance to explore the performance sensitivity of the OOFDMA PON system to symbol offset between different upstream signals.
To examine the abovementioned issue, ONU BERs versus symbol offset between upstream signals of different ONUs are plotted in Fig. 7 , based on a fixed ONU launch powers of 6dBm and the DMLs at their optimum operating conditions. In measuring Fig. 7, perfect synchronisation between all the upstream signals is first achieved to ensure that the FFT window in the OLT locates at the optimum position for the entire system. Then various symbol offsets are introduced to the upstream signal of a particular ONU via adjusting the ‘STO compensation’ function block in its transmitter. For each sample offset introduced, BER measurements are conducted simultaneously for both ONUs.
It can be seen in Fig. 7 that the OOFDMA PON system is tolerant to any symbol offsets within a dynamic range of approximate 6 samples, which is slightly less than the adopted cyclic prefix length of 8 samples. Within the optimum symbol offset range, relatively flat BER curves of less than the FEC limit are obtainable for all ONUs. Whilst outside the optimum range, perfect subcarrier orthogonality in the OLT cannot be maintained, this causes the BERs of all the involved ONUs to increase sharply and simultaneously, as shown in Fig. 7. The duration of the optimum symbol offset range can be further prolonged when long cyclic prefix lengths are adopted in the OOFDM transceiver design .
3.4 Impact of adaptive DBA
The considered OOFDMA PON system utilizes the DMLs having significantly large differences in modulation characteristics, and the adopted real-time OOFDM transceivers exhibit system frequency response roll-offs as large as 12dB within the 2-GHz signal spectral region . Inevitably, these system/transceiver design aspects cause considerable variations in achievable SNRs experienced by individual subcarriers. Therefore, the BER performance of a specific ONU is expected to vary with the adopted adaptive DBA scheme due to alterations in assigning subcarriers of different frequencies to the ONU. Detailed explorations of such an effect and, more importantly, examinations of the feasibility of utilizing adaptive DBA to reduce the effect without affecting the signal bit rate offered to each ONU form the main objectives of this section.
Making use of the adaptive DBA scheme presented in Section 2.3, BERs versus received optical power for two ONUs involved are shown in Fig. 8(a) for two cases: Case I, represented by solid lines, where both ONUs are activated, and Case II, represented by dash lines, where one ONU is activated and the other ONU is deactivated. In the deactivated states, the DML driving currents are turned off and their optimum bias currents are still turned on. The supply of the bias currents in the deactivated states is to distinguish the power penalty caused by multiple access interference (MAI) between upstream signals, as discussed below.
It can be seen in Fig. 8(a) that, for each of the aforementioned two cases, almost identical BER developing curves are obtained. Considering the significant differences in DML modulation characteristics between different ONUs, and the 12dB system frequency response roll-offs, the identical BER developing curves indicate that the adopted adaptive DBA scheme is very effective in adaptively exploiting the available channel characteristics without affecting the BER performances of any ONUs. This statement is confirmed by Fig. 8(b), where even numbers of subcarriers, i.e., 2nd, 4th, ... 14th, are assigned to ONU1 and odd numbers of subcarriers, i.e., 1st, 3rd, … 15th, are allocated to ONU2, and the definition of two cases are identical to those in Fig. 8(a). Figure 8(b) shows that, although the subcarrier-interleaved DBA scheme is capable of providing each ONU with the same signal bit rate compared to that offered by the scheme used in Fig. 8(a), ONU1, however, suffers a 2dB increase in received optical power at the FEC limit BER, compared to ONU2.
In addition, Fig. 8(a) also shows that the minimum received optical powers required for achieving the FEC limit BER are −8.6dBm for Case I and −9.3dBm for Case II. The existence of a 0.7dB difference in the minimum received optical power between these two cases is mainly contributed by the MAI effect, which arises due to the imperfection in subcarrier orthogonalily. As such imperfection is free from the adopted adaptive DBA scheme, a very similar minimum received optical power difference between Case I and Case II is thus observed in Fig. 8(b) for each ONU.
It is well known that taking low signal modulation formats on OFDM subcarriers can considerably enhance the system BER performance tolerance to the effects of DML modulation nonlinearity, MAI and system frequency response roll off. From the discussions in Fig. 8, it is, therefore, expected that a reduction in signal modulation format decreases the dependence of system BER performance on variations in adaptive DBA schemes but at a price of lowering aggregated signal bit rates. Such an expectation is experimentally confirmed in Fig. 9 , where almost identical BER curves for all these cases are observed when 32-QAM is considered. In Fig. 9, all other system parameter values and the adaptive DBA scheme are identical to those adopted in Fig. 8(a). This gives an aggregated upstream signal bit rate of 9.375Gb/s, of which 4.375Gb/s and 5Gb/s are occupied by ONU1 and ONU2, respectively.
Based on modified real-time OOFDM transceiver architectures, end-to-end real-time experimental demonstrations have been reported, for the first time, of aggregated 11.25Gb/s over 26.4km SSMF, IMDD OOFDMA PONs with adaptive DBA. The demonstrated OOFDMA PON system consists of two ONUs, each of which utilizes a DFB-based DML or a VCSEL-based DML. Extensive experimental explorations of dynamic OOFDMA PON system properties have been undertaken in terms of several crucial system design aspects, which include impact of DML operating conditions, ONU launch power variation range, performance susceptibility to symbol offset between different upstream signals, as well as impact of adaptive DBA on the BER performance of each individual ONU. Optimum DML operating conditions have been identified. It has been shown that, for achieving acceptable BERs for all ONUs simultaneously transmitting upstream signals to the OLT, the OOFDMA PON system has a >3dB dynamic ONU launch power variation range, and the BER performance of the system is insusceptible to any upstream symbol offsets slightly smaller than the adopted cyclic prefix. In addition, experimental results have also indicated that, in addition to maximizing the aggregated system transmission capacity, adaptive DBA can also effectively reduce imperfections in transmission channel properties without affecting signal bit rates offered to individual ONUs.
This work was partly supported by the PIANO+ under the European Commission’s (EC’s) ERA-NET Plus scheme within the project OCEAN, the Welsh Assembly Government and The European Regional Development fund.
References and links
1. E. Wong, “Current and next-generation broadband access technologies”, Optical Fiber Communication/National Fiber Optic Engineers Conference (OFC/NFOEC), (USA, 2011), Paper NMD1.
2. L. G. Kazovsky, W.-T. Shaw, D. Gutierrez, N. Cheng, and S.-W. Wong, “Next-Generation Optical Access Networks,” J. Lightwave Technol. 25(11), 3428–3442 (2007). [CrossRef]
3. N. Suzuki, K. Nakura, T. Suehiro, M. Nogami, S. Kosaki, and J. Nakagawa, “Over-Sampling based Burst-mode CDR Technology for High-speed TDM-PON Systems”, Optical Fiber Communication/National Fiber Optic Engineers Conference (OFC/NFOEC), (USA, 2011), Paper OThT3.
4. J. Kani, “Enabling technologies for future scalable and flexible WDM-POJN and WDM/TDM-PON systems,” IEEE J. Sel. Top. Quantum Electron. 16(5), 1290–1297 (2010). [CrossRef]
5. D. Qian, J. Hu, J. Yu, P. N. Ji, L. Xu, T. Wang, M. Cvijetic, and T. Kusano, “Experimental demonstration of a novel OFDM-A based 10 Gb/s PON architecture,” European Conference on Optical Communication (ECOC), (Berlin, 2007), Paper Mo 5.4.1.
6. D. Qian, N. Cvijetic, J. Hu, and T. Wang, “A novel OFDMA-PON architecture with source-free ONUs for next-generation optical access networks,” IEEE Photon. Technol. Lett. 21(17), 1265–1267 (2009). [CrossRef]
7. Y.-M. Lin and P.-L. Tien, “Next-generation OFDMA-based passive optical network architecture supporting radio-over-fiber,” IEEE J. Sel. Areas Commun. 28(6), 791–799 (2010). [CrossRef]
8. X. Q. Jin, J. L. Wei, R. P. Giddings, T. Quinlan, S. Walker, and J. M. Tang, “Experimental demonstrations and extensive comparisons of end-to-end real-time optical OFDM transceivers with adaptive bit and/or power loading,” IEEE Photonics J. 3(3), 500–511 (2011). [CrossRef]
9. N. Cvijetic, D. Qian, and J. Hu, “100 Gb/s optical access based on optical orthogonal frequency-division multiplexing,” IEEE Commun. Mag. 48(7), 70–77 (2010). [CrossRef]
10. R. P. Giddings and J. M. Tang, “Experimental demonstration and optimisation of a synchronous clock recovery technique for real-time end-to-end optical OFDM transmission at 11.25Gb/s over 25km SSMF,” Opt. Express 19(3), 2831–2845 (2011). [CrossRef] [PubMed]
11. R. P. Giddings, X. Q. Jin, E. Hugues-Salas, E. Giacoumidis, J. L. Wei, and J. M. Tang, “Experimental demonstration of a record high 11.25Gb/s real-time optical OFDM transceiver supporting 25km SMF end-to-end transmission in simple IMDD systems,” Opt. Express 18(6), 5541–5555 (2010). [CrossRef] [PubMed]
12. E. Hugues-Salas, R. P. Giddings, X. Q. Jin, J. L. Wei, X. Zheng, Y. Hong, C. Shu, and J. M. Tang, “Real-time experimental demonstration of low-cost VCSEL intensity-modulated 11.25 Gb/s optical OFDM signal transmission over 25 km PON systems,” Opt. Express 19(4), 2979–2988 (2011). [CrossRef] [PubMed]
13. X. Q. Jin and J. M. Tang, “Optical OFDM synchronization with symbol timing offset and sampling clock offset compensation in real-time IMDD systems,” IEEE Photonics J. 3(2), 187–196 (2011). [CrossRef]
14. X. Q. Jin, R. P. Giddings, and J. M. Tang, “Real-time transmission of 3 Gb/s 16-QAM encoded optical OFDM signals over 75 km SMFs with negative power penalties,” Opt. Express 17(17), 14574–14585 (2009). [CrossRef] [PubMed]
15. M. Morelli, C.-C. J. Kuo, and M.-O. Pun, “Synchronization Techniques for Orthogonal Frequency Division Multiple Access (OFDMA): A Tutorial Review,” Proc. IEEE 95(7), 1394–1427 (2007). [CrossRef]
16. M. Nölle, L. Molle, D.-D. Gross, and R. Freund, “Transmission of 5x62 Gbit/s DWDM coherent OFDM with a spectral efficiency of 7.2 bit/s/Hz using joint 64-QAM and 16-QAM modulation”, Optical Fiber Communication/National Fiber Optic Engineers Conference (OFC/NFOEC), (USA, 2010), Paper OMR4.
17. J. L. Wei, C. Sánchez, R. P. Giddings, E. Hugues-Salas, and J. M. Tang, “Significant improvements in optical power budgets of real-time optical OFDM PON systems,” Opt. Express 18(20), 20732–20745 (2010). [CrossRef] [PubMed]
18. X. Zheng, X. Q. Jin, R. P. Giddings, J. L. Wei, E. Hugues-Salas, Y. H. Hong, and J. M. Tang, “Negative power penalties of optical OFDM signal transmissions in directly modulated DFB laser-based IMDD systems incorporating negative dispersion fibres,” IEEE Photonics J. 2(4), 532–542 (2010). [CrossRef]