The Full Services Access Network group has recently selected the time and wavelength division multiplexed passive optical network (TWDM-PON) as the base technology solution for next-generation PON stage-2 (NG-PON2). Meeting the core requirements of NG-PON2 necessitates the following additional features in the transceivers of the optical network unit (ONU) that is located at subscriber premises: (a) legacy system compliant; (b) wavelength tunable; (c) cost-efficient; and (d) energy-efficient. To address these features, we investigate the properties of short-cavity vertical-cavity surface-emitting lasers (SC-VCSELs) for implementation as colorless ONU transmitters in future TWDM-PONs. Specifically, we investigate the tunability and transmission performance of the SC-VCSEL across the C-minus wavelength band for legacy system compliance. We report on error-free transmission across a 800 GHz tuning range with a potential aggregate upstream capacity of 80 Gbps over a system reach of 40 km and with a split ratio of 1:128 per wavelength channel. Results were achieved without dispersion compensation and electronic equalization. We also evaluate the energy efficiency of the SC-VCSEL in active, doze, and sleep mode. When in active mode, the SC-VCSEL transmitter block consumes 91.7% less power than a distributed feedback (DFB) laser transmitter block. When transitioning between doze and active modes, the transmitter block has a short settling time of only 205 ns, thus increasing the power-saving duration and consequently reducing the overall power consumption of the ONU. Through numerical analysis, evaluation of the energy-savings of the SC-VCSEL ONU over the DFB ONU under various modes of operation, demonstrates up to 84% of energy-savings. The capacity, tuning range, split ratio, system reach, and energy-savings arising from SC-VCSEL ONU implementation as reported in this work, exceed the minimum requirements of NG-PON2 for future TWDM-PON deployments.
© 2013 OSA
The combination of an exponential increase in bandwidth-intensive applications and customer base, has resulted in the proliferation of fiber networks in the access network segment in recent years. In terms of fiber access technology, the point-to-multipoint passive topology in the form of the passive optical network (PON) has been proven to be beneficial to both customers and operators. In future-proofing the optical access segment, the Full Services Access Network (FSAN) group has selected the time and wavelength division multiplexed PON (TWDM-PON) as the base technology solution to meet the core requirements of next-generation PON stage-2 (NG-PON2) [1,2]. To optimize investment costs whilst providing a graceful upgrade path, TWDM-PON deployments are recommended to possess the following criteria: (a) coexistence with legacy PON technologies for flexible bandwidth upgradeability and management, (b) reuse of existing optical distribution network (ODN) for cost-savings, (c) provide a minimum 40 Gbps aggregate downstream capacity through 4 × 10 Gbps downstream wavelength channels, (d) provide a minimum 10 Gbps aggregate upstream capacity through 4 × 2.5 Gbps wavelength channels (e) provide a minimum 1:64 split ratio per wavelength channel, and (f) achieve a minimum system reach of 40 km [1,2].
Figure 1 illustrates the baseline TWDM-PON architecture with (a) an optical line terminal (OLT) comprising N transceivers, (b) an ODN that is completely passive, and (c) multiple optical network units (ONUs) that are each capable of transmitting on any one of the N upstream wavelengths and receiving on any one of the N downstream wavelengths. Such an architecture effectively facilitates N TDM-PONs which are concurrently shared by the same ODN and ONUs. Tunable transmitters and tunable receivers at the ONUs provide cost-effective inventory management through colorless operation and also support of multiple service providers through remote reconfigurability . Potential candidates for tunable transmitter at the ONUs include tunable distributed feedback (DFB) lasers  and tunable vertical cavity surface emitting lasers (VCSELs) . Though potentially cost prohibitive, the use of tunable lasers avoids the need for centralized broadband light source(s) as compared to other solutions, and subsequently the Rayleigh backscattering penalty from using these broadband source(s).
Further, unlike legacy TDM-PONs, optical amplifier(s) can be deployed at the OLT to optically amplify both upstream and downstream signals to compensate for high optical losses from increased system reach and split ratios . It is worth noting that the ODN retains its passive nature since all active equipment of the network is located centrally at the OLT. A TWDM-PON with 4 pairs of upstream (1535 nm to 1540 nm) and downstream wavelengths (1554 nm to 1559 nm) residing within the C-band, was reported in . The employed wavelength plan facilitated the co-existence with legacy PONs, i.e. GPON, GE-PON, XG-PON, and 10GE-PON. An aggregate 40 Gbps downstream and 10 Gbps upstream capacity over a 20 km distance with a 1:512 total split ratio was achieved. The ONU transmitter used was a thermally-tuned directly-modulated distributed feedback (DFB) laser with a 400 GHz wavelength tuning range.
In addition to the NG-PON2 requirements discussed previously, power-savings in future access networks is also important to both network operator and customer. To reduce energy consumption in access networks, optical transceivers and electronic circuits of low power consumption are being developed. Sleep and doze mode operations in both the central office (CO) through OLT power shutdown, and in the subscriber premises through ONU power shutdown have been proposed [6,7]. In terms of the ONU, both the User Network Interface (UNI) through power shedding operation, and the PON interface through dozing and sleeping, can be powered down. When dozing, an ONU powers down its transmitter (TX) block and when sleeping, the ONU powers down both its TX and receiver (RX) blocks. In terms of energy-efficiency, a tunable DFB transmitter based ONU which depends on temperature tuning and requiring thermo-electric control, may not be the most suitable candidate to achieve low power consumption at the ONU, as will be discussed in Section 2. Further, to satisfy the aggregate downstream and upstream capacities, split ratio, and system reach of TWDM-PON, the implementation of higher order modulation formats and equalization techniques may be required .
Colorless ONU transmitters based on external seeding source(s) such as the reflective semiconductor optical amplifier (RSOA)  and Fabry-Perot laser diode (FP-LD)  may too be inappropriate for TWDM-PON operation due to the need for high-power external seeding source(s) to compensate for the increased transmission distance and high split ratio. Further, self-seeding RSOAs  and FP-LD  used in WDM-PONs may not be feasible in this case. The self-seeding technique is based on reflecting narrowband filtered lights from the RSOA and FP-LD back to the devices to self-seed upstream channels. As the wavelength filtering component is implemented in the OLT and not in the ODN, the high split ratio and long transmission distances will result in weak self-seeded lights resulting in unstable self-seeding performance . Further, the self-seeding of multiple ONUs with the same upstream wavelength channel as per the TWDM-PON architecture shown in Fig. 1, may cause instability in the self-seeding performance.
In this work, we address the requirements of NG-PON2 through investigating the colorless operation and transmission performance of short-cavity vertical-cavity surface-emitting lasers (SC-VCSELs) for ONU transmitter implementation in future TWDM-PONs. With low bias and modulation currents, and the potential for cost-efficient mass-fabrication, these devices have the potential to provide high performance in speed, cost-efficiency, and energy-efficiency . Through experiments, we show that the SC-VCSEL has a 800 GHz tuning range in the C-minus band, potentially providing eight 100 GHz spaced 10 Gbps upstream channels, and thus an aggregate upstream capacity of 80 Gbps. Transmission within the C-minus wavelength band extends the aggregate upstream capacity whilst still meeting system coexistence requirements. Through experiments, we will report on error-free performances without using dispersion compensation and electronic equalization for 40 km and with up to a 1:128 split ratio per wavelength channel.
Further, we show for the first time that the TX block using the SC-VCSEL transmitter can achieve a very short settling time of 205 ns between active and dozing intervals, thus allowing the ONU to transition into doze mode even during very short polling cycles and high network loading levels. Also, when in active mode, the SC-VCSEL TX block consumes 91.7% less power than that of an ONU with a DFB laser. We numerically analyze the power-saving effectiveness of the SC-VCSEL ONU over the DFB ONU under various modes of operation, demonstrating an energy-savings of up to 84% when in sleep mode. The demonstrated capacity, tuning range, split ratio, system reach, and energy-savings arising from SC-VCSEL ONU implementation as shown in this work, exceed the minimum requirements of NG-PON2 for future TWDM-PON deployments.
2. Characterization of high-speed short-cavity VCSEL
The transmitter (TX) block of the SC-VCSEL ONU used in our experiments is shown in Fig. 2(a), and comprises a 10 Gbps SC-VCSEL with a buried tunnel junction structure and a 11.3 Gbps VCSEL driver. The SC-VCSEL is mounted on a TO-46 header, which is then packaged in a transmitter optical sub-assembly (TOSA) with an electrical flex-line connector. The threshold current was measured to be 0.9 mA. The temperature tuning coefficient of the SC-VCSEL is 0.09nm/°C across the entire temperature range from −10°C to 90°C , which is comparable to that of a DFB laser at 0.1 nm/°C.
When compared to the long-cavity VCSEL (LC-VCSEL) which was previously proposed for upstream operation in the O-band of TDM-PONs , SC-VCSELs have the potential to be more energy-efficient under direct modulation. Benefitting from reduced photon lifetimes due to a reduced cavity length, SC-VCSELs feature increased D-factors and modulation current efficiencies (MCEFs), as listed in Table 1. Observe that for SC- and LC-VCSELs with the same 3 dB bandwidth of 10 GHz, the SC-VCSEL consumes 1.7 mW of electrical power as compared to 9.5 mW of the LC-VCSEL. Also, the SC-VCSEL can achieve a certain modulation bandwidth at a comparably smaller current than for conventional LC devices. Consequently, the overall power consumption of a SC-VCSEL is lower and smaller energy-per-bit values are achieved. The possibility of operating SC-VCSELs at lower DC-currents also reduces current-induced device heating.
For colorless operation in a TWDM-PON, the same SC-VCSEL TX block can be used in multiple ONUs in which the SC-VCSEL is voltage-biased tuned across the C-minus band. In our experiments, the bias current of the SC-VCSEL is controlled by an external voltage that is applied to the laser driver. This voltage is referred to as the bias voltage. Changing this bias voltage therefore changes the bias current and hence the output wavelength of the SC-VCSEL. Specifically, the step-wise tuning of the bias voltage between 0.48V to 0.85 V yields eight 100 GHz-spaced channels corresponding to λ1 (1522.4 nm) to λ8 (1527.92 nm). At CW, the output launch power of λ1 to λ8 was measured to be between 0.65 dBm and −1.65 dBm, and the extinction ratio between 6.23 dB to 8.9 dB. Figure 2(b) shows the optical spectra of λ1 to λ8 at CW operation. In our experiments, passive temperature stabilization is achieved through effective integrated heat sinking of the SC-VCSEL, thereby removing the need for active temperature control and improving the overall power-efficiency of the TX block.
In general, VCSELs provide both cost-efficiency as well as power-efficiency at the recommended optical power levels for upstream transmission in the access network. To explain, Fig. 2(c) which compares the optical power (solid red) and voltage drop (dash red) of the SC-VCSEL as a function of the bias current to those of the DFB laser (solid and dash black lines, respectively). For the SC-VCSEL, operating at a bias current of 10 mA yields an output power of 5 mW. In contrast, the threshold current of a typical DFB laser at room-temperature is ~10 mA, and therefore would require a higher bias current to achieve the same output power.
In terms of energy-efficiency, out of the two power saving modes of the ONU, sleep mode is considered to be the more energy-efficient mode due to the powering down of both TX and RX blocks during the sleeping interval, but an overhead Trec of 2 ms is incurred for clock recovery and resynchronization to the CO clock after transitioning out of sleeping . Under heavy traffic conditions, a long Trec prevents an ONU from transitioning into sleep mode. In comparison, doze mode provides lower energy-savings through the powering down of only the RX block but a smaller overhead of typically 760 ns  which corresponds to the settling time Tsett of the TX block is incurred. In order to maximize the energy-efficiency of an ONU, (a) its power consumption when active, dozing, and sleeping should be minimized, and (b) the dozing and sleeping intervals should be maximized . Since both Trec and Tsett reduce the duration in which the ONU can sleep or doze, these values must be minimized.
A summary of the power consumption of the TX and RX blocks of the SC-VCSEL ONU and DFB ONU when active, dozing, and sleeping, is listed in Table 2. Note that these values do not include the power consumption of the User Network Interface, but only the PON interface. The parameters for the previously proposed LC-VCSEL ONU are also listed for comparison. Note that the RX blocks of the ONUs were chosen to be identical to highlight the energy savings arising from using VCSEL based transmitters. As listed in Table 2, the 10 Gbps SC-VCSEL TX block when powered up in active mode, consumes an average of 0.1 W  as compared to 1.202 W of DFB [19,20] and 0.134 W  of the LC-VCSEL transmitter blocks. Though using comparable laser drivers, the DFB TX module results in a higher power consumption value due to the use of higher modulation voltage, higher bias voltage, higher supply current, and embedded temperature control within the laser package which is necessary wavelength channel tuning. With the power consumption of the RX block included, the SC-VCSEL ONU consumes a total of 3.95 W [21,22] whereas a LC-VCSEL ONU and DFB ONU consumes 3.984 W [21,22] and 5.052 W [19,20,22], respectively.
The Tsett values of the three types of ONUs are also listed in Table 2. The Tsett value of 205 ns of the SC-VCSEL was experimentally measured by measuring the individual transition times of the SC-VCSEL TX block between active and doze modes and vice versa. Figure 3 shows the oscilloscope traces of the doze mode control and the resulting TX block output. A ‘0’ to ‘1’ transition of the doze mode control powers down the TX block and vice versa. In general, VCSELs have shorter settling times than DFB lasers due to the following reasons. The VCSEL has a differential series resistance that is in the range of 50 Ohms. Consequently, as compared to the DFB laser, the VCSEL is inherently well-matched to 50 Ohm circuits and its connection to well-designed, commercially available 50 Ohm driving circuits (such as that used in our work) results in very short settling times. The entire applied modulation bias drops at the 50 Ohm-VCSEL and is not dissipated in matching circuits which would otherwise be necessary for DFB lasers with low differential series resistance around 10 Ohm.
3. Transmission experiments and results
To investigate its colorless operation, the TX block in our experiments, as described in Section 2, was voltage-biased tuned between 0.48V to 0.85 V to yield eight 100 GHz-spaced channels corresponding to λ1 (1522.4 nm) to λ8 (1527.92 nm) in the C-minus band. In evaluating the transmission performance of the VCSEL at these wavelengths, Fig. 4(a) shows an illustrative example of the TWDM-PON implementing such a SC-VCSEL at the ONUs. In our experiments, the SC-VCSEL was directly modulated using the VCSEL driver with 10 Gbps NRZ data (PRBS length of 215-1) from a pulse pattern generator (Anritsu MP1763B). Figure 4(b) shows the modulated optical spectra of λ1 to λ8 measured after the 3 dB coupler at the output of the ONU in the upstream direction. Under direct modulation, the TX block consumes 0.095W for λ1 through to 0.11 W for λ8 . These values are approximately 10% of the 1.202 W consumed by a comparable 10 Gbps DFB TX block [19,20]. The modulated upstream signals were then launched into the ODN, comprising an optical attenuator that emulated the function of the optical splitter, and standard single mode fiber of lengths 40 km or 60 km. The optical attenuation was set to correspond to the optical splitter loss of the network. No dispersion-compensating fiber or electronic equalization was used in our setup.
Prior to reception at the OLT, the upstream signals were amplified by an optical amplifier, the C-band EDFA (Highwave EDFA), with a measured optical gain from 20 dB to 28 dB for λ1 to λ8, respectively. Figure 4(c) shows the superimposed plots of the measured ASE noise spectrum and optical gain spectrum. The combination of different optical launch powers, non-flat EDFA gain and ASE noise within the C-minus band, results in upstream channels with differing optical signal-to-noise ratio (OSNR) when measured at the OLT. The amplified upstream signals was then fed to a tunable optical bandpass filter (OBPF) with a 30 GHz bandwidth and then to a second optical attenuator for upstream bit-error-ratio (BER) performance evaluation. The OBPF based on a Fabry-Perot etalon emulates the demultiplexer, removes out-of-band ASE noise, and provides optical filtering of the chirped upstream signals. Unlike external modulation, directly modulation results in chirp-induced optical spectrum broadening that is asymmetrical. As such, off-center optical filtering was used in our experiments to increase the extinction ratio of optical signals through suppressing the low frequency components that correspond to the “0”s in data sequences, and hence the zero level [23,24]. This technique has been widely implemented in long-haul transmission for dispersion management, and has recently been investigated for use in the access segment .
Figure 4(d) and 4(e) shows the optical spectra of the amplified upstream signals after 40 km and 60 km SSMF transmission, respectively. Each optical spectra, measured at the output of the OBPF when tuned from λ1 to λ8, displays a non-uniform peak power across the wavelength channels. Note that this variation in peak power when measured at the OLT, is caused by the combination of varying SC-VCSEL launch power and non-flat EDFA gain across the wavelength channels. Finally, a 10 GHz photoreceiver comprising an avalanche photodiode and a limiting amplifier is used for upstream signal detection. The detected signals are then fed to an error detector (Anritsu MP1764A) for upstream BER measurements.
Figures 5(a) and 5(b) plot the BER of the upstream signals on λ1 to λ8 for 1:64 and 1:128 split ratio TWDM-PON configurations, respectively. The length of the ODN was 40 km in both cases. Observe that the BER performance across all wavelengths degrades with increasing split ratio, due to reduced OSNR at the receiver. As an example, the OSNR for wavelength channel λ4 of the 1:64 and 1:128 split ratio configurations were measured to be 33.96 dB and 30.70 dB, respectively. Further, when considering either the 1:64 or 1:128 configurations separately, observe that λ1 and λ2 have lower BER performances as a result from lower OSNR due to lower optical launch power and EDFA gain. As a comparison to λ4 with OSNR of 33.96 dB and 30.70 dB for the 1:64 and 1:128 configurations, the OSNR of λ1 was measured to be 31.86 dB and 28.80 dB, respectively.
Figure 6(a) compares the 10 Gbps upstream BER performance of four different combinations of ODN: (a) two different system reach of 40 km and 60 km; and (b) two different split ratios of 1:64 and 1:128. The corresponding electrical eye diagrams measured at a BER of 10−9 using Agilent DCA 86100A, are shown on the right hand side. The upstream channel λ4 was selected as the reference channel for these measurements. Results show an error floor for the case of 60 km may be due to signal degradation from the combination of chromatic dispersion, receiver noise, and low OSNR. At a BER of 10−9, the power penalty for the case of the 60 km with 1:64 and 1:128 split ratios when compared to the back-to-back (B2B) measurements are 2.5 dB and 4.5 dB respectively. To address the minimum requirements of NG-PON2, the experiments were then repeated at an upstream transmission bit-rate of 2.5 Gbps. Figure 6(b) shows the upstream BER performance for the four combinations of ODN. The corresponding eye diagrams measured at a BER of 10−9 are shown on the right. As expected, a SC-VCSEL directly modulated at 2.5 Gbps shows improved BER performances over the transmission performance at 10 Gbps.
Finally, the performance of the SC-VCSEL in transmitting data beyond 10 Gbps for future proofing the TWDM-PON, was also investigated. Results in Fig. 6(c) show that increasing the transmission bit-rate from 2.5 Gbps to 12.5 Gbps results in power penalties of 8.5 dB and 10 dB for 1:64 and 1:128 split ratios, respectively. This degradation in transmission performance is expected, being a result of modulation bandwidth limitation at the ONU whereby the VCSEL driver limited to 11.3 Gbps modulation, and at the OLT whereby the 10 GHz photoreceiver is used. Regardless, error-free upstream performance (BER ≤ 10−9) could be achieved for all configurations considered. In practice, minimal variation in the output launch powers, optical gain, and ASE noise profile within the C-minus band would be beneficial to reducing the performance deviation between the eight wavelength channels considered. Further, with the implementation of forward error correction (FEC) in which the error-free BER threshold is increased to 4 × 10−3 , receiver sensitivity can be further improved and the deviation between the four ODN combinations considered can be reduced. For example, considering the upstream bit-rate of 2.5 Gbps, 40 km, and 1:64 split per wavelength to be the baseline requirement, an upstream power budget of 33 dB can be achieved with FEC.
4. Energy-savings of SC-VCSEL ONUs in doze and sleep mode operations
In this section, we analyze the energy-efficiency of SC-VCSEL ONUs by evaluating its percentage of energy-savings in (a) active, doze, and sleep mode operation over a DFB ONU in active mode, (b) doze and sleep mode operation over a DFB-ONU in doze mode, and finally in (c) sleep mode over a DFB ONU in sleep mode. In our analysis, we consider a TWDM-PON with a 1:64 split per wavelength channel, and an upstream transmission bit-rate of 10 Gbps per channel. In order for all ONUs to efficiently share bandwidth on a single upstream wavelength, we consider a dynamic bandwidth allocation (DBA) access scheme whereby each ONU is dynamically allocated bandwidth by the OLT during each polling cycle. Controlled messages such as the REPORT to request bandwidth by an ONU, and GATE to dynamically allocate bandwidth by the OLT, are used for this purpose. The REPORT control message with a duration of TREPORT, is a fixed 64 byte overhead that is transmitted by the ONU once per polling cycle. The polling cycle TCYC is defined as the time interval between consecutive transmissions from the same ONU, and is an important network parameter that impacts the packet delay, jitter, and bandwidth utilization . For DBA upstream access, TCYC is proportional to the total amount of demanded upstream bandwidth from all ONUs or equivalently the total network load. Therefore, depending on the requested bandwidth per ONU, TCYC differs from cycle to cycle. Some DBA access schemes specify a maximum polling cycle time TCYCMAX in order to limit the allocated bandwidth per polling cycle to a maximum value, and to limit the transmission delay to a maximum value.
The network and protocol parameter values used in our analysis are listed in Table 3. The arrival rate of the Ethernet packets for upstream transmission at each ONU follows the Poisson distribution with exponential arrival times. The Ethernet packet length is chosen to be uniformly distributed between 64 and 1518 bytes, and with an average packet length of 791 bytes. The guard time between consecutive ONU transmissions is 1 µs . In our analysis, we vary TCYCMAX and normalized network loading L to investigate their impact on the energy-savings arising from using SC-VCSEL ONUs.
A. Comparison of SC-VCSEL ONU in active, doze, and sleep modes vs. active DFB-ONU
When comparing an active SC-VCSEL ONU to an active DFB ONU, the percentage of energy-savings η arising from implementing the former is given by (1) below. The parameters Pactive,SC (3.95W) and Pactive,DFB (5.052 W) are the power consumed by the SC-VCSEL ONU and DFB ONU, respectively when both their TX and RX blocks are powered up. As can be deduced from (1), energy-savings η is a constant 22% irrespective of L and TCYCMAX. Even without transitioning into dozing and/or sleeping, the SC-VCSEL ONU is able to provide a constant 22% of energy-savings over the DFB ONU. This result is reflected in Fig. 7(a).
The general formula for evaluating the percentage of energy-savings η of the SC-VCSEL ONU in either doze or sleep mode over an active DFB ONU is given by:Fig. 7(b) and Fig. 7(c), respectively. Under doze mode operation, the SC-VCSEL ONU yields minimum η when the network load and TCYCMAX is minimum. Likewise, η is maximum at maximum network load and TCYCMAX. Results obtained for sleep mode operation shown in Fig. 7(c) indicate a similar energy-saving behavior to that of the doze mode. However, no energy savings is observed for TCYCMAX < Trec (~2 ms) where the ONU does not have sufficient time to transition between active and sleeping. Beyond that, the energy savings of up to 84% can be achieved when network load and TCYCMAX are maximum. This energy-saving behavior can be explained as follows. For DBA upstream access, the polling cycle is dynamic and proportional to the network load. Therefore, a heavily-loaded network and one with a long TCYCMAX will maximize the dozing or sleeping interval within the polling cycle and hence the energy-savings, η.
B. Comparison of doze and sleep mode SC-VCSEL ONU vs doze mode DFB-ONU
In (5)-(7), the parameter Tactive,DFB is the time interval within a polling cycle when the TX and RX blocks of the DFB ONU are powered up, and Tdoze,DFB is the time interval within a polling cycle when the DFB ONU is dozing. The duration Tdoze,DFB is reduced by the overheads Tsett,DFB (760 ns) and TREPORT during each polling cycle. The parameter Pdoze,DFB is the power consumed by the DFB ONU during the dozing (3.85W). Results in Fig. 7(d) show energy savings that is contrary to that of Fig. 7(b). That is, higher energy-savings is observed when the network load and TCYCMAX is low. Another observation point to the fact that even though both types of ONUs consume the same amount of power of 3.85 W when dozing, results show that the doze mode SC-VCSEL ONU is still more energy-efficient than the DFB ONU. The reasons are as follows: (a) the Tsett of 205 ns of the SC-VCSEL ONU is shorter than that of the DFB ONUs, thus giving rise to a longer dozing interval for the same polling cycle time, and (b) when transitioning out of dozing into the active interval, the power consumption of the SC-VCSEL ONU when active is lower than the DFB ONU. When comparing a sleep mode SC-VCSEL ONU to a doze mode DFB ONU, results shown in Fig. 7(e) exhibit a similar behavior to that of Fig. 7(c). For TCYCMAX < Trec (~2 ms), energy savings of up to 80% can be achieved when network load and TCYCMAX are maximum. Another point to note is that the maximum achievable energy-savings in Fig. 7(e) is slightly less than that obtained in Fig. 7(c) as the difference in power consumption between a sleep mode SC-VCSEL ONU and a doze mode DFB ONU is less than that of a sleep mode SC-VCSEL ONU and an active DFB ONU.
C. Comparison of SC-VCSEL ONU in sleep mode against DFB-ONU in sleep-mode
The general formula for evaluating the percentage of energy-savings η of a sleep mode SC-VCSEL ONU over a sleep mode DFB ONU is given by (8)-(10) below.
For the same network loading level and maximum cycle time, the SC-VCSEL and DFB-ONU in sleep mode have identical active time intervals i.e. Tactive,SC = Tactive,DFB, sleep intervals i.e. Tsleep,SC = Tsleep,DFB, power consumption i.e. Psleep,DFB = Psleep,SC = 0.75 W, and recovery time i.e. Trec,DFB = Trec,SC = Trec = 2ms. However, due to the fact Pactive,SC < Pactive,DFB, an appreciable amount of energy-savings reaching a maximum of 21% can still be achieved. This can be observed in Fig. 7(f) for a range of network loading levels and for TCYCMAX > Trec (~2 ms). Our results in this section of the paper confirm that the SC-VCSEL ONU is a suitable candidate to provide energy-savings in TWDM-PONs.
In this work, we addressed the efforts in facilitating a tunable, cost-efficient, power-efficient, and legacy system compliant transmitter for implementation in future TWDM-PONs. Through experiments, we demonstrated upstream error-free transmission performances of the SC-VCSEL ONU across a tuning range of 800 GHz in the C-minus band. Results indicate an aggregate capacity, reach, and split ratio, which exceed the minimum requirements of NG-PON2. We reviewed the properties of the SC-VCSEL which enable: (a) low power consumption, and (b) a fast settling time Tsett of 205 ns between dozing and active intervals. These features allow its power consumption to be minimized during both active and dozing intervals, and its dozing interval to be maximized. We also investigated the energy-savings from the implementation of the SC-VCSEL when compared to the DFB laser. Through numerical analysis, we showed that the low power consumption and short settling time features of the SC-VCSEL allows considerable energy-savings of up to 84% as compared to the DFB laser. In summary, the demonstrated colorless SC-VCSEL with a 800 GHz tuning range in the C-minus band is highly-suited for deployment as the upstream transmitter of future TWDM-PONs. In addition, the low-cost and low power consumption features of this device are significantly beneficial to the cost-sensitive environment of these networks.
The authors would like to thank VERTILAS GmbH for providing the VCSELs used in our work, and also the Australian Research Council and Go8-DAAD Germany Joint Research Co-Operation Scheme for funding the research.
References and links
1. H. Nakamura, “NG-PON2 Technology,” Proc. of IEEE/OSA Opt. Fiber Commun. Conf., Anaheim, USA, NTh4F.5, (2013).
2. Y. Luo, X. Zhou, F. Effenberger, X. Yan, G. Peng, Y. Qian, and Y. Ma, “Time- and Wavelength-Division Multiplexed Passive Optical Network (TWDM-PON) for Next-Generation PON Stage 2 (NG-PON2),” J. Lightwave Technol. 31(4), 587–593 (2013). [CrossRef]
3. R. A. Kumar, R. Murano, M. Cahill, J. Aufiero, and G. Bartolini, “Crosstalk performance analysis of low cost 8 X 10 Gb/s tunable receiver for TWDM-PON,” Proc. of IEEE/OSA Opt. Fiber Commun. Conf., Anaheim, USA, NTh3F.5 (2013).
4. H. Suzuki, M. Fujiwara, T. Suzuki, N. Yoshimoto, H. Kimura, and M. Tsubokawa, “Wavelength-tunable DWDM-SFP transceiver with a signal monitoring interface and its application to coexistence-type colorless WDM-PON,” in Proc. Eur. Conf. Opt. Commun., PD3.4 (2007).
5. C. Chang-Hasnain, “Optically-injection locked tunable multimode VCSEL for WDM passive optical networks,” in Proc. Int. Nano-Optoelectron. Workshop (i-NOW), 98–99 (2008). [CrossRef]
6. H. Mukai, T. Fumihiko, and J. Nakagawa, “Energy-efficient 10G-EPON system,” Proc. of IEEE/OSA Opt. Fiber Commun. Conf., Anaheim, USA, OW3G.1 (2013).
7. E. Wong, M. Mueller, M. P. I. Dias, C. A. Chan, and M. C. Amann, “Energy-efficiency of optical network units with vertical-cavity surface-emitting lasers,” Opt. Express 20(14), 14960–14970 (2012). [CrossRef] [PubMed]
8. Q. Guo, A. V. Tran, and C. J. Chae, “Extended-reach 10 Gb/s RSOA-based WDM-PON using partial response equalization, ” Proc. of 23rd Annual Meeting of the IEEE Photonics Society,345 – 346(2010). [CrossRef]
9. Y. Katagiri, K. Suzuki, and K. Aida, “Intensity stabilisation of spectrum-sliced Gaussian radiation based on amplitude squeezing using semiconductor optical amplifiers with gain saturation,” Electron. Lett. 35(16), 1362–1364 (1999). [CrossRef]
10. H. D. Kim, S.-G. Kang, and C.-H. Lee, “A low-cost WDM source with an ASE injected Fabry–Perot semiconductor laser,” IEEE Photon. Technol. Lett. 12(8), 1067–1069 (2000). [CrossRef]
11. E. Wong, K. L. Lee, and T. A. Anderson, “Directly modulated self-seeding reflective semiconductor optical amplifiers as colorless transmitters in wavelength division multiplexed passive optical networks,” J. Lightwave Technol. 25(1), 67–74 (2007). [CrossRef]
12. S. Hann, T.-Y. Kim, and C.-S. Park, “Direct-modulated upstream signal transmission using a self-injection locked F-P LD for WDM-PON,” Proc. European Conf. of Communications (ECOC), We3.3.3 (2005) [CrossRef]
13. M. Mueller, W. Hofmann, T. Grundl, M. Horn, P. Wolf, R. D. Nagel, E. Ronneberg, G. Bohm, D. Bimberg, and M. C. Amann, “1550nm high-speed short-cavity VCSELs,” IEEE J. Sel. Top. Quantum Electron. 99, 1158–1166 (2011).
14. M.-C. Amann, E. Wong, and M. Mueller, “Energy-efficient high-speed short-cavity VCSELs,” Proc. of IEEE/OSA Opt. Fiber Commun. Conf., Los Angeles, USA, OTh4F.1 (2012). [CrossRef]
15. W. Hofmann, M. Muller, G. Bohm, M. Ortsiefer, and M.-C. Amann, “1.55um VCSEL with enhanced modulation bandwidth and temperature range,” IEEE Photon. Technol. Lett. 21(13), 923–925 (2009). [CrossRef]
16. M. Mueller and M. C. Amann, “State-of-the-art and perspectives for long-wavelength high-speed VCSELs,” in Proceedings of Conference on Transparent Optical Networks (ICTON) 2011, Mo.C5.2 (2011).
17. J. Mandin, “EPON powersaving via sleep mode,” IEEE P802.3av 10GEPON Task Force Meeting (2008).
18. E. Igawa, M. Nogami, and J. Nakagawa, “Symmetric 10G-EPON ONU BMT employing dynamic power save control circuit,” Proc. of IEEE/OSA Opt. Fiber Commun. Conf., Los Angeles, USA, NTuD5 (2011).
19. Analog Devices ADN2531 data sheet, [Online]. Available: http://www.analog.com/static/imported-files/data_sheets/ADN2531.pdf”.
20. NEL Laser Diodes, [Online]. Available: http://www.nttelectronics.com/en/products/photonics/pdf/NLK5C5EBKA.pdf”.
21. Analog Devices ADN2530 data sheet, [Online]. Available: http://www.analog.com/static/imported-files/data_sheets/ADN2530.pdf”.
22. S.-W. Wong, L. Valcarenghi, S.-H. Yen, D. R. Campelo, S. Yamashita, and L. Kazovsky, “Sleep mode for energy saving PONs: Advantages and drawbacks,” Proc. of IEEE GLOBECOM Workshops, (2009). [CrossRef]
23. L. S. Yan, Y. Wang, B. Zhang, C. Yu, J. McGeehan, L. Paraschis, and A. E. Willner, “Reach extension in 10-Gb/s directly modulated transmission systems using asymmetric and narrowband optical filtering,” Opt. Express 13(13), 5106–5115 (2005). [CrossRef] [PubMed]
24. T. B. Gibbon, K. Prince, T. T. Pham, A. Tatarczak, C. Neumeyr, E. Rönneberg, M. Ortsiefer, and I. T. Monroy, “VCSEL transmission at 10 Gb/s for 20 km single mode fiber WDM-PON without dispersion compensation or injection locking,” Opt. Fiber Technol. 17(1), 41–45 (2011). [CrossRef]
25. FEC ITU-T G.975.1, Forward error correction for high bit rate DWDM submarine systems [S], (2003).
26. B. Skubic and D. Hood, “A comparison of DBA for EPON, GPON, and NG TDM PON,” IEEE Commun. Mag. 47, 540–548 (2009).