Open Access
Add to BibSonomyAbstract
We demonstrate how loss-optimised, gain-saturated SOA-REAM based reflective modulators can reduce the burst to burst power variations due to differential access loss in the upstream path in carrier distributed passive optical networks by 18dB compared to fixed linear gain modulators. We also show that the loss optimised device has a high tolerance to input power variations and can operate in deep saturation with minimal patterning penalties. Finally, we demonstrate that an optimised device can operate across the C-Band and also over a transmission distance of 80km.
©2011 Optical Society of America
1. Introduction
The Gigabit Passive Optical Network (GPON), in both its ITU- [1,2] and IEEE- [3] standardised forms, is now a well-established broadband access technology that is currently undergoing widespread deployment around the world. As a consequence, research attention has recently turned to the development of new access technologies that may supersede GPON in the future. One potential candidate is the hybrid Dense Wavelength Division Multiplexed, Time Division Multiple Access Passive Optical Network (DWDM-TDMA PON) [4] shown schematically in Fig. 1 . These systems differ from today’s optical access systems in three major aspects. First of all, DWDM is used to dramatically increase the capacity of the network. Second, the physical reach is increased significantly from today’s usual 20km (60km in the case of ITU-T G984.6 extended reach GPON [5]) to over 100km. Finally, a single system will serve several thousand optical network units (ONUs) rather than the typical 32 of today’s deployed optical access systems [6,7]. Therefore this architecture exploits the wavelength domain for capacity growth whilst retaining the capability to share the (potentially very large) per wavelength bandwidth across multiple subscribers using power splitters. A key challenge for the realisation of these networks is the DWDM source at the subscriber’s ONU. The source must stay tuned to the DWDM grid with a high level of accuracy. This must be achieved at the lowest possible cost and therefore network operators do not want to keep a large stock of ONUs with wavelength specific lasers as this brings additional inventory and management costs. A low cost colourless (wavelength agnostic) transmitter is the preferred choice in the ONU. A carrier distribution scheme with reflective modulators at the ONUs provides one promising approach to achieve this goal [8]. If the subscriber modulator is based on a semiconductor optical amplifier (SOA) and a reflective semiconductor electro-absorption modulator (REAM) integrated together in a single monolithic device (SOA-REAM), then it is possible in principle to achieve sufficient net gain to overcome splitting loss in the network and achieve sufficiently low-chirp modulation for 10Gb/s operation over extended distance [8–10]. However, one issue that has received comparatively limited attention to date is the impact of differential loss (DL) on this scheme. DL in PONs, which arises from variations in splitter port loss, non-symmetric splitter distributions and fibre path length variations, causes large burst-to-burst amplitude variations in the upstream signals from ONUs at different locations in the PON. In GPON, for example, DL values of 15dB are typically assumed. If we now consider a scenario in which a deployed GPON is being upgraded to a hybrid DWDM-TDMA PON, the impact of DL is multiplied due to the fact that the optical carrier also experiences the DL en-route to the ONUs. Hence if the ONUs all employ fixed gain reflective modulators the upstream dynamic range will be 30dB or even larger if manufacturing-induced variations in device gain and input/output coupling loss are included. This is beyond the reach of today’s 10Gb/s burst mode receiver (BMRx) technology. The large variation of ONU input power due to the DL experienced by the carrier is also of concern, since it will typically lead to the SOA sections in different ONUs operating with different degrees of gain saturation. The latter leads to gain recovery effects when the device is required to operate with modulated signals which have high input powers and results in data patterning which in turn leads to errors at the receiver.
Fig. 1 Carrier-distributed hybrid DWDM-TDMA PON. Downstream not shown for clarity (US: Upstream, AWG: Arrayed-Waveguide Grating).
In this paper, we demonstrate a novel solution to this problem based on optimisation of the internal loss of the SOA-REAM. Whilst a typical approach to integrated device design would be to minimise such losses, we show here the counter-intuitive result that large values of internal loss are in fact highly beneficial, since this allows the device to operate deep in saturation with minimal pattering penalties and significant dynamic range compression, enabling the use of standard BMRx technology.
2. Experimental Set-up
The experimental set-up is shown in Fig. 2 . An integrated SOA-REAM was emulated using a discrete SOA (CIP SOA-S-OEC-1550) and a discrete REAM (CIP REAM-1550-LS) connected together via a variable optical attenuator (VOA2) with a double pass loss, LINT, which simulates additional internal device loss. This configuration allowed the effects of variation in LINT to be investigated which would not be possible with a fully integrated device, albeit with the trade-off of longer (~60ns) round-trip delays. When driven at the designed operating current of 100mA, the SOA had a small signal gain of + 23dB at the gain peak wavelength of 1550nm, a saturated output power, Psat of + 8dBm (i.e. 3dB gain saturation), a polarization dependent gain of less than 0.5dB and a noise figure of 6.2dB. The REAM was modulated at 10Gb/s with 3Vpp non return-to-zero (NRZ) (231-1 pseudo-random bit sequence (PRBS)) data superimposed on a DC bias, which was set to −1.4V in the back-to-back case (no transmission fibre). A CW optical carrier (wavelength: 1550nm) generated by an external cavity tuneable laser was injected into the emulated SOA-REAM via a circulator. The input power to the emulated device was varied using VOA1 to give a range of typical input powers present in a PON due to the variation in loss experienced by the carrier. The eye patterns were measured at the output of the device using a digital sampling oscilloscope with a 30GHz bandwidth optical input. The bit error rate (BER) was measured using a 10Gb/s APD receiver (−26dBm sensitivity 10−9 BER) coupled to an error detector. VOA3 was used to emulate the loss experienced by the modulated upstream signal on route to the local exchange. An erbium doped fibre amplifier (EDFA) was added to the setup to represent the local exchange amplifier which would be present in a long reach PON. A 1nm bandpass filter tuned to 1550nm was used to minimise the amplified spontaneous emission (ASE) falling on the receiver. The transmission performance of the emulated SOA-REAM was also investigated over a transmission distance of 80km of single mode fibre. The REAM DC bias was increased to −2V to optimise the device for transmission.
Fig. 2 Experimental set up. PPG: Pulse pattern generator, APD: Avalanche photodiode, PC: Polarisation controller
3. Experimental Results and Discussion
Figures 3a -3e show the measured mean modulated output power (Pout) vs. CW input power (Pin) characteristics for the emulated SOA-REAM, obtained for LINT values in the range 3dB to 15dB (3dB step size). The output eyes measured for Pin values in the range −20dBm to 0dBm (5 dB step size) are also shown, and Fig. 3f shows all the Pin vs. Pout curves plotted together on the same vertical scale for LINT values from 3dB to 21dB (3dB step size). The results show a number of unique and striking features; firstly the device has a highly non-linear, peaked transfer function which results from SOA gain saturation effects, and secondly the degree of data patterning and eye closure induced by the gain saturation depends strongly on the value of LINT and is greatly reduced at the higher loss values. A conceptual understanding of these effects can be obtained if the forward propagating CW carrier and the reverse propagating modulated beam through the SOA are considered sequentially. As the input carrier power Pin is increased, the SOA eventually enters saturation on the first pass causing the output carrier power to clamp at a value close to Psat (the output saturation power of the SOA).
Fig. 3 (a)-(e): Pin vs. Pout and eye diagrams for LINT values of 3dB to 15dB in 3dB steps. (f): Pin vs. Pout on same scale
At this point, further increase in Pin leads to an approximately linear reduction in gain, with minimal increase in carrier power that enters the REAM. If the internal loss of the device is relatively high, the modulated return signal generated by the REAM does not contribute significantly to gain saturation and instead simply ‘probes’ the gain of the SOA on the second pass. The result is an initial rise in Pout with Pin followed by a peak and fall as the carrier pushes the device into saturation. The exact value of internal loss determines the relative contributions that the CW carrier and modulated return signal make to SOA gain saturation. For example, if the loss is small the modulated signal contributes significantly to saturation and patterning occurs (e.g. Figure 3a), whereas if the loss is large the saturation is dominated by the carrier, which acts as a ‘CW holding beam’ [11] that speeds up gain recovery and reduces patterning (e.g. Figure 3d). This picture explains previous results obtained with a prototype fully-integrated device with high SOA to REAM coupling loss [8]. We now consider the implications of these properties in DWDM-TDMA PONs. To simplify the analysis we consider the specific scenario of upgrade of a GPON with 30dB maximum loss and 15dB minimum loss for the access distribution network (i.e. 15dB differential loss). We further assume a carrier launch power of + 15dBm from the local exchange (LE) into the access network, which is representative of the typical achievable values if standard Stimulated Brillouin Scattering mitigation techniques are employed [9,10]. Consequently the carrier power at the inputs to the SOA-REAMs on the network will vary between −15dBm (soft) and 0dBm (loud). The same differential loss is encountered by the reflected modulated signal and in the experimental setup this is emulated using attenuator VOA3. To account for the differential loss 15dB difference in attenuation is applied for the loud and soft signals respectively. For these two Pin values Fig. 4 shows, as a function of LINT, the B2B receiver sensitivity penalty at 10−9 BER relative to an REAM alone. It can be seen how the penalty increases rapidly for loss values below 12dB due to patterning-induced eye closure. Best case performance is achieved for an LINT value of 15dB, which gives a 0.5dB penalty for the loud packet and 1.5dB penalty for the soft packet. This 1dB higher penalty of the soft packets is attributed to the amplified spontaneous emission from the EDFA, which reduces the OSNR of the soft packets. As LINT is increased further the SOA-REAM output power decreases, which in turn reduces the signal power at the input to the LE amplifier and hence increases the OSNR penalty, particularly for the low power soft packets. As mentioned, for the current SOA-REAM design, the optimum LINT value that minimises total power penalty is 15dB.
When added to the REAM return loss (~9dB) this gives a total optimum device loss of 24dB, which is close to the single pass SOA gain. In order to investigate the performance of the loss optimised device in transmission, 80km of standard single mode fibre was then added after the EDFA as shown in Fig. 2. As we have established the optimised region to operate the device, only LINT values of 12dB and 15dB were tested under transmission. The BER was measured for loud and soft signals after 80km and compared to the B2B results (Fig. 5(a) ). We observe less than 1dB dispersion penalty for both loud and soft signals. This is a typical dispersion penalty for an EAM over 80km transmission and indicates no additional impairments from the SOA or high values of LINT.
Finally, a characterisation of the performance of the loss-optimized device at representative wavelengths across the C-Band was carried out. To this end, the EDFA, transmission fibre and VOA3 were removed. As above, the plot in Fig. 5(b) was normalised to the optimal B2B receiver sensitivity achieved using an REAM alone operating at 1550nm. We observe a maximum power penalty of 2dB in the wavelength range 1530nm to 1570nm.
As described earlier, the wide dynamic range at the upstream receiver due to the differential loss in the access network is a major implementation challenge for a carrier distributed PON. Here we show the dynamic range for the loud/soft signals at the input to the LE (and hence at the BMRx) as a function of differential loss in Fig. 6 . It can be seen that by using an SOA-REAM the dynamic range can be reduced compared to a system where all reflective modulators have the same gain regardless of their position in the PON (linear gain case). Also it is evident that as LINT is increased the dynamic range compression increases due to the beneficial effects of the gain saturation, so not only does increasing LINT suppress patterning effects and enable the device to operate over a wide range of input powers but it also dramatically reduces the loud/soft dynamic range of the upstream signal at the input to the LE. For example, for a differential loss of 15dB and an LINT value of 15dB, the loud/soft dynamic range is only 12dB compared with 30dB for the linear gain case. As mentioned earlier it is not desirable to increase LINT to values greater than 15dB due to the reduced output power, but it can also be seen in Fig. 6 that increasing LINT past 15dB gives almost no increase in dynamic range compression.
Therefore the device can operate in the optimum region for patterning reduction without significantly compromising output power while achieving almost maximum dynamic range compression. The plot also shows that the dynamic range compression continues for differential losses greater than 15dB. In other words the loss optimised SOA-REAM also has potential to operate in a network with increased splits and/or tolerate variation in reflective modulator coupling efficiency.
4. Conclusion
A novel solution for dynamic range reduction in carrier-distributed hybrid DWDM-TDMA PONs has been demonstrated based on loss-optimised SOA-REAMs. It is demonstrated that, counter-intuitively, larger internal device losses can actually improve system performance. The scheme allows the device to operate over a wide range of input powers and in deep saturation with minimal pattering penalties and significant dynamic range compression, enabling the use of standard BMRx technology. Measurement of dispersion penalties and a characterisation over the C-band confirms that the device can be used for transmission distances and wavelength plans typical for hybrid DWDM-TMDA carrier distributed PONs.
Acknowledgments
We thank Science Foundation Ireland for financial support (Grant 06/IN/I969) and R. Manning and R. Webb for helpful advice. G. Talli is currently with Intune networks.
References and links
1. ITU-T G987.2, “Ten gigabit - capable passive optical networks: Physical dependent layer specifications,” Jan. 2010.
2. ITU-T G987.2, “Ten gigabit - capable passive optical networks: Service requirements,” Jan, 2010.
3. IEEE P802.3av, “10Gb/s Ethernet Passive Optical Networks,” 2009.
4. R. P. Davey, D. B. Grossman, M. Rasztovits-Wiech, D. B. Payne, D. Nesset, A. E. Kelly, A. Rafel, S. Appathurai, and S.-H. Yang, “Long-reach passive optical networks,” J. Lightwave Technol. 27(3), 273–291 (2009). [CrossRef]
5. ITU-T G.984.6, “Gigabit-capable passive optical networks (GPON) Reach extension,” Mar. 2008.
6. G. Talli and P. D. Townsend, “Hybrid DWDM-TDM long reach PON for next generation optical access,” J. Lightwave Technol. 24(7no. 7pp), 2827–2834 (2006). [CrossRef]
7. A. Banerjee, Y. Park, F. Clarke, H. Song, S. Yang, G. Kramer, K. Kim, and B. Mukherjee, “Wavelength-division-multiplexed passive optical network (WDM-PON) technologies for broadband access: A review [Invited],” J. Opt. Netw. 4(11), 737–758 (2005). [CrossRef]
8. E. K. MacHale, et al., “Extended-Reach PON employing 10Gb/s Integrated Reflective EAM-SOA,” presented at the European Conf. Optical communication (ECOC), Brussels, Belgium, 2008, Paper Th.2.F.1.
9. C. Antony et al., “Demonstration of a carrier distributed, 8192-split hybrid DWDM-TDMA PON over 124 km field-installed fibres,” Optical fiber conference, San Diego, CA, Mar. 2010, Postdeadline paper PDPD8.
10. P. Ossieur, C. Antony, A. M. Clarke, A. Naughton, H.-G. Krimmel, Y. Chang, C. Ford, A. Borghesani, D. G. Moodie, A. Poustie, R. Wyatt, B. Harmon, I. Lealman, G. Maxwell, D. Rogers, D. W. Smith, D. Nesset, R. P. Davey, and P. D. Townsend, “A 135km, 8192-Split, Carrier Distributed DWDM-TDMA PON with 2x32x10Gb/s Capacity,” J. Lightwave Technol. 29(4Issue 4), 463–474 (2011). [CrossRef]
11. R. J. Manning and D. A. O. Davies, “Three-wavelength device for all-optical signal processing,” Opt. Lett. 19(12), 889–991 (1994). [CrossRef] [PubMed]
