Open Access
Add to BibSonomyAbstract
We show how dynamically adjustable modulation formats can be used to reduce link margins in flexgrid networks, reverting to lower order QAM due to reduced OSNR, if ageing occurs. Spectral savings amount to as much as 63% gain in capacity across a network using 64QAM with a fine frequency granularity of 6.25 GHz, with variable baud rate transponder. Further, a fixed baud rate, demand multiplexed transponder with adaptive modulation has been suggested. These transponders provide twice as much network capacity as compared to widely used fixed baud rate transponders operating at fixed grid of 50 GHz.
© 2013 Optical Society of America
1. Introduction
The network traffic continues to grow exponentially due to high data rate multimedia and internet services. These services use high bandwidth and put the onus on the network operators to efficiently use the network resources [1]. However, with the current 50 GHz ITU fixed grid this becomes difficult due to its lack of flexibility.
A network that is flexible, adaptive and can adjust to the required traffic can help resolve these issues. The combination of a flexible wavelength grid and components such as WSS and coherent transceivers has led to the new paradigm of flexgrid optical networks. There is currently a great deal of consideration being given to how to capitalize on this flexibility.
One area, unexplored until now, is the potential to use adaptable modulation formats to meet the requirements of the optical path, without additional margin. In this concept, the modulation format occupying the minimum spectrum is used, calculated on a path by path basis [2]. Each modulation format (ranging from BPSK, QPSK through to 64QAM) has a different associated OSNR requirement for good error rate performance. So for a channel with high link OSNR, a matching high order M-ary modulation format can be assigned to transmit more bits per symbol and thereby save on the spectrum resources [2].
Unallocated margins, design margins and system margins mentioned in [3] constitute the overall Link Margin (LM) and arise from ageing and uniformity effects: both operators and equipment vendors have good reasons for including margins in traditional optical network planning and these are typically several dB. In general, the LM is subtracted from the estimated OSNR as a precaution to guarantee that the link remains operational in the event of degradation. This gives the Link Operational OSNR which forms the basis of traditional system design – the system should continue to run error free even if all the margins are eroded. If the link degrades further than this then the link will fail – so margins are set relatively high. Dynamic modulation format control can be used to operate the system with reduced margin, but with the scope to adjust to lower order QAM under any equipment or link degradation [4]. This can lead to reduced spectrum utilization and hence more traffic on day one, together with a targeted response to ageing.
In this paper, we explore the magnitude of the benefits of reduced margins by doing simulations over a BT network for 100 Gb/s demands using PM-BPSK, PM-QPSK, PM-16QAM and PM-64QAM with variable baud rate transponders. We show the impact of reducing LM by 4 dB and 6 dB using dynamic modulation format operation. Our approach measures the amount of random traffic supported by the network at a given Cumulative Blocking Probability (CBP). We compare the results for different Frequency Granularities (FG) of 50 GHz, 25 GHz, 12.5 GHz and 6.25 GHz. Further, we introduce a demand multiplexed, fixed baud rate transponder and compare its performance with traditional fixed baud rate transponders for a given CBP over 50 GHz, across all the LMs.
2. Estimation of link OSNR
For a pair of source and destination nodes, we first find the shortest path on our network. As an example suppose we have a link shown in Fig. 1. Intermediate nodes 2, 3 and 4 are ROADMs. Each ROADM is modeled having a span loss of 22 dB which is immediately compensated by an EDFA. The line EDFAs are symmetrically placed with a maximum EDFA inter span spacing of 60 km. The resulting noise model [2] determines the estimated link OSNR, assuming an EDFA noise figure of 4.5 dB. In the example shown, an estimated link OSNR of 25.4 dB is calculated. Comparing this with OSNR thresholds for the key modulation formats given in Table 1, derived from [5] with pre-FEC BER of 4*10−3, we choose to operate on PM-64QAM and allocate a corresponding bandwidth of 16 GHz. This would then use either a single 50 GHz, a single 25 GHz, two 12.5 GHz or three 6.25 GHz flexgrid channels to accommodate this demand.
Table 1. OSNR Threshold for 100 Gb/s
3. Simulation results and discussion
Simulations were carried out on 100 different traffic matrices, each with up to 3000 demands of 100 Gb/s between random node pairs. Figure 2 shows the network under study. Demands were routed using shortest path and the highest QAM rate was found for each path using the noise model described above. Then, the required number of slots was allocated depending on which FG was in operation. Blocked demands were counted and summed to provide the CBP.
It is clearly observed in the bar graphs of Figs. 3(b), 4(b) and 5(b) that during the simulation most of the demands are served using 64QAM indicating high OSNR in the network. This is because most of the optical links are less than 200 km as the UK is a small country. This important result shows the need for extensive research on higher modulation formats, e.g. reducing the impact of laser phase noise to enable the potential for even higher QAM rates.
Looking at the CBP plots in Fig. 3(a), it is clear that a FG of 50 GHz fails to capitalize on the benefit of reducing the margin: the CBP curves for 0, 4 and 6 dB margin are identical. This is due to the generally high OSNRs for paths in the network leading to a modulation of 64QAM for many demands. Using a single 50 GHz slot to cover 16 GHz of spectrum wastes about 34 GHz and soon the network runs out of 50 GHz frequency slots.
Given there is no need for BPSK in such a small network, all the modulation formats fit into a single 50 GHz slot, making this option equivalent to the current fixed grid approach.
The CBP performance curves Fig. 4(a), for FG = 25 GHz and 12.5 GHz are the same with the curves for no margin and LM = 4 dB overlapping. From Fig. 4(b), due to high OSNR in the network, 64QAM and 16QAM are mainly used. However, the spectrum allocated to the demands served by 16QAM and 64QAM are the same because they both use either a single 25 GHz slot or two 12.5 GHz slots – identical total spectrum usage. For LM = 6 dB more number of demands are operating at 16QAM and QPSK which causes increased spectrum utilization and hence increases the CBP.
In contrast to the other FGs, a FG of 6.25 GHz results in distinct CBP curves, Fig. 5(a). Here we need to allocate only 3 slots or 18.75 GHz to cover the required spectrum of 16 GHz for 64QAM which is much less than for previous FGs. However, for 16QAM the spectrum utilization remains the same as previously (i.e. 25 GHz or 4 x 6.25 GHz slots). In Fig. 5(b), as the number of demands served by 64QAM in the LM = 0 dB system is higher than for LM = 4 dB, this results in increased spectrum saving and thus gives a lower CBP than LM = 4 and 6 dB.
It is seen in the CBP curves in Figs. 3(a), 4(a) and 5(a), that as we move towards finer granularity, the CBP decreases. This is because with finer FG, there is less wasted spectrum, resulting in a higher availability of contiguous frequency slots and thus leading to allocation of more demands. This is illustrated in Table 2, where a 62.9% increase in capacity is measured for a system with margin reduced by 6 dB, for FG = 6.25 GHz and evaluated at CBP = 0.1 (10% blocking probability). Finally the increased 64QAM demands for LM = 0 dB in Fig. 5(b) show that the 6.25 GHz grid is successful at preserving spare spectrum for 64QAM.
Table 2. Capacity Advantage of reducing margin by 4 or 6 dB at CBP = 0.1
4. Fixed baud rate transponders
In the above approach the baud rate has been allowed to change by keeping the data rate fixed [6]. Although, designing of variable baud rate transceiver is an open area of research, it is easier to implement transceivers with fixed baud rate. For our initial investigations of fixed baud rate transponders, we restrict use to a fixed 50 GHz grid and compare its performance with the traditional fixed baud rate transponders. For fixed baud rate of 27.75 Gbaud, a new OSNR Threshold table, Table 3, has been derived from [7]. For PM-64QAM the following Eqs. (1) and (2) have been used [8,9] and a non-linear penalty of 1.76 dB has been added from [7] with pre-FEC BER of 4*10−3. Each subcarrier occupies 30 GHz of spectrum [7].
Table 3. OSNR Threshold for Fixed Baud Rate of 27.75GBaud/s
In the traditional transponder approach (Design 1), we use currently available fixed baud rate, fixed modulation format transponders - e.g. fixed PM-QPSK. These fit well into 50 GHz fixed grid. There is no point in using higher order QAM unless we can access the additional capacity made available by it.
To access the additional capacity requires a more flexible transponder design, and so we introduce a second approach (Design 2). Here, multiple 100 Gb/s demands between same pair of source and destination nodes fill up a single carrier. Therefore, a PM-64QAM can cater to three 100 Gb/s demands and does not waste any capacity. We call them demand multiplexed transponders.
Looking at Fig. 6 we clearly see that using Design 2 allows the network to accommodate twice as much demands compared to traditional Design 1 transponders. Further, Design 2 shows increase in capacity with reduction in LM, because as LM is reduced we operate near to the OSNR Limit and higher modulation format. This provides much more capacity in terms of data rate, shown in Table 3, therefore allowing increased number of multiple 100 Gb/s demands to be allocated over a single carrier. The 0.1 CBP performance of Design 1 is equivalent to a variable baud rate transponder as both would use 50 GHz of spectrum to allocate a single 100 Gb/s demand for the operating modulation formats.
5. Conclusion
Flexgrid with dynamic modulation format can increase the day 1 capacity by over 60% by operating at reduced link margin. In the BT network, the shorter distances suggest the use of higher order QAM and a small frequency granularity, with 6.25 GHz by far the best. For smaller networks a bigger range of high order QAM modulation formats, fitting the OSNR profiles of the paths will increase the traffic more. However, the above approach is for variable baud rate transponder which is an open area for research. Further, a demand multiplexed transponder is introduced that takes advantage of the high link OSNR in the UK network. Using adaptive modulation and flexibility this transponder provides twice as much capacity compared to traditional fixed baud rate transponders. In future we will look at the impact on longer distance networks with different bit rate demands, and a wider range of QAM formats to better track the range of optical path OSNRs.
References and links
1. M. Jinno, B. Kozicki, H. Takara, A. Watanabe, Y. Sone, T. Tanaka, and A. Hirano, “Distance-adaptive spectrum resource allocation in spectrum-sliced elastic optical path network [Topics in Optical Communications],” IEEE Commun. Mag. 48(8), 138–145 (2010). [CrossRef]
2. A. Mitra, S. Kar, and A. Lord, “Effect of link margin on spectrum saving and advantages of flexgrid optical networking,” in Proceedings of National Conference on Communications (NCC) (Indian Institute of Technology Delhi, New Delhi, 2013), pp.1–5. [CrossRef]
3. J. Auge, “Can we use flexible transponders to reduce margins?” in Proceedings of Optical Fiber Communication Conference/National Fiber Optic Engineers Conference(OFC/NFOEC) (OFC/NFOEC, 2013), OTu2A.1. [CrossRef]
4. O. Gerstel, M. Jinno, A. Lord, and S. J. B. Yoo, “Elastic optical networking: a new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012). [CrossRef]
5. E. Lach and W. Idler, “Modulation formats for 100G and beyond,” Opt. Fiber Technol. 17(5), 377–386 (2011). [CrossRef]
6. A. Mitra, A. Lord, S. Kar, and P. Wright, “Effect of link margin and frequency granularity on the performance of a flexgrid optical network,” in Proceedings of European Conference and Exhibition on Optical Communication (ECOC, 2013),We.2.E.3.
7. G. Bosco, V. Curri, A. Carena, P. Poggiolini, and F. Forghieri, “On the performance of Nyquist-WDM terabit superchannels based on PM-BPSK, PM-QPSK, PM-8QAM and PM-16QAM subcarriers,” J. Lightwave Technol. 29(1), 53–61 (2011). [CrossRef]
8. D. Yoon, K. Cho, and J. Lee, “Bit error probability of M-ary quadrature amplitude modulation,” in Proceedings of Vehicular Technology Conference, 2000 (IEEE-VTS Fall VTC, 2000), pp. 2422–2427.
9. R. Essiambre, J. Kramer, P. J. Winzer, G. J. Foschini, and B. Goebel, “Capacity limits of optical fiber networks,” J. Lightwave Technol. 28(4), 662–701 (2010). [CrossRef]
