Open Access
Add to BibSonomyAbstract
An open converged metro-access network approach allows for sharing optical layer resources like fibers and optical spectrum among different services and operators. We demonstrated experimentally the feasibility of such a concept by the simultaneous operation of multiple services showing different modulation formats and multiplexing techniques. Flexible access nodes are implemented including semiconductor optical amplifiers to create a transparent and reconfigurable optical ring network. The impact of cascaded optical amplifiers on the signal quality is studied along the ring. In addition, the influence of high power rival signals in the same waveband and in the same fiber is analyzed.
© 2014 Optical Society of America
1. Introduction
More and more it is accepted that optical networks are becoming an important factor for enhancing the economic attractiveness of cities and regions. However, the deployment of fiber infrastructures needs a large investment, slowing down the proliferation of optical networks, particularly in the access space. Today’s optical networks in the metro and access area are mostly being owned and operated by single operators or companies for a dedicated purpose. For instance, operators install and manage as separate networks fiber-to-the-home (FTTH) and fiber-to-the-building (FTTB) services as well as optical backhauling systems. However, when looking at the slow progress of fiber network deployments these business cases do not seem sufficiently compelling or are just too risky, especially in view of regulatory uncertainties [1].
The remedy to this deficiency is the operation of an open access network implementation on the optical layer, making efficient use of the large optical bandwidth of fibers. It would allow the utilization of a single infrastructure for multiple communication purposes at the same time. The optical layer resources, such as fibers and optical spectrum, are shared among different services and operators and/or other client systems. Such a network is certainly challenging to implement and operate, but conceptually it opens attractive business opportunities. In order to allow for the most open and flexible resource utilization and to avoid the restriction to a certain modulation format or multiplexing technique, the infrastructure should remain transparent as far as possible. Therefore, the transmitted optical signals should not be converted to the electrical domain, and should not be processed in a protocol-specific way. The target for the infrastructure design is to make it a versatile and flexible grid for any kind of services in the access and metro area [2].
The term open access network is used in the literature in different meanings [3, 4]. We define an open access network as an infrastructure for multiple operators utilizing a multitude of services using all kinds of modulation formats, i.e. a virtually transparent metro access network. A possible scenario is shown in Fig. 1.The network covers the area of a big city, has a diameter of up to 50 km, and supports typical residential services like FTTH and FTTB, enterprise networks via high-speed optical channels, and also wireless backhauling.
Fig. 1 Open metro-access network configuration with basic services such as FTTH/FTTB (residential access), wireless backhauling and business access on a common infrastructure.
The connection to the core network is established via a few active nodes only (green boxes) which include the network’s intelligent routing functionality. By implementing all-optical nodes (orange boxes) into an existing metro-access fiber network the infrastructure can be shared and utilized in a flexible way. These all-optical and system-agnostic nodes represent access points (AP) that include network functions like all-optical signal regeneration (e.g. amplification), dispersion compensation, and performance monitoring.
The basic idea of that novel network concept is to establish multiple optical pipes, shown in Fig. 2,that run in parallel through the network, within which different clients/network providers can independently allocate different kinds of service. These optical pipes are used to generate independent connections between the client equipment / networks, thus enabling a transparent optical infrastructure. The pipes can be considered as virtual “dark fibers” usable with almost any bit rate, modulation format, multiplexing technique, protocol and network topology (ptp, ptmp, mesh, ring, etc.). All-optical fiber switches will allow for a flexible reconfiguration of the virtual sub-networks, an option which is especially important for temporarily used pipes. The optical pipes are established and managed at the access points, where certain regions of the fiber transmission spectrum are administered. Exemplarily, Fig. 2 depicts a typical waveband structure which is compliant with common standards for, e.g., G-PON, XG-PON, NG-PON2, EPON and 10G-EPON [5, 6]. The fiber spectrum is divided into a variety of wavebands with possibly unequal spectral width (a few nanometers to 20 nm, variable from fiber to fiber). The wavebands can carry services for residential subscribers like single channel G-PON (TDM-PON) and stacked XG-PON (TWDM-PON) as well as services for business subscribers like multi-channel DWDM systems (100 Gbit/s, 400 Gbit/s).
To ensure the transparency and independence of optical pipes, particularly, loss-compensating linear optical amplifiers [7] need to be deployed. Special requirements of such optical amplifiers are the amplification of data signals with different modulation formats, bit rates, or multiplexing techniques, especially for a chain of them [8]. The optical amplifiers must not disturb significantly the signal quality neither by non-linear effect nor by noise. Dependent on the network realization these amplifiers should offer a moderate gain (> 10…15 dB), low polarization dependence (< 0.5 dB) and a moderate noise figure (~8 dB). The efficient utilization of the fiber spectrum requires optical amplifiers with a gain that can be centered at any waveband. Especially, semiconductor optical amplifiers (SOA) can fulfill this demanding requirement, and thus, they are a beneficial technology for the realization of such networks.
In this paper, we focus on the physical layer implementation and on experimental results of a field trial which demonstrates the feasibility of the described network concept. Flexible access nodes are implemented including amplification and switching to create a transparent and reconfigurable optical ring network. Different kinds of services are transported and their performance is studied along the SOA chain. Results of that field trial have been presented for the first time in [9]. In this paper, a more detailed description of the underlying concept and the performed experiments are discussed. In addition, results on the influence of rival signals within the wavebands as well as interaction between wavebands are investigated. The overall measurement results show very good performance and a highly robust network infrastructure.
2. Field trial scenario
The architecture of the field trial is shown in Fig. 3(a).Four APs have been implemented into fiber links in the network of the Deutsche Telekom in Berlin, see Fig. 3(b), forming a double fiber-ring network topology. It consists of already deployed fibers with a length of 18.1 km between the access points AP1-AP2 and AP1-AP4, respectively, and coiled fibers on spools with a length of 9 km (AP2-AP3) and 6.5 km (AP3-AP4). A set of services has been selected showing different modulation formats, multiplexing techniques and bit rates for connecting residential customers via FTT(H/B) (G-PON), for connecting enterprises via high speed optical channels (4x10 Gbit/s or 1x100 Gbit/s), for connecting radio antenna with switched radio-over-fiber (RoF) signals, and also providing wireless backhauling (X2 traffic between eNodeB (base stations)), e.g., by using orthogonal frequency-division-multiplexing (OFDM) signals. As shown in Fig. 3(c), four wavebands are allocated to the different services. The G-PON system operates at 1311 nm (upstream, US) and 1491 nm (downstream, DS), respectively. The other two wavebands are both shared for different systems/services. The re-use of wavelength bands as well as the add/drop switching functionality per waveband have been successfully demonstrated:
Fig. 3 Architecture / traffic configuration of the field trial infrastructure (a) incorporating installed fiber links of the Deutsche Telekom in Berlin (b) and the allocation plan of the operational wavebands (c).
- Within the 1551 nm waveband, at AP 3, the OFDM signal is dropped and simultaneously the RoF signal is added, see Fig. 3(a).
- Within the 1531 nm waveband the business access services are either added at AP1, 10 Gbit/s, or at AP2, 100 Gbit/s, see Fig. 3(a). The used service is dropped from the ring at AP4.
Thus, the signals make use of the same optical path between AP2 and AP4. No readjustment of optical amplifier gain at the involved APs is needed.
The AP consists of reconfigurable optical add/drop multiplexers (ROADM) enabling the management of optical wavebands throughout the network. The ROADM within our approach can be distinguished from a typical state-of-the-art high-speed metro ROADM by the following:
- Processing of a few wavebands (some nm up to 20 nm) always containing several data signals (on varying wavelength grids) vs. processing of individual wavelength channels, e.g., on a fixed 50 GHz DWDM grid.
- Individual waveband amplification vs. common amplification of all channels simultaneously
- Nonrecurring power level adaptation of wavebands at implementation vs. permanent power level adaptation of individual wavelength channels during operation (at amplifier input/output)
Figure 4depicts the set-up of a typical ROADM implementation used in the field demonstrator. It enables the independent handling of, in our case, 6 wavebands which are 13 nm wide (1 dB bandwidth). Each band can be individually passed through or added/dropped by optical cross/bar switches employing mechanically steered beam deflectors (provided by Leoni Fiber Optics in the framework of the BMBF-project CONDOR).
Fig. 4 Setup of ROADM used within the field trial demonstrator. The ROADM enables individually passing through or adding/dropping of data signals on the wavebands as well as their amplification by SOA and monitoring.
In each waveband losses are compensated by optical amplifiers which ensure the independent operation of the different services running on the optical pipes. In our experiments, we implemented and tested commercially available SOAs which have gain of about 15 dB, a polarization dependence of 0.5 dB, and a noise figure of 7.5 dB. In the 1.55 μm (1.31 μm) window, we found a 3 dB saturation input power of 0 dBm (−1 dBm) and a 3 dB gain bandwidth of 90 nm (60 nm). In all measurements, the SOA gain is used to compensate for fiber and ROADM insertion losses. The remaining device gain is compensated by attenuators to obtain a net gain of 0 dB at a total SOA input power of about −10 dBm.
All wavebands are monitored via −15 dB tap couplers at the input ports of the cross / bar switches and at the output of the SOAs. Subsequent −3 dB power splitters allow for simultaneous measurement of the integral optical power on all ports (continuously) and of the optical spectrum at one port at a time (on demand). The optical spectrum analyzer (OSA) is connected via a 24 x 1 switch.
3. Measurement results
The main objective of the field trial was to demonstrate the simultaneous operation of multiple heterogeneous services on a common shared optical infrastructure. As explained in the previous sections, the data signals are passed over the fibers from AP to AP which results in a chain of SOAs. The number of cascaded SOAs depends on the service transmission path. Due to the fact that the amplification process adds noise and nonlinearities, an in-depth study of performance limitations is required.
In the following, results of various experiments of the field trial are presented, highlighting the feasibility of the concept which shows exceptionally good performance of the involved services even when cascading several SOAs. We demonstrate the following:
- Multiple use of the same waveband (optical pipe) for different services
- Impact of SOA cascade on different system/service types
- Influence of high power rival signals on signals within the same waveband or between wavebands
3.1 Use of the same waveband for different services
The open network concept has to allow for multiple use of the same optical pipe for providing different services without the need of physical parameter re-adjustments of the involved network elements, especially without changing the SOA gain.
Exemplarily, a business access service centered around a wavelength of 1531 nm is established either by using 4 decorrelated 10 Gbit/s NRZ-OOK DWDM (200 GHz spacing) channels (different XFPs with different PPGs), or by using a single channel 100 Gbit/s signal with 28 GBd dual-polarization QPSK. Figure 5shows the detailed measurement set-up as an enlarged partial view of Fig. 3(a). At AP1 the DWDM signals with wavelengths of C54, C56, C58 and C60 (ITUT-T grid) are added to the fiber ring and transported over the field fiber (including a dispersion compensating fiber to compensate for 20 km SMF), two APs (AP2 to AP3, connected by lab fibers). The signals are dropped at AP 4 for measuring the bit-error ratio (BER).
Fig. 5 Multiple use of the same waveband (optical pipe) for different services: (1) by 4 x 10 Gbit/s DWDM system (Txs located at Wannsee) or (2) by 1 x 100 Gbit/s system (Tx located at Winterfeldstraße, see dashed boxes), the inserts show the corresponding optical spectra at 1531 nm.
In the second part of that experiment, the DWDM signals were dropped at AP2, and simultaneously the 100 Gbit/s data signal (see dashed boxes in Fig. 5) was added to the same waveband and transported to the receiver at AP4. Therefore, both systems make use of an identical optical path between AP2 and AP4. The performance analysis presented in Fig. 6 is obtained without readjustment of any SOA gain within the respective APs.
Fig. 6 Multiple use of same pipe for establishing different business access connections: (a) power penalty of 4 ch with 10 Gbit/s OOK and (b) Q2 degradation of 100 Gbit/s DP-QPSK. The insets show in (a) a typical spectrum and in (b) constellation diagrams for 3 characteristic input power levels (noise limited, optimum, nonlinearity limited).
In Fig. 6(a) the power penalty (receiver sensitivities of signals after transmission compared with back-to-back (BtB) receiver sensitivity at a BER of 10−9) of one 10 Gbit/s test channel (C58) versus the total input power into the 1st SOA in the chain is shown. In Fig. 6(b) the Q2 degradation (compared to a Q2 factor of 17 dB) of the 100 Gbit/s signal is presented. The measurement results are obtained by changing the input power levels at the first SOA in the chain only. The curves indicate a degradation of the system performance due to SOA noise at low input powers, and due to gain saturation-induced distortions at large input powers. Even after 4 SOAs, Fig. 6(a), or 3 SOAs, Fig. 6(b), the input power dynamic range in which the penalties remain below 2 dB exceeds 15 dB. Different services using the same optical pipe are possible by cascading SOAs which operate in their linear range.
3.2 Impact of SOA cascade on different system/service types
The impact of several cascaded SOAs on different system/service types has been investigated. Two GPON systems with bit rates of 2.5 / 1.25 Gbit/s (in down/up-stream) have been established within the demonstrator: GPON 1 for providing residential access for users as FTTH/FTTB service (from AP1 to AP2 in Fig. 3(a)), and GPON 2 for optical backhauling (S1 interface) in mobile networks (from AP1 to AP3) [4]. A typical burst mode upstream path has been installed using a burst of 10 µs duration with a repetition rate of 8 kHz, inset Fig. 7.The power penalty for a target BER of 10−9 versus the total input power into the 1st SOA in the chain is depicted for two cases in Fig. 7:
- The black curve shows the performance of the GPON 1 upstream path where only one SOA located at AP2 is used.
- The green curve shows the performance for the GPON 1 upstream path where 4 cascaded SOAs are used.
The SOA cascade is adapted again to a net gain of 0 dB at individual SOA input power levels of −10 dBm. Thus, if the input power into the first SOA in the chain is increased/decreased, also the input power levels into subsequent SOAs increase/decrease. As a result, we found that the power penalty increases at low SOA input powers due to the accumulation of noise, which deteriorates the optical-signal-to-noise ratio (OSNR). At large input powers, the OSNR worsens due to gain saturation-induced signal distortions. Comparing the results of the two cases we see that for cascaded SOAs the impact on the signal quality due to nonlinear effects is stronger than the influence of noise. It should be mentioned that these specific measurements were performed in the laboratory, and not during the field trial.
In our field trial, a switched radio-over-fiber (RoF) system is incorporated, which has significantly increased system performance requirements [10] compared to, e.g., OOK signals. The RoF system mimics the centralized supply of base station antennas with RF-signals. The required transmission lengths need a few SOAs only. The RoF signal is based on a pulse-width-modulation with 5 MHz bandwidth. The most important parameter for this RoF signal is the adjacent channel leakage ratio (ACLR), see inset in Fig. 8.It defines the power that leaks from a transmitted signal into adjacent channels. Data signals at a 0.9 GHz electrical carrier frequency have been transported from AP3 to AP2 (passing 2 SOAs) as well as from AP3 to AP4 (passing 4 SOAs). Figure 8 shows for both cases the ACLR degradation after the SOA chain (compared to optical BtB measurements) versus the input power into the first SOA. The ACLR in our experiments should be better than 45 dB which means that an ACLR degradation of 4 dB can be tolerated to fulfill 3GPP specification [11]. This is achieved in our experiments within a limited SOA input power range only. Nevertheless, significant improvements can be expected by an integration of the discrete setup. It should be mentioned that the error-vector magnitude (EVM) is always near a very good performance level of 10%, and it does not severely deteriorate by the cascading of SOA.
Future broadband wireless networks need an efficient backhauling design. Beside the transport of signals from base stations (eNodeB) back to the central office (GPON2 in Fig. 2), a cooperative multipoint processing (CoMP) will be applied for increasing network performance. Here, high capacity traffic between neighboring cells (eNodeB/eNodeB communication; X2 traffic [4]) will occur which can be served by using OFDM signals.
The OFDM backhauling approach works as follows: Each eNodeB transmits at a different optical carrier frequency an eNodeB-specific OFDM signal on a different intermediate electrical carrier frequency. Following transmission over the fiber ring, an eNodeB receives the OFDM signals from neighboring cells. At the receiver site only the electrical OFDM signals from different eNodeB transmitters are detected. Here, the high peak-to-average power ratio (PAPR) of OFDM signals provides a challenging requirement for the SOA chain.
Within our field trial, two OFDM transmitters are used. The first one uses a FPGA to generate an OFDM signal with 64 subcarriers carrying a QPSK data signal with a total data rate of 1.2 Gbit/s. The second one uses an arbitrary-waveform generator (AWG) to generate an OFDM signal with 256 subcarriers carrying QPSK data with a total data rate of 5 Gbit/s. A single receiver comprising a PIN and an 80 GS/s real-time oscilloscope detects the signals from two OFDM transmitters after 2 SOAs and 6.5 km single-mode fibre. Figure 9shows results for the successful proof-of-concept experiment. Figure 9(a) shows constellation diagrams of the OFDM Txs with low EVM values, and Fig. 9(b) shows the corresponding optical spectrum. Figure 9(c) shows the constellation diagrams of the received OFDM signals with EVM values corresponding to BER < 10−3. Figure 9(d) shows the electrical spectrum of the simultaneously received signals.
3.3 Influence of high power rival signals on signals within the same waveband or between wavebands
The discussed open access network approach allows the simultaneous operation of multiple services in separate optical pipes as well as services using multiple wavelength channels in a specific optical pipe. In this regard it is of special interest how the signal performance will be influenced when rival signals, especially with the largest possible optical power for the network scheme, are present in different wavebands on the same fiber (inter-waveband) or within the same waveband on the same fiber (intra-waveband).
Inter-waveband distortion
For investigating the inter-waveband influence, the field trial set-up as shown in Fig. 3(a) has been modified to allow the transmission of four different services in parallel through the fiber ring network. Four wavebands with neighboring channels lying on the CWDM grid have been selected for carrying the services over varying numbers of APs (1…4 SOAs). The following services are selected: An GPON downstream signal in the 1491 nm waveband, an optical Gbit-Ethernet data signal in the 1511 nm waveband, the 4 channel DWDM system with 10 Gbit/s signals in the 1531 nm waveband, and the switched RoF signal in the 1551 nm waveband. This study is used to investigate the cross-talk which can occur by nonlinearities of the SMF between the wavebands.
The chosen scenario mimics a typical realization of the network in which the power levels between wavebands are within 10 dB. Cross-talk induced by non-ideal filter suppression is not studied here (30 dB suppression of used CWDM filters on adjacent channel). Thus we focus our study on fiber nonlinearities, add an erbium-doped-fiber-amplifier (EDFA) in the RoF path, and achieve a high output power signal as a rival signal to the 10 Gbit/s system. The total optical power at the input of the first fiber span with about 9 km amounts to 10 dBm. The power levels of the individual wavebands are about + 8 dBm for the RoF rival signal and 0…-2 dBm for the other wavebands.
The performance of one (C56) of the four DWDM 10 Gbit/s signals after cascading 4 SOAs is tested with and without the presence of the 3 rival wavebands (RoF peak power level is 10 dB larger than peak channel power of 10 Gbit/s signal). The results of the measurements are shown in Fig. 10 for both cases (spectra: Fig. 10(a): only 10 Gbit/s system, Fig. 10(b): all wavebands are present). Obviously, no influence on the BER is to be seen, which indicates that the rival signals (wavebands) have no impact. Thus, for realistic power levels inside our scenario, no inter-waveband distortion due to fiber nonlinearities is expected.
Fig. 10 Measurement of impact of rival signals with high power levels in same fiber: Resulting BER measurement for C56 of 4x10Gbit/s DWDM system without (spectra (a)) and with rival signals in neighboring wavebands (spectra (b)). The inset shows a typical eye diagram of the 10 Gbit/s data signal.
Intra-waveband distortion
For investigating the intra-waveband influence, the field trial set-up as shown in Fig. 3(a) has been modified again. Two experiments are performed which differ in the number of channels, data rate and modulation format. In a first experiment, see Fig. 11(a).the 1531 nm waveband carrying a single 100 Gbit/s (C57) system in the field trial is extended with a second 100 Gbit/s (C58, DP-QPDK) signal.
Fig. 11 Measurement of impact of rival signal with high power in same waveband (intra-waveband): set-up with two 100 Gbit/s carriers in a 1531 nm waveband, resulting Q2-factor degradation of the original 100 Gbit/s signal compared to ideal Q2-factor of 17 dB (b).
Both signals are decorrelated and added to the network at AP2, then amplified by 3 cascaded SOAs (AP2…4), transported over 9 km and 6.5 km SMF, and finally dropped at AP4 to the coherent receiver. The optical power of the rival signal (C58) is increased by an EDFA and is adjusted by an attenuator before launching it into the first SOA of the chain. The test signal (C57) is always operated in its optimum for the single channel experiment corresponding to an input power of −10 dBm, see Fig. 6(b). The inset shows a signal spectrum for the two channels in front of the SOA chain for the case in which the rival signal has 15 dB higher power level than the test channel.
We measured the Q2 degradation of the 100 Gbit/s test signal in dependence of the optical power of the rival signal into the first SOA of the chain, and compared to the results for the Q2-factor degradation in the single channel case. Figure 11(b) shows that the observed intra-waveband distortion of the SOA via cross-gain (XGM) and cross-phase modulation (XPM) is relatively low. Compared to the optimum operating point for the single-channel case, the rival signal can have 9 dB higher input power levels (about −1 dBm) for a Q2-factor degradation of 2 dB.
In a second experiment the impact of 3 × 10 Gbit/s OOK channels on a 10 Gbit/s test channel (C58) is studied. The four decorrelated DWDM channels (C60, C58, C56, C54) are spaced by 100 GHz. As shown in Fig. 5 the four channels are transmitted over the fiber links and amplified by 4 cascaded SOAs. Finally, at the receiver site, the BER of the test channel is measured. In this scenario, the optical input power into the first SOA of the chain in the test channel has been decreased successively with respect to the total power of the 3 rival signals which is −10 dBm. Thus, in all measurements, the rival signals have a constant total first SOA input power representing an optimum operating point.
The power penalty after 4 SOAs compared to the single channel results at a BER of 10−9 is shown in Fig. 12.Due to intra-waveband distortions, namely XGM, the signal performance degrades when the optical input power is reduced. In this case, the test channel power level can be more than 10 dB below the optimum total input power of the rival signals for a power penalty below 2 dB. The inset of Fig. 12 shows the DWDM spectrum in front of the first SOA for the case when the optical power of the test signal is 10 dB below the total rival signal power. Within this context the influence of rival signals in the same waveband has been investigated. The power penalties as well as the Q2 degradation has been measured considering a signal quality requirement for the BER of 10−9 and for the Q2 factor of about 17 dB, respectively. Since this approach is a worst-case analysis, even better results are expected in case of a system using strong forward error correction (FEC) coding, reducing the requirements to a BER of around 10−3.
Fig. 12 Measurement of impact of 3 x 10 Gbit/s OOK rival signals with high power on a single 10 Gbit/s OOK signal within the same waveband (1531 nm, intra-waveband): Resulting power penalty at 10 Gbit/s test channel (C58) compared to the results of the power penalty for the single-channel measurements at 10−9. The inset shows the spectrum in front of the first SOA for the case in which the test signal power is −20 dBm and the total power of the rival signals is 10 dB larger.
4. Conclusions
Within an open converged metro-access infrastructure the simultaneous operation of multiple services operating in separate optical pipes has been successfully demonstrated. In addition, the use of a pipe for different services without any readjustment of optical amplifier gain has been shown. The impact of the SOA chain (1…4 SOAs cascaded) on system performance has been studied. The results confirm the applicability of linearly operating SOAs in future open converged metro-access networks with heterogeneous services. The influence of rival signals with high power in same waveband and fiber has been investigated as well. The results indicate that residual distortions are low enough for using the scheme for a real network application.
Acknowledgments
This work was partly funded by the BMBF project CONDOR (16BP1015) and ATOB (16BP1037).
References and links
1. T. Pfeiffer, “New avenues of revenues - open access and infrastructure virtualization,” in Optical Fibre Conference (OFC) (2012), paper NTh4E. [CrossRef]
2. T. Pfeiffer, “Converged heterogeneous optical metro-access networks,” in European Conference on Optical Communication (ECOC) (2010), paper Tu.5.B.1. [CrossRef]
3. Open access regulation in the digital economy, GSR 2011 discussion paper, ITU, http://www.itu.int/ITU-D/treg/Events/Seminars/GSR/GSR11/documents/03-Broadband%20Policies-E.pdf (2011).
4. OECD, “Broadband networks and open access,” OECD Digital Economy Papers, No. 218 (OECD, 2013).
5. ITU-T Rec. 984.x, IEEE P802.3 ah/av.
6. “Next-Generation PON- Part I – III,” IEEE Commun. Mag. 47, 43–64 (2009).
7. R. Bonk, G. Huber, T. Vallaitis, S. Koenig, R. Schmogrow, D. Hillerkuss, R. Brenot, F. Lelarge, G. H. Duan, S. Sygletos, C. Koos, W. Freude, and J. Leuthold, “Linear semiconductor optical amplifiers for amplification of advanced modulation formats,” Opt. Express 20(9), 9657–9672 (2012). [CrossRef] [PubMed]
8. S. Koenig, R. Bonk, R. Schmogrow, A. Josten, D. Karnick, H. Schmuck, W. Poehlmann, T. Pfeiffer, C. Koos, W. Freude, and J. Leuthold, “Cascade of 4 SOAs with 448 Gbit/s (224 Gbit/s) dual channel dual polarization 16QAM (QPSK) for high-capacity business paths in converged metro-access networks,” in Optical Fibre Conference (OFC) (2013), paper OTh4A.3. [CrossRef]
9. H. Schmuck, R. Bonk, W. Poehlmann, C. Haslach, W. Kuebart, D. Karnick, J. Meyer, D. Fritzsche, E. Weis, J. Becker, W. Freude, and T. Pfeiffer, “Demonstration of SOA-assisted open metro-access infrastructure for heterogeneous services,” in European Conference on Optical Communication (ECOC) (2013), paper We.4.F.2.
10. D. Markert, C. Haslach, G. Luz, G. Fischer, and A. Pascht, “Wideband measurements and linearization of a simplified architecture for analog RF-PWM,” in European Microwave Conference (EuMC) (2013), pp. 1–4.
11. 3rd Generation Partnership Project (3GPP), “Base station (BS) radio transmission and reception (FDD),” http://www.etsi.org/deliver/etsi_ts/125100_125199/125104/09.04.00_60/ts_125104v090400p.pdf (2011)
