NGPON2

NGPON2 is a terminology describing what might the next optical access networks look like with a possibility to have afrom scratch approach not considering any existing systems.

The aim of these architectures is to achieve bit rates of 40Gbit/s minimum in the downstream, where the target is to have 1Gbit/s sustained per user. In the upstream the goal is also to provide 1Gbit/s connection per customer.

Therefore several concepts are currently discussed¹¹: OFDMA-PON, WDM-PON, and WDM-TDM-PON. These concepts can set the bounds for the fixe/mobile infrastructure convergence.

OFDMA-PON

Technique: The electrical signal which modulates the optical source is modulated (either directly or indirectly) according to an adaptive OFDM-scheme in the electrical domain.

Once the electrical frequency response of the channel is estimated, the bandwidth is sliced into sub-channels. According to the SNR of each sub-channel, the sub-carrier’s modulation scheme is individually optimized: the higher the SNR of each sub-channel is, the more complex the modulation scheme is (from 4 to 32-QAM in [43]).

In downstream, this technique allows to use the full bandwidth of the transmission system.

In the upstream, the electrical spectrum is shared among the users, and user emits a portion of the electrical spectrum. If a single wavelength is used by all customers (which is the simplest implementation), the different (and not overlapping portions of electrical spectra) are recombined at the CO. However due to the drift of the lasers’ central wavelengths, frequency guard bands have to be introduced.

11meaning not standardized yet

Figure 2.19.: Wavelength allocations of IEEE GE-PON & 10G-EPON and ITU-T G-PON &

XG-PON (Picture from [40])

Benefit: The main benefit of this technique is the increased spectral efficiency (from 1/bit/s/Hz for NRZ to 4-5bits/s/Hz for OFDMA [43]), which allows to increase the bit rate while using standard components with limited bandwidth (since designed for NRZ-applications in Intensity Modulation and Direct Detection (IM-DD) systems).

For instance [43] transmitted a 10Gbit/s OFDMA-signal¹² using a DFB laser with a RF frequency bandwidth of 2.1GHz (at -3dB, Fig. 2.20) over a 26dB optical budget including 110km of S-SMF.

Figure 2.20.: Frequency response of the laser used by [43] for OFDMA modulation

When using standard components designed for 10Gbit/s NRZ applications, then thanks to the electrical OFDM modulation scheme, bit rates of 40Gbit/s are achievable, and this over optical link with an optical budget of 20dB including 100km of fiber. Also 255 users are served per wavelength.

Reach extension The reach or the splitting ratio of a PON-OFDMA can be increased by using a SOA or an Erbium Doped Fiber Amplifier (EDFA) (Fig. 2.21). This was demonstrated by [44] for 10Gbit/s OFDM signals.

Figure 2.21.: Total optical budget performances (direct modulation) comparison with NRZ or OFDM signals and with and without SOA or EDFA, (picture from [44])

Practical implementations issues: Yet, the afore reported bit rates are off-line per- formances. The real-time processing results show lower bit rates since the Fast Fourier Transform (FFT) and Inverse FFT (IFFT) operations are very computing intensive, and depend upon the number of processed sub-carriers.

WDM-PON

In a WDM-PON several wavelengths are used in the downstream (DS) as well as in upstream (US) [45], typically 32 to 40 per propagation direction are aimed (Fig. 2.22).

Figure 2.22.: Basic WDM-PON architectures: a) broadcast-and-select (BS) WDM-PON with splitter/combiner in passive node; b) AWG-based, (picture from [45])

Basic WDM-PON architectures , as shown in Fig. 2.22, make use of a fixed-wavelength laser array or a multi-frequency laser (MFL).

For the broadcast-and-select architecture shown in Fig. 2.22a, the OLT broadcasts all wavelengths in the DS through a passive 1:N splitter. Each ONU selects one of the DS wavelengths using an individual filter, and uses another individual wavelength for the US.

As in the downstream, the US wavelengths are passively combined in the 1:N coupler. No identical ONUs can be used unless both the receiver filters and transmitters are tunable.

The latter aspect is the weakest point of such an architecture [45].

In Fig. 2.22b, an AWG-based (arrayed waveguide grating) wavelength-routing PON is shown. Here, the AWG wavelength router replaces the passive splitter/combiner.

This scheme offers lower insertion loss¹³ In addition, no wavelength selective (individual) receivers (Rx) are necessary, thus simplifying the ONUs. However the transmitters (Tx) at the ONUs still have to emit at different wavelengths or to be at least tunable.

Whether a 200 or 100GHz array wave grating filter is used for separating the wavelengths, respectively 16 or 32 users can be served. The current state-of-the-art consists in 1.25Gbit/s per wavelength, with one wavelength per user. The typical reach is 20km for an optical budget of 10dB.

Colorless ONUs can resolve the issue of having different ONUs (not emitting at the same wavelength) as well as having ONUs with tunable laser sources.

The combination of a single fiber network and colorless, seed-based ONUs leads to reflective SOAs (RSOAs). Here, one end-facet of the SOA is fully reflective so that input and amplified and modulated output are both coupled to a single fiber at the other end.

Alternatives to RSOAs are reflective electro-absorption modulators (REAM).

A block diagram of an PON with RSOA-based colorless ONUs is shown in Fig. 2.23:

dedicated seed wavelengths from a multi-frequency laser (MFL) are used as input to the reflective ONU transmitters. The seed signals are not modulated in the OLT and can easily be separated in the ONUs by simple WDM splitters. Since dedicated wavelengths for the seeds are used, the number of ONUs is limited. In Fig. 2.23, transmit and receive signals in the OLT are also amplified (A) for higher maximum reach.

Figure 2.23.: Colorless ONUs for single-fiber working based on RSOAs, REAMs, or (Injection- Locked Fabry Perot Lasers) IL-FP lasers: different seed and DS payload wavelengths, (picture from [45])

WDM-PONs can also run at different per wavelength bit rates for DS and US, i.e., 1G, 2.5G, 4G, or 10G, respectively. This provides scalability for both splitting ratio and per-ONU bit rate.

Comparing WDM-PON and 10GPON: according to [46] the 10GPON has more ad- vantages in terms of standardization and maturity, costs and energy consumption than the WDM-PON. Yet the latter offers higher bandwidths and longer reaches.

For these reasons, a tendency according to [46] would be to use 10GPON for residential customers, whereas the WDM-PON could be envisaged for business and and backhaul applications where higher bandwidths are needed.

WDM-TDM-PON

The idea is reproduce the way a legacy PON works (TDM/TDMA) but for several wavelengths. This results in a wavelength pool shared (statically or dynamically) among all the users of the PON instead of allocating a couple of wavelengths to each ONU [4].

Current state-of-the-art is to have a 2.5Gbit/s bit stream per wavelength. 32 of such channels are aggregated by a 100GHz AWG. Finally this allows to serve 1024 customers for a reach of up to 60km and for a maximal optical budget of 20dB.

Conclusion

We have shown the main architectural variations of current optical access networks.

More specifically the performances (bit rates and optical budgets) and main properties (wavelengths, multiplexing protocol) of PON networks have been discussed. This allows to set the framework within which the discussed (in Part 2) and the proposed (in Part 4) Radio over Fiber (RoF) architectures will have to evolve.

Also the optical properties of the components (emitters, power splitter, optical fiber, and receivers) building the optical link of PONs have been presented and discussed.

Finally the properties of the optical access networks of the close-future and more long-term have been summarized.