Reconfigurable Optical Networks with Self-Tunable Transceivers: Implementation Options and Control Michael H. Eiselt ADVA SE, Maerzenquelle 1-3, 98617 Meiningen, Germany meiselt@adva.com Abstract This paper reviews methods for autonomous tuning of optical transceivers, based on an overhead management channel between the modules on both sides of the link. Different implementation options for the tuning principle, as well as for the tunable laser are introduced. Introduction The increasing deployment of 5G mobile stations also places a challenge on the optical network supporting the features of 5G. A section of the optical network with a particularly rapid growth is the fronthaul between the central office and the radio unit. Depending on the functional split, data rates of 10 Gbit/s and above need to be transported to each of typically three antenna sectors. With a high antenna density, this part of the network is highly cost-sensitive and requires an implementation supporting simple operation. These requirements can be supported by autonomously tunable transceivers. ITU-T Recommendation G.698.4 [1] (ex: G.metro) has been developed to allow interoperability between transceivers of different vendors. Since publication of this standard, several solutions have been marketed by different vendors, all of them only partly implementing the specifications of the Recommendation. This paper will first describe the system and network requirements and discuss implementation options of tunable lasers as well as two methods to tune the laser to the target frequency. Network architecture A typical architecture of the fronthaul network connects the central office (CO) with a number of remote radio units (RU) by point-to-point links. While, in a fiber-rich environment, the connections can run on individual fiber pairs, the increasing number of radio units requires sharing of the transmission fiber by multiple connections, using wavelength division multiplexing (WDM). The fiber layout can form a tree structure with a trunk fiber connecting the CO to a remote node, to which all RUs are connected via drop fibers. As the connections between CO and RU use individual wavelengths, the remote node can include a wavelength demultiplexer to route the wavelength channels. Alternatively, the fiber can be deployed in a drop line structure, passing the radio units and dropping/adding individual wavelengths or sets of wavelengths to/from each radio unit. The drop line structure can be expanded to a “horseshoe” type network by adding a second CO to the end of the line, such that each radio unit can be connected to two central offices, using the same wavelength on separate sections of the ring. This provides redundant connectivity in case of a fiber interruption. In any case, it is important to note that there is no direct connectivity between different radio units, but only between the central office and each radio unit. Wavelength agnostic transceivers To keep the operational complexity of the front haul system low, tunability of the transceivers at least at the radio units is required. At the same time, however, the cost of these transceivers, which are expected to be deloyed in large numbers, must be kept low. This resulted in the attempt to move control functionality from the individual RU transceivers to a shared function in the central office. One important aspect is here the tuning of the transmitter to a target frequency, used for the RU-to-CO transmission, and to maintain a stable frequency. The target tuning of the transmitter can be performed in two ways. Either, (1) the transceiver is characterized, before shipping to the customer, generating a calibration table with tuning parameters for all operating frequencies or, (2) upon turn-on, the transmitter sweeps over the operating frequency range and expects a feedback from the receiver if the correct frequency has been achieved. After tuning, a stable transmission frequency needs to be maintained. While ITU-T standards do not specify a parameter for frequency stability, the ”maximum spectral excursion” [2] limits the modulation bandwidth (including chirp) as well as the central frequency deviation. With a low- chirp modulation, a maximum spectral excursion of 12.5 GHz, as specified for many DWDM applications, leaves a central frequency tolerance of approximately 7.5 GHz for the tunable laser.