The Top 5 CapEx Considerations When Choosing Coherent Transceivers
Digital ASIC/DSPs based on 7-nm CMOS technology and advanced photonic integration based on indium phosphide or silicon photonics are enabling a wide range of new coherent transceiver types, including 100ZR and 400ZR pluggables, OpenZR+ 400G pluggables, Open ROADM 400G ZR+ pluggables, XR optics, and embedded 800G. Selecting the optimal engine for a given application requires the careful consideration of a wide range of factors, as shown in Figure 1.
Figure 1: Coherent transceiver selection considerations
In this blog, the first in a multi-part series, I will examine the top five considerations related to CapEx.
Transceiver Cost/Cost Per Bit
A key consideration is the transceiver cost per bit for a given reach requirement, which is primarily a function of the transceiver’s wavelength capacity-reach: the maximum data rate that the transceiver can achieve for a given path through the optical network. If less than the full capacity is required, then the cost per bit needs to consider the required capacity rather than the maximum capacity. If the full capacity will only be required in five years, then a solution that enables CapEx to be more closely correlated with the actual required capacity, for example XR optics with 25 Gb/s increments, has an advantage.
Cost per bit is also a function of the unit cost of the optical transceiver. Factors that influence this include the cost of the individual components, packaging, and manufacturing. These costs will in turn be impacted by volumes, and the cost to the network operator will also be heavily influenced by competition, both direct (i.e., the same type of transceiver) and indirect (i.e., different types of transceiver), with more suppliers driving more competition. which typically reduces unit prices.
Cost of Xponders, Xponder Shelves, and Grey Interconnect Pluggables
In addition to the cost of the transceiver, the cost of any Xponders (transponders, muxponders, switchponders, etc.), the shelves that house them, and the grey interconnect pluggables on the client side of the Xponder and in the router should also be considered. Plugging the coherent transceiver directly into the router can eliminate the Xponder, Xponder shelf, and grey interconnect pluggable costs. If high-capacity wavelengths require statistical multiplexing or switching for efficient utilization, then this additional cost also needs to be considered.
Optical Line System CapEx
In brownfield scenarios, compatibility with the existing optical line system needs to be considered. As the spectral width of the wavelength is primarily a function of its baud rate, higher-baud-rate wavelengths may be incompatible with the existing DWDM grid. Even 100 GHz grid systems based on older filter and wavelength selective switching (WSS) technology will have a passband (~50 GHz) too narrow to support a 400 Gb/s wavelength with a baud rate of 60+ Gbaud. Other considerations include transmit power compatibility and out-of-band noise if colorless add/drop is a requirement, with smaller QSFP-DD-based pluggables typically facing some challenges in this regard as they lack the space for a micro-EDFA or a tuneable optical filter, which help with transmit power and out-of-band noise, respectively. Another brownfield consideration is whether the new wavelengths will interfere with the existing wavelengths, thus requiring guard bands or reducing the performance of existing wavelengths.
For greenfield scenarios, which will be the case if you do not already have flexible grid (or wide-passband fixed grid) optical line systems and wish to leverage 400G+ coherent technologies, a particular coherent transceiver may enable a more cost-effective optical line system, for example, a filterless broadcast one based on splitter/combiners or one with a reduced need for amplification. Higher-capacity wavelengths can also reduce the number of ROADM add/drop ports, thus reducing line system CapEx. Conversely, any extra line system costs incurred by a specific optical transceiver also need to be considered – for example, if a more expensive optical line system is required to compensate for any deficiencies in the coherent transceiver such as low transmit power or high out-of-band noise.
Fiber Costs: Spectral Efficiency and Fiber Capacity
The cost of the fiber itself is an important consideration, especially for long-haul, submarine, and fiber-constrained metros where the cost of acquiring and lighting new fibers is high. In these scenarios, spectral efficiency and fiber capacity can become key transceiver considerations. Spectral efficiency is largely a function of how many bits per symbol the modulation can deliver for a given reach requirement. A secondary consideration is how tightly you can pack the wavelengths together, which in turn is related to the shape of the wavelength (i.e., the percentage roll-off).
Figure 2: A wavelength with tight roll-off uses less spectrum
For example, a 400 Gb/s wavelength (~60 Gbaud, PM-16QAM) with no Nyquist shaping (i.e., 400ZR) uses more spectrum than an equivalent wavelength (~60 Gbaud, PM-16QAM) that uses Nyquist shaping and has a tight roll-off, as shown in Figure 2. With no Nyquist shaping and a relatively large roll-off, anyone deploying 400ZR has to choose between a 100 GHz grid with better performance but lower fiber capacity or a 75 GHz grid with higher fiber capacity but reduced reach due to inter-channel interference (ICI). Even Open ROADM CFP2s at 63.1 Gbaud typically require 87.5 GHz or more per channel in a mesh ROADM network. Another factor is how much correlation there is between the movement/drift of each wavelength, with shared wavelocker technology that enables multiple wavelengths to drift in unison, enabling better spectral efficiency.
Fiber capacity also needs to consider the amount of spectrum that can be used on the fiber – for example whether a particular transceiver type can support an extended C-band or the L-band (i.e., C+L). Embedded optical engines are more likely to support the L-band, though L-band coherent pluggables are also possible. A related consideration is the amount of wasted spectrum due to wavelength blocking. This can be an issue when mixing wavelengths with different baud rates on the same optical line system, which complicates channel plans, especially in mesh ROADM networks.
Router CapEx also needs to be considered. If the optical transceiver technology forces the purchase of new routers with the required port form factor, power, and thermals to support it, that can increase router CapEx. If the transceiver form factor (i.e., CFP2) or data rate (200 Gb/s for extended reach instead of 400 Gb/s) reduces router or line card efficiency in terms of faceplate density and/or throughput, that may also increase router CapEx.
On the other hand, a pluggable form factor and power/thermal envelope that is compatible with existing routers can avoid router upgrade costs. Router CapEx may also be reduced if the coherent transceiver enables more cost-effective router form factors (i.e., high-density QSFP-DD only) or the elimination of intermediate switch/router aggregation layers. Another factor to consider is load balancing efficiency, due to well-known hashing algorithm limitations in load-balancing mechanisms such as link aggregation (LAG) and equal cost multi-path (ECMP); a smaller number of high-capacity wavelengths will typically be more efficient than a large number of lower-speed wavelengths.
So, to summarize, if you want to minimize CapEx, you should consider the costs of the transceivers themselves but also any additional costs or savings related to the Xponders, grey interconnect pluggables, optical line system, fiber, and routers. In the next blog in this series, I will move to the key OpEx considerations for next-generation coherent transceiver selection.