The driver for any development in optical communications technology is, almost without exception, a reduction in cost-per-bit/km travelled (although there’s one interesting exception to this which I’ll talk about at some future point). This intensely capitalistic and utilitarian approach has enabled the dramatic growth in the internet (YouTube needs very low cost/bit to be viable), but are we approaching the point where the rate of cost reduction may start to slow down? In the US, for example, there’s recently been a move away from ‘all you can eat’ data plans to something approaching (very slowly) the capped plans found in many other parts of the world – partially in response to the recognition that there is a finite amount of bandwidth available at any point in the network.
In the optical space, we have, over the past 15 years, moved to higher and higher per-channel bit rates running on more-closely-spaced WDM channels. However, we are (as I mentioned previously) reaching the limit of what can be achieved on the ITU Grid-based systems introduced in the 1990s. Consequently, fibre bandwidth, which only a few years ago was being proselytised by Gilder as being effectively ‘infinite’, is increasingly being seen as a finite resource which needs to be managed as effectively as possible using all the techniques at our disposal.
One solution, which is gaining a great deal of traction at the moment is to move away from the constraints of the ITU Grid to what we term a flexible grid architecture where the channel optical bandwidth can be dynamically adjusted to meet the requirements of the signal being sent through it and hence maximise the data carrying capacity of the fibre.
This ability to control the channel bandwidth and position with GHz resolution has been utilized in many research papers as I discussed in my previous blog. In practice, however, the complexity of managing a network with such fine granularity may outweigh its advantages. Market requirements indicate that a channel bandwidth granularity of 12.5 GHz will meet future channel bandwidth requirements and that even 25 GHz channel bandwidth increments may be sufficient.
First generation WSS (typically based on MEMS and/or Liquid Crystal technologies) allocate a single switching element (pixel) to each channel which means that the channel bandwidth and centre frequency are fixed at the time of manufacture and cannot be changed in service. However, second generation WSS, based on Liquid Crystal on Silicon (LCoS) or 2D MEMS mega-pixel matrix switching arrays, permit dynamic control of channel centre frequency and bandwidth through ‘on the fly’ modification of internal pixel arrays via embedded software.
Furthermore, not only must the core switching elements in a ROADM be capable of supporting flexible grid architectures, but the multiplexer/demultiplexers and filter arrays must support the same degree of flexibility. The flexibility provided by LCoS technology can also be applied to these high-port-count (e.g. 1×23) multiplexer/demultiplexers and programmable filter arrays.
This discussion has focused on the requirements for the wavelength-selective elements in a flexible-grid ROADM as these will be the first part of the flexible grid network that has to be deployed to ensure the network is future-proofed. However flexible grid networks will also require additional component developments including scanning optical channel monitors capable of handling polarization multiplexed signals with varying bandwidths and signal formats and signal (and local oscillator, for coherent systems) lasers capable of operating at the finer frequency increments implied by flexible grid architectures (6.25 GHz for 12.5 GHz channel increments and 12.5 GHz for 25 GHz channel increments). This provides a continuing challenge to those of us in the optical space as we continue to improve the price/performance of our components and modules which underpin the networks of the future.
Flexible grid technology (not grid-less) will allow the optimum usage of the finite operating window in a fibre and allow operators to continue down the path of reducing the cost per bit/km of data travelling through the network.
Any comments are welcome.