Meeting the demand for optical bandwidth over the next decade
As I visit our customers across the globe, I often get asked the question: what is next for scaling optical transmission? Given that the capabilities of the latest generation of networking solutions are now achieving performance near the nonlinear Shannon limit, this is a great question.
In the past two decades, we have witnessed a more than one-hundred-fold increase in transport capacity over optical fiber. That works out to an average annual growth rate of 40%. If projected forward, this scaling trend, akin to Moore’s law for silicon integrated circuits, would mean that 25 years from now, optical transmission capacity would need to scale to over 40 petabits per second per fiber — a tall order!
While optical fibers are commonly celebrated for their almost infinite capacity, they are still limited by physics. The physical limits for transmitting data over a transmission medium are represented by the nonlinear Shannon equation, C ~ M x B x P x log2(1 + SNR), where M= the number of spatial paths, B = bandwidth, P = the number of polarization states used, and SNR, the signal-to-noise ratio needed for successful transmission of the data channel.
Given the world’s insatiable and constantly growing demand for bandwidth, the fact that performance in some areas has now reached the Shannon Limit means that new approaches are required to continue increasing optical fiber transmission capacity. This blog is the first in a series where I will explore possible technology innovations needed for optical fiber capacity to keep up with future demand growth.
The vectors of scale for optical fiber capacity
Before we look at how we can scale optical fiber capacity another hundred-fold, we must understand what attributes define capacity, and what factors can be influenced to increase them. For optical fiber transmission, there are three vectors that define the maximum capacity that can be transmitted over an optical fiber, as depicted in Figure 1.
As described by the Shannon equation, these are defined by the Spectral Efficiency (SE) of a single channel, and the number of multiple data channels that can be transmitted in parallel, either across the WDM spectrum of a single fiber core (B), or across multiple fibers, or fiber cores (M).
Figure 1. The three vectors of technology that define Shannon’s Limit for increasing optical fiber capacity
The options for further increases in the capacity that can be carried over an optical fiber are both promising and complex. Enabling innovation to achieve this requires a careful understanding and engineering of the optical transceivers as well as the fiber, its propagation characteristics, how it can carry more wavelengths, manage dispersion, and its structure and manufacturing.
Spectral efficiency (SE) defines how many bits per second can be transmitted within a given amount of spectrum used by a single data channel. It can be scaled either by increasing the number of bits transmitted, reducing the amount of spectrum needed, or both. The earliest optical fiber transmission systems used simple on-off keying (OOK) modulation schemes to code binary information into 1’s and 0’s of light pulses, with typical data channels operating at 2.5Gb/s and using approximately 50GHz of optical fiber spectrum.
Increasing capacity relied on increasing the bit rate, and this became economically limited beyond 10Gb/s, and technologically limited after about 40Gb/s. Thus, innovation was needed to enable further increases in capacity, leading to a transition to coherent modulation, which encoded binary information into both phase and amplitude.
With coherent transmission, the innovation engine that drove continuous increases in optical fiber capacity was the digital signal processor (DSP), which implemented the complex algorithmic coding needed to recover the original information — the data bits — buried in the modulated wave form.
Rapid progress in scaling SE was achieved over the first few generations of coherent optics, by increasing the baud rate and modulation order. Leveraging new silicon node geometries enabled by the progress of Moore’s Law, coherent baud rates increased from 30GBaud to over 60GBaud in only three generations. In parallel, coherent modulation scaled from initial Quadrature Phase Shift Keying (QPSK) enabling the transmission of 2 bits per symbol to 16QAM, able to transmit 4 bits per symbol. After a decade of innovation, coherent modems today are capable of high order modulation such as 64QAM transmitting 6 bits per symbol.
Beyond baud rate and modulation order, new innovations in coherent modem algorithms such as Probabilistic Constellation Shaping (PCS), further refinements in Forward Error Correction (FEC), and nonlinearity compensation have combined to increase SE tenfold over the first coherent systems. The latest industry-leading coherent optics, such as Nokia’s recently announced PSE-6s, enable wavelength speeds of 1.2 Terabits per second.
All these improvements have enabled increases in SE while simultaneously reducing the number of optical transponders that need to be deployed for a given network capacity. With the introduction of 67GBaud coherent optics, however, increases in SE have been limited, with little to no improvement foreseen.
Moving forward, the two variables of the Shannon Limit that can be manipulated further are M, the spatial component primarily defined by fiber parallelism, and B, by expanding the available amplified spectrum. Thus, further increases in total fiber capacity require ongoing innovation across multiple vectors of scaling.
In the next blog in this series, I will explore just that; looking at how the parallelism of wavelength division multiplexing (WDM) and recent innovations aimed at increasing spectrum use are providing continued increases in total fiber capacity, and how future technologies will continue this trend.
Visit our Scale made simple webpage to learn more!