Nokia Innovation in single-carrier and multi-carrier coherent optics
With the introduction of higher wavelength capacities and baud rates, some optical networking vendors have begun implementing WDM coherent interfaces using multiple-carriers (i.e. n x subcarriers), instead of traditional single-carrier wavelengths used with previous generations of 100G/200G transponders. This new approach has led to questions regarding any performance benefits or limitations compared to single-carrier wavelengths.
The good news is that both approaches support high-performance coherent interfaces, each operating within a few tenths of a dB of one another. However, understanding how and why single-carrier and multi-carrier wavelengths are implemented, will help explain some of the confusing industry information surrounding both approaches and their relative performance. Finally, there’s an even better, more innovative technique that combines that best of both methods, and it’s been available since 2015.
Single Carrier vs Multi-Carrier
A single-carrier optical interface is exactly as the name implies, a transponder generates a single wavelength that is transmitted out of the line port, as shown in figure 1a. In this example, a single 90 Gbaud wavelength is transmitted over a 100 GHz channel.
A multi-carrier optical interface transmits “n x subcarrier” wavelengths as a combined group, as shown by the figure 1b example using (8) x subcarriers transmitted over a 125 GHz channel. The n x subcarriers are generated digitally, within the DSP, so both approaches only require a single Tx laser and single Tx modulator. However, since there is a limit on how close wavelengths can be spaced together, while avoiding interference between subcarriers, small gaps of unused spectrum result when using multi-carriers.
Figure 1a) Single-Carrier and 1b) Multi-Carrier Optics
CD Compensation Effects
The introduction of coherent digital signal processors (DSP) and coherent optics has led to tremendous increases in wavelength speeds and network capacities, while significantly lowering transport costs per bit ($/Gb). In addition, coherent DSPs allow for digital compensation of network impairments, including chromatic dispersion (CD) and polarization mode dispersion (PMD).
Using the DSP for chromatic dispersion (CD) compensation provides enormous benefits over previous methods used on older generations of 10G based WDM systems. However, digital CD compensation comes at the cost of significant DSP resources and power consumption within the ASIC. Within the DSP, chromatic dispersion is compensated using digital filters (FIR), essentially a series of mathematical calculations that reverse and remove the chromatic dispersion effects. Unfortunately, very large sized digital filters are needed to remove the chromatic dispersion effects of thousands of kilometers of fiber, requiring large number of gates (die area) and high-power consumption within the DSP.
As the industry transitioned to higher wavelength capacities and baud rates, from 33 Gbaud to 60 Gbaud to +90 Gbaud, the problem of CD compensation filter size and power consumption has become ever more problematic. Chromatic dispersion increases with the square of the baud rate, so as baud rates increase the amount of chromatic dispersion generated grows enormously, requiring very large CD compensation filter sizes and power consumption. Multi-carriers offer a solution for implementing smaller sized and lower power CD compensation filters within the DSP.
With multi-carrier wavelengths, each subcarrier operates at a fraction (1/n) of the overall baud rate. A 96 Gbaud signal with eight subcarriers results in each subcarrier operating at only 12 Gbaud. Using a bicycle rider theme shown in figure 2, chromatic dispersion causes the individual bicyclists (subcarriers) to spread as they travel down a fiber. However, the lower subcarrier (bicyclist) baud rates, enable the use of smaller sized and lower power CD compensation filters within the DSP. This multi-carrier technique requires parallel compensation filters in the DSP, one for each bicyclist, but results in lower overall power consumption, which is a primary reason many vendors are adopting multi-carrier coherent optics.
Figure 2) Multi-Carrier CD Compensation
Multi-Carrier Challenges
Using multi-carrier wavelengths reduces the size and power consumption of DSP compensation filters, however this method comes with its own set of “challenges”. There is a limit of how tightly subcarriers can be spaced together, while still avoiding interference between each of the individual subcarrier wavelengths. The small gaps of unused spectrum between each subcarrier can result in slightly lower spectral efficiency compared to single-carrier wavelengths.
On the DSP receive (Rcv) side, additional complexity is needed to support parallel DSP blocks to receive, recover, and filter the n x subcarriers in a multi-carrier wavelength. In addition, any skew that develops between the subcarriers results in additional performance penalties. With multi-carrier wavelengths, there are trade-offs between the benefits of smaller sized, lower power CD compensation filters with the added DSP complexity, slightly lower spectral efficiency, and potential skew and implementation penalties caused by multiple subcarriers.
Nonlinear Performance Impacts
There has been some conflicting and confusing industry information regarding nonlinear performance differences between single-carrier and multi-carrier optics. In reality, many of the performance penalties that are labelled as “nonlinear”, actually result from component tolerances and variations that would be better categorized as “implementation penalties”.
Traditional optical networking nonlinear effects include, self-phase modulation (SPM), cross-phase modulation (XPM), and four-wave mixing (FWM). Self-phase modulation effects are due to changes in optical power levels (Kerr effect), while both XPM and FWM are related to interactions between closely spaced wavelengths. Since XPM/FWM are related to how close wavelengths are spaced together, a multi-carrier signal using closely spaced “n x subcarriers” should, at least qualitatively, incur higher XPM/FWM penalties.
Implementation penalties result from designing coherent optics with real world components, each with their own tolerances and variances. Transmit and receive lasers have specified line widths, drivers aren’t perfectly linear, modulators have imperfect isolation and phase, clock recovery is imperfect, and so forth. These component limitations result in many tenths of dB “penalties”, compared to an ideal solution. These penalties are typically just lumped into the nonlinear penalty “bucket”, confusing the issue of which effects are caused by component issues, such as laser line widths, and which are due to traditional SPM, XPM, or FWM nonlinear effects.
In particular, laser line widths have a strong correlation and impact on coherent optics performance. Coherent optics rely on phase-modulated signals, which use receive lasers as local oscillators to recover the original transmitted signal. High performance lasers used in modern coherent optical systems have line widths ranging from 100 KHz to 300 KHz wide. The fixed (i.e. non-zero) laser line widths result in some laser drift over time, referred to as “laser phase noise”. Since these lasers are used in the coherent receiver to recover the transmitted (phase-modulated) signal, any laser phase noise can result in additional errors and penalties.
Multi-carrier wavelengths enable smaller sized CD compensation filters in the DSP, resulting in fewer DSP gates and lower DSP power consumption. In addition, the small CD filters also benefit from having shorter “time spans”, essentially less time is needed to pass through the smaller sized digital filters. With shorter times, the receive laser has less time to “drift”, resulting in less impact due to laser phase noise.
It’s not that multi-carriers are somehow more immune to nonlinear (SPM, XPM, FWM) effects, but rather the multi-carriers are the mechanism that enables smaller-sized CD compensation filters in the Rcv DSP. It’s the smaller sized CD filters that are impacted less by laser phase noise caused by imperfect, real world lasers. Although frequently lumped together with nonlinear penalties, these types of component tolerance effects would be better categorized as “implementation penalties”, since the underlying cause is imperfect laser line widths. Nonlinear XPM/FWM nonlinear effects are likely to be a bit higher with multi-carrier wavelengths, due to the closely spaced subcarriers.
Fortunately, there is a better, more innovative technique that combines single-carriers and small-sized CD compensation filters, providing the best of both worlds.
Taking Light to the Limit
Figure 3) Nokia PSE sub-band CD compensation
Sub-band compensation relies on a single-carrier wavelength on the transmit side, as shown in figure 3 using the same bicyclist theme. Chromatic dispersion causes the single bicyclist to spread as it travels down the fiber. The receive DSP digitally slices the single-carrier signal into many, many sub-bands and performs CD compensation across each, individual sub-ban using very small-sized CD compensation filters. Nokia’s Photonic Service Engine (PSE) gains the benefits of small sized and low power CD compensation filters, along with reduction in nonlinear (i.e. implementation) penalties due to imperfect laser line widths, but without the 1) “gaps” of unused spectrum, 2) higher receive DSP complexity, or 3) increased XPM/FWM found on multi-carrier implementations.
Nokia - innovative coherent optic solutions
