Have It Your Way
July 12, 2022
By Paul Momtahan
Director, Solution Marketing
The Power of Coherent Optical Engine Programmability
Infinera’s ICE6 optical engine is being adopted across a broad range of optical networking segments, by communications service providers (CSPs), wholesale long-haul operators, internet content providers (ICPs), research and education (R&E) network operators, and on all types of submarine networks (dispersion managed, uncompensated large area, space-division multiplexing, etc.). You can read more about it in a previous blog by my colleague Geoff Bennett. A large part of this application breadth is down to the flexibility enabled by ICE6’s programmability.
Figure 1: ICE6 programmability currently supported on Infinera GX G42
ICE6 supports baud rate tuneability from 31 to 96 Gbaud, and to just over 100 Gbaud with ICE6 Turbo. Bits per symbol range from 2 to 12, with high granularity enabled by probabilistic constellation shaping (PCS). In addition to PCS-64QAM, ICE6 also supports conventional modulation (64QAM, 32QAM, 16QAM, 8QAM, QPSK). And, while its second-generation Nyquist subcarrier capability is one of the key enablers of ICE6’s superior performance, (frequency-domain) hybrid modulation further leverages this capability, putting different conventional modulation formats on different subcarriers. For example, 64/32QAM puts 32QAM on the outer four subcarriers and 64QAM on the inner four subcarriers to get 11 bits per symbol, as shown in Figure 2. This provides some additional granularity beyond conventional modulation, without using PCS.
Figure 2: Frequency-domain hybrid modulation example: 64/32QAM with 11 bits per symbol
ICE6 also supports specialist multi-dimensional modulation, including 4D (i.e., ME-8QAM) and 8D formats (i.e., FD-eBPSK, FD-2.5QAM, FD-3QAM). Although four and eight dimensions sound like something from science fiction or advanced quantum physics, these “dimensions” are actually more straightforward than they sound. Four dimensions is phase and amplitude (or in-phase carrier and quadrature carrier/I and Q in a constellation diagram) on one polarization plus phase and amplitude (or I and Q) on the other polarization. With eight dimensions, we add frequency in the form of paired subcarriers.
Figure 3: ME-8QAM balances a higher-power outer constellation point on one polarization with a lower-power inner constellation point on the other polarization
What makes these formats multi-dimensional is a set of rules for how symbols can be combined across these dimensions to maximize performance in specific scenarios, such as dispersion-managed subsea fibers.
Combined with a variety of forward error correction (FEC) settings, FEC overhead is configurable between 20% and 33%. Framing modes include an overhead-efficient Ethernet-only mode and a flexible “mixed” mode that can support both Ethernet and OTN client types. ICE6 also supports specialized submarine modes that tweak multiple parameters for optimized performance. In total, the ICE6-based CHM6 sled for the GX G42 currently supports more than 400 modes to ensure optimal configurations to meet network operator objectives for virtually any scenario. On top of all these modes, additional parameters that can be configured include dynamic bandwidth allocation (DBA), PCS distribution (Gaussian or super-Gaussian), state-of-polarization (SOP) tolerance (normal or high), non-linear compensation, encryption, SD-FEC gain sharing, center frequency, and transmit power, as shown in the bottom half of Figure 1.
But why does any of this matter? Aren’t a handful of the highest-performance modes, together with center frequency tuning, sufficient? Well, in a world where every network had an ideal set of conditions and every operator wanted to prioritize the same things, maybe we would only need a handful of modes. But in the real world, we need the flexibility to choose from the many configurations ICE6 supports. Here is my top five list of ICE6’s programmability benefits for service providers and their networks:
1. Optimize What Matters to You
As discussed in a previous blog (Think Spectral Efficiency and Wavelength Capacity-reach are the Same Thing? Think Again.), spectral efficiency and wavelength capacity-reach are not the same thing and no longer evolve in tandem. For some scenarios and applications, it will make more sense to prioritize spectral efficiency and the resulting fiber capacity benefits. For other scenarios and applications, it will make more sense to prioritize wavelength capacity-reach and the resulting benefits in terms of cost per bit, power consumption, and footprint. Power consumption can be further optimized by turning down (e.g., chromatic dispersion compensation) or off (i.e., non-linear compensation) power-consuming features that are not required. Latency is another parameter that can be optimized for with certain modulations. However, in practice, it will often be a case of optimally balancing these metrics by selecting the appropriate configurations.
2. Address Challenging Conditions
Challenging conditions might include high-loss fiber spans, G.655 LEAF fiber, poor fiber quality, high ROADM cascades, and aerial fiber in lightning-prone geographies. The wide range of modes, including lower baud rate options, specialized 4D and 8D formats with low bits per symbol, and 33% FEC, enable ICE6 to maximize performance even under these challenging conditions. Additional ICE6 configurable parameters that address challenging conditions include DBA for high ROADM cascades, super-Gaussian PCS for high-wavelength power scenarios (e.g., uncompensated large-effective-area subsea fibers), and the high SOP/lightning tolerance setting for aerial fibers.
3. Squeeze Out Every Last Gb/s
The ability to select from a wide range of possible configurations enables network operators to squeeze the maximum performance from their fiber, monetizing any unrequired margin. Even on the same subsea fiber or terrestrial fiber path, performance in different parts of the spectrum varies with tilt, ripple, dispersion, and nonlinearities. ICE6 enables operators to select the best mode for each part of the spectrum. Turning on SD-FEC gain sharing can also help, where two wavelengths operate in parts of the spectrum with different performance.
4. Choose Your Wavelength Spectrum
One key benefit of baud rate flexibility is the ability to tune the spectrum consumed by the wavelength. For maximum wavelength capacity-reach, the maximum baud rate is usually best. However, as discussed in a previous blog (Coherent Baud Rates: Is Higher Always Better?), there are a number of scenarios where a lower baud rate might be required. For example, legacy filters and wavelength-selective switches (WSSs) will typically have a passband significantly smaller than the grid spacing might suggest.
A lower baud rate may also align better with operationally simplified flexible-grid granularity. As examples, considering factors like roll-off and filter narrowing, 84 Gbaud might be required for 100 GHz of spectrum, while a 96 Gbaud wavelength would require 112.5 GHz or more. For some operators, managing spectrum in 100 GHz increments may have operational simplicity advantages. And finally, with trans-Pacific fibers of up to 14,000 km, the additional chromatic dispersion effects of ultra-high baud rates might require a lower baud rate.
5. Maximize Network Availability
One final benefit of ICE6’s programmability relates to network availability. After a fiber cut or other failure, a longer alternative path may have higher impairments than the original working path. If the original mode did not work on this new path, a lower-performance mode would provide a better alternative to failure and the resulting loss of capacity. Another example of this is fiber degradation, which occurs over time due to aging and fiber cut repairs. As the fiber degrades over time, ICE6 can be reconfigured for lower capacity modes rather than waiting for the wavelength to eventually fail.
So, to summarize, ICE6 now supports over 400 modes based on baud rate, modulation, FEC, and framing, plus multiple additional configurable parameters, as shown in Figure 1. This enables network operators to select the optimal configuration to address the specific network and path conditions and their specific priorities, with the flexibility to adapt to future changes in the network.