Ever increasing data demands push optical networks to transport data at faster and faster rates. Modulation and baud rate affect the distances that optical signals can travel.
The mission of optical networks is to transport astronomically high volumes of service traffic over wide geographic areas at an infinitesimally low cost per bit. The dominant type of service traffic in recent years of 100GbE (Gigabit Ethernet) is now starting to be complemented by 400 GbE.
The reason for this originates in data centers, which are the sources and sinks of most traffic in our global cloud economy. 400GbE has begun replacing 100 GbE as the standard communication interface within data centers. A prime mission of optical networks is to extend these interfaces beyond data center walls for applications interworking and data mirroring.
As shown in Figure 1, this imposes a practical requirement for optical networks to provide line rates at 400G (gigabits per second) and 800G to transport this service traffic effectively. Note that an optical network line – which physically is a wavelength in a dense wavelength division multiplexed (DWDM) network – can support multiple services. For example, an 800G line can transport 4x100GbE and 1×400 GbE services traffic.
Gold standards for modulation
Figure 2 illustrates the two main knobs for achieving these line rates: Modulation and Baud Rate.
Modulation combines different phases and amplitudes of an optical sine wave carrier to create different “symbols,” where each symbol represents a string of bits. The simplest scheme, QPSK (quadrature phase-shift keying), uses four phases at a constant amplitude. This creates four symbols that encode two bits (00, 01, 10, 11). Figure 3 shows a time domain representation and constellation diagram for 16QAM (quadrature amplitude modulation), which adds amplitude levels to the phases to create 16 symbols that encode 4 bits.
You might think that a quick method to achieve a higher line rate is to increase the modulation density. Physics, however, dictates that there is no free lunch. Modulation density is inversely proportional to distance. Denser modulations mean shorter distances and less dense modulation means longer distances. The reason is that optical transmission signals “spread out” ever so slightly with distance. Because the symbols of denser modulations are packed more closely together, they become more easily confused with neighboring symbols due to noise encountered as they propagate along the fiber, making them more difficult to distinguish at the receiver. Technology can’t really improve modulation effectiveness beyond recent advances using probabilistic constellation shaping.
For optical transmission engineering, the industry broadly recognizes two modulations as gold standards: 16QAM (4 bits/symbol) for metro transport to about 500 miles, and QPSK (2 bits/symbol) for long haul transport to about 2,000 miles. When transmitting with these modulations, it is possible to address all applications within these two transport domains.
Given these target modulations, you can easily calculate the minimum baud rates required to support them for 400G and 800G line rates. The results shown are based on 20% forward-error correction (FEC) overhead.
|Metro (16QAM)||Long Haul (QPSK)|
|400G line rate||>62.5 Gbaud||>125 Gbaud|
|800G line rate||>125 Gbaud||>250 Gbaud|
Capacity-reach and power-cost optimized transceivers
DWDM optical transport networks rely on transceivers to launch wavelengths with high power from 0 dBM to 3 dBm (1 mW to 2 mW). These power levels are needed to transmit the wavelengths without signal regeneration across multiple nodes in metro and long haul network applications. (Another application called IP-over-DWDM, not covered here, directly integrates transceivers into routers to launch low power (-10 dBm) wavelengths for point-to-point or minimal hop connectivity.)The industry supports two types of coherent optical transport solution using minimum 0 dBm wavelength launch power, as illustrated in Figure 4.
- Capacity-reach optimized: These solutions maximize the spectral efficiency for any distance and fiber condition. They are used in most long-haul applications to squeeze every bit of capacity from a channel and can also be economical in high-density metro applications. They rely on proprietary transceiver technologies that maximize the bit rate by programmatically creating the optimum mix of baud rate, modulation scheme, and channel width. These proprietary transceivers are not absolutely constrained by size and electrical power to deliver higher performance.
- Power-cost optimized: These solutions optimize for lower electrical power and cost, while providing strong enough performance for most metro applications, and as the technology evolves, also for regional transport. They rely on multisource pluggable transceivers that implement industry agreements from organizations like the Open ROADM Multi-Source Agreement (MSA) that emphasizes multivendor interoperability, as well as OpenZR+ and the Optical Internetworking Forum (OIF).
There are three main types of pluggables for coherent optical transceivers with different size and power dissipation constraints, enabling different levels of performance:
- CFP2: 55 cm3 / 25 W – 30 W
- OSFP: 29 cm3 / 30 W
- QSFP-DD: 14 cm3 / 25 W
Transceiver technology roadmap
Figure 5 summarizes the industry roadmap for coherent transceiver technologies supporting minimum 0 dBm wavelength launch power. In 2020, the initial commercial (made available to any optical network equipment vendor) capacity-reach optimized transceiver used dual wavelengths, each operating at 70 Gbaud. This transceiver enabled combining two 400G wavelengths using 16QAM modulation to provide an equivalent 800G line rate for metro applications, and two 200G wavelengths using QPSK modulation to provide an equivalent 400G line rate for long haul applications. This same commercial technology is now being updated with state-of-the-art 5 nm silicon to operate at 140 Gbaud. Available in early 2023, it will provide the same metro-16QAM and long haul-QPSK functionality but using a single much more economical wavelength. Other proprietary capacity-reach-optimized transceivers developed by optical equipment vendors for their own use operate only at 95 Gbaud. Based on the math, to support 800G wavelengths they require denser modulation schemes than 16QAM, which limits their reach for metro applications.For power-cost optimized 0 dBM optical transport solutions, a variety of multi-source pluggables exist or are being planned.
- 400G CFP2 @ 64 Gbaud (since 2021). This first-to-market 0 dBm pluggable provides strong 400G metro coverage. Its downside is a relatively large form factor.
- 400G QSFP-DD @ 64 Gbaud (1H 2023). This will provide the same 400G functionality as the CFP2 pluggable in a smaller more economical form factor.
- 800G OSFP @ 124 Gbaud (1H 2024). This will provide superior 800G performance to current 95 GBaud proprietary capacity-reach optimized transceivers, with all the size, electrical power, and cost benefits of a pluggable.
An interesting observation about the rollout of these transceiver technologies is that the higher-performing proprietary transceivers precede similar but more economical pluggable varieties. This is because there is always a demand for the best possible transceiver for long-haul applications and based on what technology allows at a point in time, this can be achieved more easily in proprietary form factors not constrained by size and power. Then as a next step, this technology becomes optimized to fit into constrained pluggables, trying to preserve as much performance as possible.
Relative performance of optical-transport technologies
Figure 6 summarizes the big picture for capacity-reach and power-cost optimized optical transport solutions at minimum 0 dBm wavelength launch power. Rather than show absolute distances that are open to debate based on conditions including fiber type and amplification, the diagram shows the relative performance of approaches that at a common 0 dBm launch power are governed almost entirely by baud rate. The diagram shows that we are reaching what might be considered an ideal situation for capacity-reach optimized solutions. Although still based on proprietary technology, new 5 nm-140 Gbaud DSP transceivers deliver:
- An exclusive 1.2 Tb/sec (1.2T) single wavelength for short haul metro applications.
- Complete 800G coverage for all metro applications using 16QAM modulation, compared to current 7 nm-95 Gbaud based 800G solutions that only cover small metro regions.
- Complete 400G long haul coverage using QPSK modulation.
In parallel, for power-cost optimized solutions, multisource pluggable transceivers deliver continuous performance improvements. For metro applications these pluggables will rival the performance of capacity-reach optimized solutions, but at a much lower cost.
Orderly spectrum use
To conclude this article, we look at how these various optical technologies can co-exist on a fiber, which in the commonly used C-band has about 5000 GHz of spectrum available. Physics dictates that higher baud rates require wider channel widths, and Figure 7 summarizes the spectrum needs for all the transceiver technologies noted in this piece. Except for the proprietary 95 Gbaud technology, all other transceivers operate within an orderly 75 GHz/150 GHz spectral grid.This facilitates using the spectrum efficiently. Both capacity-reach and power-cost optimized transmission technologies can co-exist in the metro and long-haul applications spaces and transition from one to the other as the technologies evolve without creating spectral gaps.
Anuradha Udunuwara says
This is an excellent refresher from Jonathan. I always love his to-the-point crisp clear explanations. Jonathan is clearly one of the best veterans we have in the industry today.
Jonathan Homa says
Many thanks, Anuradha. I look forward to our next meeting at an industry event. Best, Jonathan
Martin Rowe says
@Anuradha, what in particular did you like?
Anuradha Udunuwara says
The simple explnation and summarization of most of the things necessary to understand modern optical networking in such a short article.