Who’s better, The Beatles or The Rolling Stones? The Smiths or The Cure?
Fans of either band usually had very strong opinions, and it was way too easy to get drawn into heated discussions while, on my side, I have always enjoyed both “rivals” for their unique peculiarities.
Interestingly, since fiber optics was invented, one of the recurring technology discussions is about “bands”.
No, not rock or pop bands, but the bands of the electromagnetic spectrum which can be used for fiber optic transmissions.
The near-infrared region of the electromagnetic spectrum which can be used for single mode fiber transmission, has been conventionally divided into five sub-bands, identified with a letter.
“O”: original; “E”: extended; “S”: short; “C”: conventional; “L”: long
Like in the rock musical landscape, two battling bands – the O-band and the C-band, emerged as the more successful, each with its unique features.
The O-band features a very low chromatic dispersion, but a higher attenuation per kilometer; the C-band features a lower attenuation per kilometer, but higher chromatic dispersion.
O-band was the “original” band, the first for which we learned how to build lasers and detectors. Over time, the very low chromatic dispersion has allowed use of particularly cheap laser sources, making O-band the preferred choice for short reach optical interconnects, for which the higher attenuation per km is not a problem.
C-band has the lower overall attenuation per km, and it has become the preferred choice for long distances, at the cost of having to cope with chromatic dispersion. Transmission penalties due to chromatic dispersion increase with the square of the bit rate and get cumulatively worse with distance. Due to this fact, dispersion management by optical or electrical compensation techniques is needed for the typical metro and long-haul links (hundreds or thousands of km).
The emergence of centralized RAN architectures has created the new fronthaul segment in the past decade: the interconnect between baseband units and radios, which used to be a simple site cabling based on short-reach optics, has gradually become a full-fledged transport network segment. This new fronthaul segment comes with unique characteristics and requirements: it has some challenging performance aspects of metro aggregation segment (fiber scarcity, distance in the 10-20 km range) but also the well-known cost sensitivity of the radio access segment.
Cost sensitivity here must always be considered from a total cost of ownership standpoint. However, shoehorning expensive technologies conceived for the metro segment and differentiation by over-engineering performance are two approaches that do not work. Instead, what pays out nicely is a “good enough” approach to performance, making sure any increase in complexity and CAPEX of solutions is compensated by a greater and monetizable value generated for the end user.
These principles worked fine for the adaptation of metro DWDM technologies in C-band. Starting from the same basic technology building blocks, cost/performance optimized solutions for DWDM fronthaul have appeared for the typical reach of 15 km and up to 48 wavelengths with 100 GHz spacing and are now widely adopted at 10 Gb/s and 25 Gb/s per wavelength.
So, do we have a winner for the battle of the bands? Well, to use what’s probably the favorite answer of any engineer all around the world – “it depends…”
The O-band has started to fight back from the short-reach domain, repurposing existing technology building blocks for new applications in fronthaul networks. In particular, the first generation of 400 G 10 km duplex fiber optics defined by IEEE802.3, 400GBASE-LR8, was using an 8x50G (25Gbaud PAM-4) architecture and 8 wavelengths of the so-called LAN-WDM (LWDM), an O-band WDM variant with approximately 800 GHz carrier spacing. The 400G optical standard for 10 km then evolved in the second generation with a 4x100G architecture (400GBASE-LR4), but the palette of 25G lasers for eight LAN-WDM wavelengths remained in the toolbox of the optical industry. With minor adaptations of existing designs, it was possible to add four more wavelengths for a total of 12, nicely supporting fronthaul use cases requiring a maximum of 6 bidirectional 25G links, and with a lower cost than the corresponding C-band DWDM fronthaul solution.
Different bands, different peculiarities: at 25G per wavelength, we have two winners!
Where maximum capacity of 24 bidirectional links per fiber is a must, the more scalable C-band wins. Where the limit to 6 bidirectional links per fiber is acceptable, the more frugal O-band wins.
The next challenge that WDM fronthaul needs to tackle is scaling up to 50 Gb/s per wavelength. And the “battle of the bands” is raging on again!
50 Gb/s per wavelength is a reasonable next step: it can be obtained from today’s 25G optical components by using PAM-4 modulation, allowing to transmit four amplitude levels (“symbols”) and thus delivering two bits per symbol. This means effectively doubling capacity the same components can provide, with respect to the two-level OOK modulation in use today.
Unfortunately, PAM-4 modulation adds additional transmission challenges.
To begin with, a four amplitude level signal means that the average distance between such levels is 1/3 of the distance one would have only transmitting two levels for the same received average power. With PAM-4, the receiver needs 3x the average power to achieve the same received signal quality. In other words, PAM-4 modulation requires more powerful transmitters.
Furthermore, the PAM-4 signal features more transitions among symbols, which are more prone to the detrimental effects of pulse broadening due to chromatic dispersion and result in increased transmission penalties and reduced transmission reach.
Here it would seem the O-band has the potential to deliver the final blow. The key problem at 50G is chromatic dispersion – C-band solutions will be really challenged distance-wise, while with O-band it doesn’t seem problematic to maintain the typical 15 km-20 km target reach of fronthaul links at 50G for today’s 12 wavelengths LAN-WDM grid.
The future of C-band solutions would seem confined to use cases characterized by shorter reaches (little km) and higher capacities (>6 bidirectional links per fiber).
Anyway, one should never underestimate the tricks in the bag of an old fighter like the C-band. There is a large arsenal of signal conditioning technologies (dispersion compensation, optical amplification) in C-band, used for ages in metro and long haul DWDM transport, which in the future could be transformed and adapted cost/performance-wise to the fronthaul segment.
For example, when the time comes for 100G per wavelength, the adaptation of one of the most successful technologies ever in metro networks – coherent transmission – will be a serious challenger for O-band 100Gb/s solutions using 50Gbaud PAM-4.
The “battle of the bands” will continue to animate the WDM fronthaul technology stage in the coming years. It is my opinion that both O-band and C-band will win, each for the use cases that will better fit based on their unique peculiarities.
After all, can we honestly tell the winner between The Beatles and The Rolling Stones, or do both provide a great listening experience 😊?