by Paul Hospodar, Product Line Manager – L-com
Multimode fiber technology (MMF) is a critical backbone to support next generation Ethernet speeds in short reach applications. As technology stretches to terabit speeds and beyond, wideband MMF (WBMMF) alternatives to the currently leveraged OM3/OM4 laser-optimized MMF (LOMMF) are rising in importance. Future iterations of 400G Ethernet have already been proposed with up to 32 channels; this increase in lanes highlights the need for a rethink on current MMF and associated optical technologies.
In order to understand the increasing need for an upgraded MMF, it may be helpful to understand the differences between the two major fiber optic cable technologies and their respective optics: Multimode Fiber (MMF) and Single Mode Fiber (SMF). The MMF cable has a bigger core (50 um to 62.5 um) while the SMF cable has a much thinner core (8.3um to 10um). This allows for the MMF cable to operate with longer laser wavelengths of 850nm; the longer wavelength makes gathering light a simpler task which permits the use of more cost-effective optics, such as incoherent LEDs and Vertical Cavity Surface Emitting Lasers (VCSELs). Typically, SMF Small Form-factor Pluggable transceivers (SFPs) are more expensive with coherent light sources such as Fabry-Perot (FP) and Distributed Feedback (DFB) lasers. This increased cost, however, comes with the benefit of better range.
While the MMF cable can handle high data rates at up to 100 meters, modal dispersion (MD) becomes worse with distance. Modal dispersion can be defined as the separation of a light pulse into multiple paths/modes where each light component arrives at the receiver at varying times. A more practical measure of MD can be found with the parameter of effective modal bandwidth (EMB); a good EMB will be gained only from a low MD. The SMF, however, is generally limited by its chromatic dispersion (CD) or the difference in propagation velocity between the various colors, or wavelengths, within a light beam. This results in spreading of the pulses, and the introduction of errors in the digital information. The CD is a parameter that is cited for both MMF and SMF cables.
The SMF SFPs will have generally granted a much more flexible link budget for long range or high bandwidth, 10G and beyond installations. While SMF cabling can be seen as more future-proof than MMF installations, the upfront capital expense may not be feasible for some installations. The MMF still functions as the backbone in short-reach optical links for local area networks (LANs) with high reliability, high energy-efficiency, and cost-effectiveness as the key considerations.
Lessening the Mess: Shortwave Wavelength Division Multiplexing
Data centers are tasked with housing the optical links that are meant to support the ever-increasing web traffic at ever-increasing data rates. This has led toa painful balancing act of data rate and cable density. For instance, the IEEE 802.3ba specification leverages ten parallel MMFs in each direction with a 24-fiber MPO connector to support 100Gbps Ethernet. The IEEE 803.2bs, 400GBASE-SR16 proposal involves sixteen, 25 Gbps lanes with an unofficial 32-fiber MPO connector for 400Gbps Ethernet (Refer to Table 1). Hence, more expensive and higher density cable can be used at the cost of greater expense and interconnect complexity up to practical limits of interconnect density.
However, there is a viable solution to this clutter conundrum by using Shortwave Wavelength Division Multiplexing (SWDM). SWDM enables the transmission of multiple channels over duplex MMF through the use of multiple wavelengths between 860nm to 1100nm; this allows for a four-fold increase in the data rate per fiber. In 2017, the SWDM Alliance was formed with an initial goal of supporting 40 Gbps using four 10 Gbps wavelengths and eventually supporting 100 Gbps with four 25 Gbps wavelengths . Recently a Multi-Source Agreement (MSA) was published by the SWDM Alliance covering optical interfaces for 100 Gbps optical transceivers . In essence, four different VCSELs with four different wavelengths 30nm apart (850, 880, 910, 940 nm) and independent bit streams can be multiplexed before transmission over a MMF link. At the other end of the MMF link, the de-multiplexed signal with four different wavelengths is received at four separate receivers (Figure 1).
This solution can provide a clean cabling installation without worrying about cost increases and power usage—the SWDM4 architecture proposed can have power dissipation as low as 1.5W . Figure 2 illustrates the decreased cable usage through WDM schemes where 400G and beyond speeds can be achieved with WDM and parallel transmission tactics.
Optimizing MMF Properties to Suit SWDM: Wideband MMF
This use of SWDM and parallel lanes provides less than adequate performance at wavelengths longer than 850nm. Standard OM3 and OM4 fibers generally have a peak effective modal bandwidth (EMB) at around 860 nm (OM3: 2000 MHz*km, OM4: 4700 MHz*km); at around 1310nm the EMB performance degrades to around 500 MHz*km (Refer to Figure 3). This is because these LOMMF cables were optimally designed to perform at the singular wavelength of 850nm. It should be noted that the chromatic dispersion of fiber tends to decrease with increasing wavelength. The decreasing EMB at longer wavelengths can, therefore, be slightly offset by the increase in chromatic bandwidth. For this reason, many manufacturers often cite the parameter of effective bandwidth (EB), which takes into account both EMB and CD.
The OM5 was released in the TIA-492AAAE in June 2017 and is optimized for WDM operation with VSCELs. Structurally, the core is the same as OM2, OM3, and OM4 at 50um with the same 125um cladding for backward compatibility. While OM3 and OM4 have aqua jackets, an OM5 cable can be identified by a distinctive lime green jacket (Figure 4). The EMB is defined across four wavelengths from 840nm to 953 nm. The EMB of both OM4 and OM5 at 850nm is specified to be 4700 MHz*km whereas the EMB at 953nm is specified to be a minimum of 2470 MHz*km for only OM5 cables. According the SWDM Alliance MSA, the OM5 cable can operate up to 150m while the OM4 cable has a range of 100m for 100G-SWDM4 operation. Moreover, this newer MMF variant exhibits bend tolerant characteristics that minimize signal loss in the event the cable is bent beyond the specified bend radius.
When OM5 is Really Useful
The first of SWDM-enabled VSCELs have already been released and, with data centers predicted to pass the vast majority of Internet traffic, it is important for the smart use of space. The OM5 cable is particularly applicable in these scenarios where it is optimized for wideband operation.
The general utility of the OM5 cable has been argued in the past where critics explain how OM4 and OM4+ MMF variants could perform on par with OM5. For instance, at the 2017 European Conference on Optical Communications (ECOC), the OM4 cable had been shown to perform up to 300m for 100G Ethernet with the SWDM4 Extended Reach QSFP28 transceiver, while the OM5 cable is meant to perform up to 400m for 100G Ethernet with the SWDM4 QSFP28 transceiver .
Another significant point of contention is the current price drop that can be seen with 100G optics; leveraging SMF cables could provide the same speeds at a similar price point. As shown in Figure 2, the true merit can be more readily seen at speeds beyond 100G, where parallel fiber transmission can grow to up to 16 transmit channels and 16 receive channels—32 channels in total (Refer to Table 1). Still, it can be argued that port density can be improved by breaking out the 40G and 100G transceivers with their additional port interfaces, where individual unused 10Gbps and 25 Gbps channels can be leveraged. In essence, the variety of data center routing scenarios warrants a variety of potential solutions.
Several whitepapers and blogs have been released already explaining the merit of OM5 despite the skepticism. The general consensus is that the standardization of multi-wavelength operation is growing more and more pertinent with wideband optics such as the QSFP 40G BiDi transceivers as well as the 40G and 100G SWDM4 transceivers recently released. While LOMMF has been leveraged with multi-wavelength technologies, there is no real support for it as most data sheets will not even list performance beyond 900 nm. After all, wideband operation is not the reason why the OM3 and OM4 cables were generated.
The OM5 WBMMF is a necessary cabling backbone to the already commercially available wavelength division multiplexed (WDM) VSCELs that dominate many short-reach data center applications. The method of multiplexing various wavelengths of light effectively mitigates the use of increasing parallel lane topologies for terabit Ethernet speeds. In essence, implementing OM5 technology can future-proof data centers for 100G speeds, all while preventing the high upfront cost of coherent optics for SMF as well as cable clutter.
1. Balemarthy, Kasyapa, et al. “Next-Generation Wideband Multimode Fibers for Data Centers.” Next-Generation Optical Networks for Data Centers and Short-Reach Links III, July 2016, doi:10.1117/12.2220700.
4. Lyubomirsky, I., et al. “100G SWDM4 Transmission over 300m Wideband MMF.” 2015 European Conference on Optical Communication (ECOC), 2015, doi:10.1109/ecoc.2015.7341778.
5. Parsons, Earl Ryan, et al. “The Impact of Effective Modal Bandwidth on 100G SWDM Transmission Over 250 m OM5 and Left-Tilt OM4 Multimode Fibers.” Journal of Lightwave Technology, vol. 36, no. 24, 2018, pp. 5841–5848., doi:10.1109/jlt.2018.2878666.
7. Kuchta, D. M., et al. “A 4-λ, 40Gb/s/λ Bandwidth Extension of Multimode Fiber in the 850nm Range.” Optical Fiber Communication Conference, 2015, doi:10.1364/ofc.2015.w1d.4.