Anyone who has pulled fiber through a crowded data center rack, routed a patch cord around a tight corner in a telecom closet, or fished a drop cable through a residential wall has likely experienced a physical limit: sharp fiber bends increase signal attenuation (loss) and degrade performance. Bend a fiber too far and light escapes the core, resulting in maximum loss and the signal never reaching the receiver at all. For decades, this constraint shaped how fiber networks were designed, often incorporating generous fiber bend radii into every cabinet, tray, and pathway to prevent excessive signal loss.
ITU-T G.657 is the international optical fiber standard that directly addresses this performance constraint. Formally titled “Characteristics of a bending-loss insensitive single-mode optical fibre and cable,” the Recommendation was first introduced by the International Telecommunication Union Telecommunication Standardization Sector (ITU-T) in 2006, with updates in 2009 and 2012, in direct response to the global rollout of fiber-to-the-home (FTTH) networks that began in the early 2000s. Operators and fiber manufacturers needed a single-mode fiber that could withstand the periodic sharp bends associated with tight corners, small enclosures, and space-constrained installations found in buildings and homes, and help to mitigate the excessive signal loss that conventional standard single-mode fiber (defined under ITU-T G.652) would experience under similar conditions.
Why G.657 Optical Fiber Was Designed
Before G.657, fiber optic engineers and technicians worked around bend sensitivity by over-engineering the physical environment: larger cabinets, wider-radius pathways, and more conservative routing. That approach worked, but it added cost and consumed space that FTTH deployments, apartment buildings, and later data centers and other high-speed communications networks could no longer spare as fiber demand and density rapidly increased. G.657 fibers were designed to effectively address the challenge of mitigating excessive bend-induced signal losses in a more efficient manner. According to the ITU, the bend-insensitive properties of G.657 fiber enable smaller cabinets, pedestals, enclosures, and terminations; more engineer-friendly installation with less rework; and a lower total cost of ownership for the access network overall.
Key Technical Benefits of G.657 Optical Fiber
The defining characteristic of G.657 fiber is macrobend loss performance, which is how much signal loss/attenuation occurs when the fiber is bent around a given radius. G.657 fiber achieves this through a modified refractive index profile in the core and cladding, engineered to keep more of the transmitted light guided within the core even under bending conditions that would cause a conventional G.652 fiber to leak signal.
Importantly, the G.657 fiber classification applies only to single-mode fiber and is designed for backward compatibility with standard G.652 single-mode fiber. As a current example, Corning's SMF-28® Ultra optical fiber data sheet confirms bend performance that exceeds G.657.A1 while remaining fully compliant with G.652.D, using the same 9.2-micron (µ) mode field diameter (MFD) as the installed base of its G.652.D single-mode fiber. Lightera's AllWave+ fiber describes the same principle: G.657.A1 with full backward compatibility for splicing, testing, and network turn-up with existing G.652.D infrastructure. Thus, each of these fibers is classified as being compliant with both G.652.D and G.657.A1 standards. This matters because it means G.657 fiber is not a specialty fiber with its own sole classification requiring a parallel ecosystem of connectors, splicers, or test equipment. In a manner of speaking, it is an additional upgrade to standard G.652.D fiber delivering enhanced bend performance.
Where G.657 Optical Fiber Is Used
While G.657 fiber is useful in network environments where periodic, sharp bends are expected and unavoidable, it still has physical bend limitations like all other fibers, so it does not support every tight-radius application. Therefore, while it can withstand a greater degree and number of bending instances, that does not mean it is immune to all bending or to withstand continuous sharp or tight bending over distances. One prominent use application is FTTx networks, where fiber cabling to and inside the home, multi-dwelling unit (MDU), or commercial building must be routed around door frames and room corners, from floor to floor, and reside in compact wall boxes. Another rapidly growing area of demand is data center cabling, where high-density racks and overhead trays force fiber into tight turns between switches, patch panels, and equipment racks. Also, mobile backhaul applications where fiber is routed up and down towers, satellite dish and radar applications, and many other types of networks also benefit from G.657 fibers.
In fact, demand for its bend performance has intensified sharply in recent years. Data center fiber consumption grew by an estimated 76 percent year over year in 2025, driven largely by AI and cloud application infrastructure buildouts. Industry projections indicate that data centers will account for roughly 30 percent of global fiber demand by 2027, up from under 5 percent in 2024. Because G.657.A2 fiber is now widely used in data centers and other high-demand applications for its slightly greater bend-insensitivity than G.657. A1 fiber, which we’ll discuss in the next section, the surge in demand has created real global supply constraints. At the time of this article in 2026, global fiber manufacturers have massive order backlogs, since the process of adding new manufacturing capacity is complex and extensive, from the initial glass preform production through fully operational draw towers, can take a few years to bring online (especially when new facilities must also be constructed along with hiring staff, etc). The result? Saying that availability is constrained would be an understatement, as G.657 optical fiber, along with many other fiber types, is largely unavailable to many people and entities worldwide, except to those with existing contracts that consume all of the current output. This is expected to continue for some time, as major global manufacturers are working to expand capacity but cannot physically produce enough G.657 fiber to meet rapidly growing demand until additional manufacturing capacity comes online.
We covered this topic in more detail in a prior article published on June 30th 2026: What Is Causing the Global Optical Fiber Shortage in 2026?
The G.657 Sub-Categories: A1, A2, B2, and B3
G.657 is not a single specification. The ITU-T Recommendation splits it into two main categories: Category A, for Access Networks where some bending is common, and Category B, for the shorter, more bending-rich stretches at the end of an access network, such as inside buildings, where further bend-insensitivity is required. Each main category has sub-categories with progressively tighter minimum specified bend radii:
- G.657 at 10mm
- G.657.A2 and B2 at 7.5mm
- G.657.B3 at 5mm
G.657.A1 and G.657.A2 fibers are fully compliant with G.652.D fiber, meaning they meet or exceed every specified attribute of that Recommendation. In slight contrast, G.657.B2 and G.657.B3 fibers are described as compatible with G.652.D, meaning they introduce negligible system impairment despite small differences in chromatic dispersion (CD) and polarization mode dispersion (PMD) optical performance characteristics.
The fiber manufacturers verify the bend performance of a fiber using a standardized test method: wrapping a length of fiber around a mandrel with a defined radius for a specified number of turns, then measuring the induced attenuation (in decibels, dB) at specific wavelengths. This is where the millimeter figures noted above come from.
As one working example, Corning's SMF-28® Ultra fiber (G.652.D and exceeds G.657.A1) is tested at the 10mm mandrel radius for one turn, showing no more than 0.5dB of induced loss at 1550nm, while Corning's SMF-28® Contour fiber (G.652.D and aligning with G.657.A2) is tested at the tighter 7.5mm mandrel radius for one turn, at no more than 0.5dB of induced loss at the same 1550nm wavelength. Corning’s ClearCurve® LBL fiber (G.652.D and exceeds both G.657.A2 and G.657.B2) is also tested at the same 7.5mm mandrel radius for one turn, at no more than 0.4db of induced loss at 1550nm. Finally, Corning’s ClearCurve® ZBL fiber, designed to exceed the most demanding G.657.B3 specification, is tested at the 5mm mandrel radius for one turn, showing no more than 0.10dB of induced loss at 1550nm.
Comparably, Lightera's AllWave+, AllWave FLEX+, and EZ-Bend fibers follow the same performance pattern across the A1, A2, and B3 categories respectively. Regardless of the fiber manufacturer, the pattern across all sub-categories is consistent: as the mandrel radius decreases, the fiber design must work harder to keep light confined to the core, and the sub-category designation simply reflects how tight a bend the fiber can tolerate before excessive loss/attenuation is significant enough to be detrimental.
Common Misconceptions About G.657 Optical Fibers
Across the general public, relatively few people have deep knowledge of optical fibers, including G.657 fibers, aside from those working directly with optical fibers at a highly technical level (e.g., designing, manufacturing, cabling, advanced characterization testing, specialized fiber application expertise, etc.). Even within the fiber optic communications industry, many have foundational knowledge, but lack detailed knowledge of fiber construction, characteristics, and intricacies. This is not due to any fault of their own, but because their technology focus is often on network devices or other fiber-based technologies. Also, since most experience with fiber is in its finished cabled form, expertise with “bare” fiber in its most basic, unprotected form is even rarer.
While understandable, the result of this knowledge gap is that it is not uncommon for web content generated by the media and others about G.657 and other optical fibers to be inaccurate, misleading, or poorly communicated from a technical perspective.
Unfortunately, there are recognized instances in which inaccurate or misleading information is deliberately generated, often by less-than-reputable companies, as part of marketing or sales activities intended to drive business to a particular fiber, cable, or product. With the widespread availability and use of AI models in recent years, anyone can now use these tools to generate topic-relevant content with no or limited prior knowledge of the topic. Since these tools are prone to inaccuracies and use existing web content as a resource, when not prompted correctly and referencing information that already contains inaccuracies, and when the mistakes aren’t identified by the user, even more of this type of information ends up in the public domain.
Let’s look at a newer, rapidly growing specialized application of bare G.657 optical fiber that’s been in the news and frequently reported over the last year or two, following the start of the Ukraine-Russia conflict: fiber-optic drone connectivity. While bare fiber is almost always cabled for network installations, in this application G.657 fiber is kept in its original bare form, re-spooled, and integrated into a canister attached to the drone, which then pays out as the drone flies, maintaining a physical communications connection between the operator and the drone. (An example of a drone fiber canister can be viewed here: OptiTether™ fiber optic drone tether solutions.).
G.657 Misconception #1: It is More Durable Than Other Fiber Types
One common misstatement out there is that bend-insensitive G.657 fibers are “more durable” than other types of single-mode fiber. However, this is not the case, as documented by the fiber manufacturer G.657 data sheets. In addition to G.657 bare optical fibers sharing a similar physical diameter and protective coating approach as other types, all G.657 variants (A1, A2, and B3) specify the same tensile proof test of 100 kpsi (0.69 GPa) applied to the entire fiber length, matching the mechanical robustness of standard G.652 fiber. What distinguishes G.657 bend-insensitive fiber is not durability or a dissimilar physical form compared to other fibers. The difference is purely optical, as G.657 fibers have a different refractive index profile, resulting in increased resistance to macrobend loss, not any difference in tensile strength, coating, or physical toughness.
G.657 Misconception #2: It Can Be Spooled or Coiled Much Tighter Than Other Fiber Types
Another common claim is that G.657 fiber can be wound or spooled/coiled around much smaller core diameters or over greater distances than other fiber types without incurring excessive bend loss. This is also not the case, as it overstates what the specification provides. Every optical fiber, G.657 bend-insensitive included, has a defined minimum bend radius. For the G.657 subcategories, exactly the 10mm, 7.5mm, and 5mm figures described above, manufacturer mandrel test data confirms that loss increases sharply once the fiber is bent tighter than that radius. G.657's purpose is to maintain low loss through occasional, sharp bends encountered while routing cable around corners or through equipment, such as between racks in a data center or from floor to floor in a building riser, not to enable continuous, tightly wound spooling over long distances at radii smaller than its rated minimum.
It is also worth noting that bare optical fiber in factory-spooled form is, in nearly all cases, an intermediate manufacturing and shipping format rather than an end-use configuration. Most fiber is manufactured, tested, and shipped on factory spools, in lengths up to 50.4km, specifically to be integrated into cable and then installed with minimal tension and as little bending as possible. In fact, most manufacturers' factory-shipping spools have a large core diameter of approximately 6 inches to ensure the fiber remains well above its minimum bending radius. If it were true that G.657 fibers could be wound around extremely small spool core diameters over longer distances, the manufacturers themselves would do the same, as it’d enable them to fit that much more fiber on every shipping spool.
Additionally, a common misunderstanding is that, because many see on paper that A2 and B3 offer greater bend-insensitivity than A1, those fibers should always be used over A1 when possible, especially in spooled or coiled formats. However, as noted, the difference among the three G.657 variants is just a single wrap around a mandrel with a radius of 10mm, 7.5mm, or 5mm, the most extreme tight-bend scenario over a very short length. Thus, there are many applications in which it may be expected that A2 or B3 is an absolute requirement when in fact A1 may be perfectly suitable. For long lengths of spooled or coiled fibers, all fibers, including G.657, require a minimum circular bend diameter of at least a few inches, as reducing it below the threshold will produce excessive loss.
Even for the few highly specialized applications that utilize bare spooled or coiled G.657 fiber, such as drone fiber payout canisters, network simulators, optical delay lines, and OTDR launch fibers, all require their manufacturers to understand and adhere to the same underlying bend-radius physics as any other G.657 application, while always taking each application’s variables and goals into account when determining the most appropriate fiber types and solution designs.
Connect With M2 Optics About G.657 Fibers
Recognized for its vast expertise in designing and manufacturing customized solutions containing spooled bare optical fibers since 2001, M2 Optics supports essential fiber-optic testing, latency, and connectivity applications for leading entities across global telecom/ISP, data center, aerospace and defense, financial trading, and other market sectors.
To learn more about bend-insensitive fibers, or if you require G.657 fibers for drone/UxV connectivity, network and latency simulation testing, or precision optical time delay applications, we welcome you to contact the M2 Optics team at your convenience.
