Cisco Cisco DWDM Transceiver Modules 白皮書

White Paper

Page 2 of 6

Limitations of Traditional Methods for Measuring Fiber Bandwidth

In January 1996, the IEEE Computer Society LAN MAN Standards Committee approved a project

to develop a supplement to the Ethernet standard to support operation at a data rate of 1000

Mbps. The new standard’s physical layers include 1000BASE-SX, a short-wavelength transceiver

operating at 850 nanometers (nm) over multimode fibers only, and 1000BASE-LX, a long-

wavelength laser operating at 1300nm over both multimode and single-mode fibers.

During September 1997, it was demonstrated that technical issues remained to be solved to

achieve the targeted Gigabit Ethernet multimode fiber operating distances for 1000BASE-LX. The

issues were related to the use of laser sources with multimode fibers. More specifically, it was

proven that the small spot size of a laser compared to an LED was causing the reduction of

bandwidth available in the multimode fiber. Figure 2 depicts the difference between these launch

conditions, showing how an LED illuminates the entire core of the fiber, while a laser source only

illuminates a small portion of the fiber core. The laser does not energize as many dispersive

modes of the fiber waveguide as does the overfilled-launch (OFL) of an LED, so the fiber modal

dispersion and bandwidth performance are different than might be expected from the OFL

measurements.

Figure 2. OFL Light Source Versus Laser Light Source

Since multimode fibers were first designed to work with LED-based systems, OFL bandwidth

(OFLBW) is the traditional method to characterize their modal properties and capacity. However,

during the standardization of Gigabit Ethernet, for operation near a wavelength of 1300nm using

OM1 fiber, it was found that the effective modal bandwidth (EMB) of center launch (CL) was

significantly less than the OFLBW specification. This was experimentally and theoretically verified

to be true for a high percentage of cases (approximately 10 to 30 percent), thus proving the lack of

correlation between CL EMB and OFLBW.