Multimode fiber testing gets international attention

almost feel obligated to apologize for writing about technical issues related to multimode fiber again. However, multimode fiber is undergoing a revival in both interest and performance that is creating new issues users and installers must be aware of to guarantee successful usage. So here we go again.

Multimode fiber research and development was shoved aside in the mid-1980s when the phone companies switched to single-mode fiber for their long-haul networks. As the telcos expanded their fiber usage, fiber manufacturers developed new and better types of single-mode fiber and perfected the manufacturing techniques, so single-mode fiber became incredibly high-performance with costs less than kite string or fishing line.

Meanwhile, the premises networking market developed around the PC and adopted multimode fiber for its use. Multimode was preferred for its easier on-site termination and its ability to use inexpensive LED sources at the lower speeds PC networks required. One fiber, 62.5/125 micron, was used for virtually all multimode premises fiber in the United States. Everything worked well on multimode fiber until network speeds approached 1 gigabit (billion bits) per second.

Networks using speeds up to 100 megabits (million bits) per second used inexpensive LED sources. These slower networks had relatively high loss margins, typically tolerating cable plant losses of 6 to 20 dB. Most were distance-limited by cable plant loss; although, bandwidth limitations caused by multimode dispersion was an issue at distances around 2 km.

With the introduction of gigabit Ethernet, LED sources were too slow. Only lasers offered adequate modulation capability. New inexpensive lasers, called VCSELs, had been developed that offered plenty of power and speed. The FDDI-grade 62.5/125 multimode fiber commonly used in premises networks had inadequate bandwidth capability and limited the distances over which gigabit links would operate, so networks moved toward laser-optimized 50/125 micron multimode fiber.

As transmission speeds get higher, the transmitter pulses get shorter, since you need to squeeze more pulses in the same time frame. A shorter pulse has less total power content, but the receiver still needs a certain power level to recognize the signal. So, the loss margin goes down. For example, if a 100 Mb/s link works with a 14 dB link margin, the same link modulated 10 times faster at 1 Gb/s would have a link margin 10 times (or 10 dB) less, or 4 dB.

In addition, the bandwidth limitation of multimode fiber reduced the link margin even further. The pulse spreading caused by the fiber dispersion reduced the link margin of gigabit Ethernet to a worst-case value of about 2.3 dB.

Now, a link that can operate over a cable plant loss range of 0 to 14 dB provides a lot of latitude in installing and testing the cable plant. It obviously can tolerate higher loss connectors, and the need for accurate loss testing is minimal. But, a link that has a margin of only 2–4 dB is more critical. Connectors must be lower loss, well below the TIA-allowed 0.75 dB per connector. Testing also needs to be more accurate, since an uncertainty of 0.5 dB in loss measurements is a lot more worrisome in testing a 2–4 dB link than a 14 dB one.

The biggest factor in testing uncertainty for multimode fiber is mode power distribution, which I covered in the April column. At a recent TIA fiber optic standards meeting, multimode fiber testing was the primary subject of discussion. This much seemed to be agreed on:

1. Multimode testing requires controlled test source launch conditions.

2. Current methods of specifying launch conditions are inadequate.

3. OTDR testing (covered in July 2006) is not comparable to light source and power meter testing (insertion loss).

4. Test equipment manufacturers are lax in understanding the issues, specifying their instruments and educating customers on the proper way to test multimode fiber.

5. U.S. and international standards are not in agreement.

As a result of discussing these issues, the TIA FO-4 fiber optic group is starting a task group with high priority to address these issues. The group includes enough technical talent that a solution is finally likely. How long will it take? Probably, at least a year, but once the work is completed, we should be able to address the above issues and create a reasonable test method that will reduce multimode testing uncertainty and facilitate the continued use of multimode fiber for gigabit and 10 gigabit networks and maybe even higher. The next step will be to get the methods incorporated in industry standards, which is, alas, another time-consuming project. EC

The FOA, which I represent at TIA meetings, is considering starting a multimode test round robin, sending a simulated cable plant around to any interested parties willing to test it using their methods and test equipment and allow the results to be included in a report to the TIA committees. Any fiber optic contractors interested in participating should contact me directly.

HAYES is a VDV writer and trainer and the president of The Fiber Optic Association. Find him at www.JimHayes.com.