What you need to know
The rush to implement Gigabit Ethernet without understanding Multimode fibre characteristics can be disastrous. This paper will cover the evolution of Multimode fibre and discuss the potential pitfalls in using Gigabit Ethernet in a Multimode fibre optic system.

Multimode fibre origins
Multimode fibre technology evolved in the 1970s to resolve technical problems associated with super high bandwidth Singlemode fibres, light sources and connectors. Problems included how to manufacture products to the demanding tolerances required by Singlemode fibres, how to increase the fibre coupling
efficiency and longer life cycles of laser diodes required for singlemode fibres.

One solution was to increase the core size of the fibre to 50, and 62.5 microns (which increases the area which transmits the optical signal) and the Numerical Aperture (N.A.) which allows for greater coupling efficiency from the LED at the transmitter and at connections and splices. The impact of this solution was the introduction of multiple modes (or light paths) from the same light source. Multiple modes transmitted down a fibre allows for the signal to disperse (modal dispersion). An example would be to compare the Light Emitting Diode (LED) to a shotgun. With the shotgun the trigger is pulled and many pellets are emitted from the shell casing. In fibre the trigger of the LED would launch a optical pulse emitting multiple
photons with each carrying the same optical signal. The end result is that when these photons are launched and transmitted down the of the optical fibre, the timing of those taking the axial path down the centre or close to the centre of the fibre is shorter than the higher order modes being transmitted in a pattern similar to a sine wave. Since these modes arrive at the receiver at separate times but carry the same information, the resulting pulse recovered at the receiver is actually the timing between from the first to the last photon of each pulse. The bandwidth (information capacity) of multimode fibre is measured as MHz-km. A multimode 62.5/125 micron fibre with a bandwidth of 160Mhz-Km @ 850nm and 500 MHz-Km @ 1300nm has a defined amount of information it can transmit over a known distance. Therefore all multimode fibres have a defined bandwidth or signal capacity relative to the kilometre (3,281 feet). The limited signal capacity was the penalty of using multimode fibres rather than singlemode fibres. Through the late 1970’s the first users of fibre optics only had the larger multimode fibres available for use. The smaller 50/125 fibre with its lower loss and higher bandwidth was used internationally for voice transmission at 850nm and later at 1300nm by telephone companies until singlemode technology caught up with demand in 1982 and 1983. From 1983 to the present, all long distance service providers (telephone companies and later CATV companies) used only the 9/125 micron singlemode fibres to meet their requirements.

What happened to Multimode fibres
Larger core multimode fibres are easier to use with low cost LEDs and to cross-connect using optical connectors. These features met the demands of the emerging data communications industry and provided a low cost, high bandwidth solution for linking buildings together in campus environments, while also providing a solution for security applications including closed circuit television (CCTV) and access
controls.

New system standards including the IEEE 802.3 Ethernet protocol operating at 10 Mb/S and IEEE 802.5 Token Ring at 4 Mb/s could now be linked beyond the distances limited by existing coax and twisted pair technology. In addition, fibre optics provided both noise immunity and greater security for network users.
In the late 1980’s a new Fibre Distributed Data Interface (FDDI) standard evolved to transmit data at 100Mb/s (125 MHz). Features included self healing capabilities using optical switches, defined connector types (ST & MIC), wavelengths, fibres, distances and access on/off the physical network. Even at this "high speed" the existing multimode fibre worked well and became very mature in use and applications. The 62.5/125 fibre was selected as the preferred fibre and the 50/125 and 100/140 were listed as alternative fibres. The selection of the 50/125 (as an alternate) would in time come back and haunt the fibre industry as data rates increased. The 1300nm wavelength provided higher bandwidth for multimode and also allowed for use with singlemode fibres.

Increased data rates influence on Multimode fibre systems
Data rates and requirements kept increasing! By the mid 1990’s data rates up to 622 Mb/s were being implemented through the Asynchronous Transfer Mode (ATM) protocol. At this data rate existing LEDs were being modulated to their limitation, and the usable distance for these systems (limited by the modal dispersion) became less than a kilometre.

Parallel to the ATM standard the enhanced "Fast Ethernet" data rate of 100Mb/s allowed most organizations to migrate to higher data levels with lower transition costs than changing to the ATM standard at (155 or 622Mb/s). At the 100 Mb/s data rate modal dispersion still wasn’t an issue. Jump forward to the late 1990’s. Demand from data users required backbone speeds at an order of magnitude greater than Fast Ethernet could transmit. The next generation became the IEEE-802.3Z Gigabit Ethernet standard transmitting 1 billion bits of information per second with the light source modulated at 1.25 GHz/second. Other high speed standards include Fibre Channel, High Performance Parallel Interface (HIPPI), and ATM (at 622Mb/s). Still next generation protocols continue to migrate to higher speeds including ATM at 2.488 Gb/s and 10 Gigabit Ethernet.

Problems resulting from the increased data rates
Problem # 1     LEDs cannot be modulated at these speeds, therefore, a faster source is required! The faster source is the LASER

Problem #2     All lasers are affected by reflections caused by connectors and mechanical splices.

Problem #3     Fibre test equipment for multimode fibres use LEDs. Lasers launch light into the axis of the core of the fibres so the test results vary.

Problem #4     Test standards need to be written or updated to address launch conditions, bandwidth measurements & optical return loss.

Problem #5     Test equipment is limited to meet the existing standards and are designed for field use. New equipment must be developed specifically to duplicate laser launch conditions & measure reflectivity as well as attenuation.

Issues for the end user
Does all this mean that the multimode backbone fibre we are currently using may not support higher speed transmission?
Actually, that possibility is very real. The reason is modal dispersion which limits the effective bandwidth of the installed plant is proportionately related to distance. If your data rate increases your distance must decrease. How is this problem solved?
What about the validity of the test reports my contractor gave me at the completion of the tests when the system was installed? Aren’t these good?
Yes and no. If you will be using the fibres with systems using LED light sources then the test reports are valid. However if you are using a transmission system using lasers (VCSELs, Fabrey-Perot and DFB types) then the measurement values will be different.

How can this be?
When multimode fibre optic test procedures (FOTP) were originally written in the late 1970’s and early 1980’s no one foresaw the need to establish procedures to work with the types of sources and high speed data rates we currently use (or are planning to use). The use of the "Overfilled Launch Condition" which duplicates how LEDs emit light into an optical fibre has been effective through the years.

With laser transmitters the launch conditions require a restricted mode launch (or "Underfilled" technique). This requires the light sources to incorporate a Vertical Cavity Surface Emitting Laser (VCSEL) at 850nm and a Fabrey-Perot laser at 1300nm which duplicate the types of sources the transmission systems use. The good news is that in most cases the underfilled launch conditions measure losses from 1-2dB lower than the overfilled conditions.

I thought from the earlier statement that reflections may not allow my system to work. This concerns me greatly!
It should! This is a potential nightmare here. Lets look at a scenario, you’ve made the decision to upgrade your backbone from Fast Ethernet to Gigabit Ethernet. The contractor long ago installed, terminated and tested your fibre spans and they have worked transparently for many years.

Now you bring on line your new system and you are having serious errors. What do you do? First you analyze the problem and having found that it isn’t the equipment (or the vendors fault) then it has to be the fibre plant. You call the contractor back and request a retest, after which the resulting test report closely approximates the original test report and the contractor says all is OK and you get a bill for the retest. But the problem still exists and needs resolution. Why?

The problem is being caused by highly reflective connectors. The contractor doesn’t know what the problem is due to changes in connector & laser technology. The contractor is still working in the physical layer of the OSI model, without regards to how system changes affect the work, products and processes required.

How do I get this problem resolved? My network is down and we’ve spent a lot of time and money on this!
This is where the problem is? We (the fibre industry) have known for years about the problem of reflectance since all the singlemode systems have used lasers for years. There are three issues here:

Issue #1 The Test Standards
What is the current status of applicable standards? As stated earlier there are many new issues which now confront the fibre industry, updating standards to provide a roadmap to the industry is essential. In many cases numbers exist for reflectivity, but many are without merit.

A system example is IEEE 802.3Z which specifies multimode return loss at 20dB (no more than 1% of the optical power is reflected back to the source). Yet, this number is considered too low by many in the fibre industry. The industry needs more research to validate this number. If in doubt specify 30 dB (.1%). In the case of singlemode systems the standard specifies 26dB, yet in the singlemode industry a minimum standard is 40dB (.01%). Why should we tolerate a lower number and then have to go through this process again in the future?

What needs to be done?
First the EIA 455 series of fibre optic test procedures (FOTPs) needs to be updated where applicable.
These standards are the building blocks of the fibre industry. Each FOTP provides the conditions and
requirements which can then be duplicated by users, manufacturers and contractors.

Key FOTP’s & Standards include:
TIA/EIA 455-34 Interconnection Device Insertion Loss Test

TIA/EIA 455-50B Light Launch Conditions for Long Length Graded-
Index Optical Fibre Spectral Attenuation
Measurements

TIA/EIA 455-107A Determination of Component Reflectance or Link
System Return Loss Using a Loss Test Set

TIA/EIA 455-171 Attenuation by Substitution Measurement for Short Length Multimode Graded Index and Singlemode Optical Fibre Cable Assemblies

TIA/EIA 455-204 Measurement of Bandwidth on Multimode Fibre *

TIA/EIA 526-7 Measurement of Optical Power Loss of Installed Singlemode Fibre Cable Plant

TIA/EIA 526-14A Measurement of Optical Power Loss of Installed Multimode Fibre Cable Plant

Issue #2 The Test Equipment
Is the equipment available to perform the required tasks and are they traceable to Standards?

Visual test equipment and standards are applicable to connector quality, but only an interferometer can
distinguish the cause of a high reflection. This instrument is primarily used in lab environments to
troubleshoot products (including cable assemblies to identify the cause of highly reflective plugs).

Optical Return Loss (ORL) test sets are available for measuring singlemode assemblies and systems using the TIA/EIA 455-107A standard. For multimode usage activities in the TIA/EIA test committees are addressing the launch conditions required for the source. Once this is agreed upon & defined, a traceable standard will exist which test equipment manufacturers can model their equipment after.

The Optical Time Domain Reflectometer (OTDR) can also be used to measure reflective connections and
splices. The OTDR though is limited by its "dead zone" caused when looking at the first reflective event. The first reflection is usually the strongest in terms of the amount of power reflected back into the laser cavity. Therefore the OTDR may not be able to measure the connection at fault. In addition most contractors are not knowledgeable about measuring reflections using the OTDR.

The Optical Loss Test Set is one of the products which is mature and is traceable to existing standards for
overfilled launch conditions using LEDs. Even though the losses may measure higher than with the VCSEL (850nm) and Fabrey-Perot (1300nm) lasers the equipment still measures a known amount of signal loss for most applications.

Bandwidth Testing has been performed in the lab and not in the field. In the early 1980s field bandwidth test sets were available to test the bandwidth of the multimode fibre networks owned by the telephone
companies. The problem of multimode fibre networks owned by the telephone companies. The problem of
modal dispersion and higher data rates is the same problem as high speed networks are facing today in the LAN environment. In the case of the telephone companies they switched to singlemode fibres and
never looked back.

To test for bandwidth today, in the field, requires a review of the test standards including launch
conditions and bandwidth testing (TIA/EIA 455-204) to bring them to current standards. Once these are
defined we may see a field bandwidth tester in the next few years which would allow for field measurements to confirm true bandwidth.

Issue #3 The Contractor
Are they qualified for these new technologies? The lack of knowledge, standards and test equipment required to address these issues will cause problems for the early GbE and other high speed users until the standards and manufacturers catch up.

How do we prevent reflection problems?
Treat the fibre issues the same as in the singlemode industry. When possible install singlemode fibres in the network (or at least 25% of the optical fibres in the cable). Do not attempt mechanical or direct termination of the connector to the fibres as in multimode (an exception is the crimp and cleave connectors
with a prepolished endface) applications. The technique of splicing factory-made pigtails (one ended jumpers) through mechanical or fusion splicing techniques will also reduce the reflections to less than 40dB. Check your contractors qualifications, techniques and specification sheets to assure that you are getting the reflection quality of the components required.

How do we fix reflection problems?
First let’s make sure that reflections are the problem.

Overdriving the Photodetector
Optical receivers can also be oversaturated with light if too much power is received (just like looking at bright lights your eye can only adjust so much). This process would require a power meter and two variable attenuators to accurately measure TX and RX power levels. You will also need to know the manufacturers receiver minimum and maximum optical power specifications.

Reflection
The use of index matching gel (IMG) on the first connections at the patch panel will reduce existing reflection levels. Your should be able to use this on the surface of the plugs closest to the transmitter as they induce the greatest reflections. Once IMG is on the plus surface this will attract contaminants so dustcaps are critical. If cleaning is required, both sides of the connection must be cleaned along with the mating adapter. Another option would be to order pre-terminated "pigtails" which have a PC (lower reflectance polish) and have these spliced to the cable. In the case of pre-existing terminations you could cut-off the existing connections and re-terminate with the newer lower reflectance (PC) plugs.

A new piece of equipment in the industry is the field retro-polisher. This portable polisher was designed for singlemode upgrades where reflections limited the effectiveness of singlemode transmission systems. The polisher can be used for new terminations or upgrading existing terminations by grinding the surface of the ferrule and fibre to PC specifications. Final polishing is done to bring the surface quality to low loss values.

The Light Brigade and Fibre Optic Training The Light Brigade (TLB) of Kent, Washington is the leader in
fibre optic training. Since 1987, the company has provided hundreds of courses internationally to over 20,000 students. One of the challenges of TLB is to be proactive of changes in technology. This philosophy encompasses the physical issues such as fibres, connectors etc., as well as active components such as lasers, systems, protocols and their relationship. Being able to research and apply these findings in our courses and training materials allows our clients to look ahead vs. Being reactive in planning and maintaining their fibre optic networks of the past, present and future.

© January 2000 The Light Brigade In other words - You haven’t seen anything yet!

Other Columns
Video Preview
White Papers
Larry Johnson June 1998 Hybrid Fibre vs. Multimode & Singlemode Fibre
Larry Johnson February 1999 What’s Your Tolerance Level?
Larry Johnson The Light colum