The Fiber Optic Association - Tech Topics
Power Penalty For Mixing 50/125 And 62.5/125 Fibers In One LinkMost premises installations are now using 50/125 fiber for 1/10/100 Gigabit Ethernet instead of 62.5/125 fiber, since the laser-optimized 50/125 fiber is a better solution for high speed systems with laser sources and is still compatible with virtually every system that will run on 62.5//125 fiber.
If one connects a 50/125 fiber to a 62.5/125 fiber, the smaller core of the 50/125 fiber easily couples to the 62.5/125 fiber and is very insensitive to offset and angular misalignment. However, the larger core of 62.5/125 fiber overfills the core of the 50/125 fiber, creating excess loss.
The traditional range of mismatch coupling losses has been covered in several documents, including The Fiber Optic Technicians Manual, Table 17-4 (Excess loss in dB) covering the full range of legacy multimode fibers:
|50/125||0.9-1.6 dB||3.0-4.6 dB||4.7-9.0 dB|
|62.5/125||-||0.9 dB||2.1-4.1 dB|
There has been much speculation as to what the power penalty would be, especially when the system uses a GBE VCSEL source with it's narrower modal fill in the fiber, rather than a LED used for slower systems and most test equipment. As part of tests of cable plant loss with LEDs and VCSELs, we decided to see for ourselves the magnitude of the power penalty.
We set up a test to find the excess loss or
power penalty that can occur when mixing 50/125 and 62.5/125 micron
multimode fibers in a single link. To make the test more relevant for
current networks, we used both a standard 850 nm LED source and a 850 nm
VCSEL for test sources. The tests were done for both three different
conditions that should have relevance to all normal circumstances.
Launch---------[-]-----------50/125 Receive (1m)-------[Meter
Test 1: Launch with 1 meter 62.5/125 cable, as with a source with 62.5/125 patchcord connected to a network cable plant of 50/125 fiber.
Test 2: Launch with 20 meter 62.5/125 cable with one connector in the middle, such as the receiver would see in a short floor-to-floor link and a 50//125 patchcord at the receiver.
Test 3: Launch with 520 meter 62.5/125 cable with three connectors in the middle, the longest normally expected link for GBE into a 50//125 patchcord at the receiver.
In all three cases, the receive cable was a
1 m 50/125 jumper, to minimize all other loss factors, such as fiber
|Fiber excess loss connecting 62.5 to 50 micron|
|Control 62.5||0.33||Not Tested||0.12||0.05||0.05||0.04|
The data follow predictions very well. Losses are higher with LEDs than VCSELs, showing the lower mode fill of the VCSEL. LED losses go down as we go to longer launches, indicating the LED mode fill decreases. Note, however, the VCSEL has slightly lower excess loss at 20 meters than at either 1 meter or 520 meters. This is similar to data we obtained in our earlier tests, where the mode mixing from the extra connectors in longer cable plants makes the VCSEL differences level out in only short cables.
The data is very definitive, however. The excess loss caused by coupling 50/125 fiber to 652.5/125 fiber enacts a severe penalty on power budget, especially when for used GBE networks with their lower power margins.
Mixing 50/125 and 62.5/125 fiber in one cable run is not sensible. If one is using 50/125 fiber instead of 62.5/125, all components - cable and patchcords should be 50/125. The best way to segregate 50/125 and 62.5/125 fiber in one installation is to color code the cables to keep them separate and/or use different connectors. Since the LC connector is becoming the de facto standard for gigabit and above, using LCs for 50/125 fiber is a bulletproof solution.
(C) 2002-8, The Fiber Optic Association,
Return To The FOA Home Page
Return To FOA Tech Topics