NETS header NETS Homepage UCAR Homepage NCAR Homepage CISL Homepage NETS Homepage About NETS Work requests & support

Very Coarse Wave Division Multiplexor (VCWDM) testing

What the WDM system is supposed to do:

We want to use inexpensive 1310/1550 nm wave division multiplexers from Fiber Instrument Sales to eliminate expensive leased-line circuits. Specifically, our plan is to use Very Coarse Wave Division Multiplexers (VCWDM) to multiplex the 1310 nm wavelength from the FRGP M20's Abilene PoS OC-12 interface with the 1550 nm wavelength of the Level3-bound Cisco CWDM GigE. The two waves will be combined at 1200 Larimer and separated at 910 15th Street (and, of course, the opposite in the other direction). At 910 15th Street the 1310 wave will connect to the Denver Abilene router's PoS OC-12 interface, and the 1550 GigE wave will drop back into ICG's fiber to the Level3 colo.

Caveats and unknowns:

There will still be a few unknowns even after these tests are complete. Caveats include:

Power Meter testing summary:

Power-meter testing has shown that two back-to-back VCWDM modules have a very reasonable insertion loss to pass-through waves of 0.4-0.9 dB and attenuate the blocked wave at least 23.9 dB.
These WDMs and most optical gear functions bidirectionally. If light passes in one direction, light will normally also pass in the other too. This means that the two multiplexed waves can travel the same direction (parallel) or in opposing directions (anti-parallel). Both configurations should work for data transport, but in the parallel configuration enough misdirected light (about -19 dB in the 1550 -> 1310 case) is emitted from the wrong egress port to maintain a false link up condition in the recieving GBIC or a Section LOF (Loss of Frame) condition in a recieving SONET interface. This effect is analyzed below and described more fully in the link test section.

Power Meter testing details:

Tests used a Noyes OLS4 Optical Light Source and OPM5 Optical Power Meter. After zeroing both 1310 and 1550 wavelengths, the OPM5 measures 1.5 dB of 1550 when 1310 is being emitted, and -1.58 dB of 1310 when 1550 is being emitted. The power-meter is not a OSA - its detector is broad-band and internally calibrated for response at different wavelengths. Simply unplugging and replugging connectors can change measurements 0.3-0.6 dB.

Demultiplexing efficiency:

A test of each WDM individually, this test measures the quality of the WDM. Ideally, all the light injected would be recieved at the desired port, and none would be recieved at the other port, i.e., when 1310 nm ligh t is injected there should be zero attenuation to the 1310 port and infinite attenuation at the 1550 port.

WDM Unit 1:
Wavelength injected at common muxed input:
1310 power measured
-0.9 dB
-24.0 dB
1550 power measured
-17.0 dB
-1.1 dB

WDM Unit 2:
Wavelength injected at common muxed input:
1310 power measured
-0.4 dB
-25.3 dB
1550 power measured
-16.7 dB
-0.4 dB

Transmission, reflection, and rejection of waves in a production configuration.

These tests measure the power of wavelengths being carried between the two desired ports as well as the reflected and misdirected power being recieved on the other wave's ports. Ideally, zero power will be reflected or misdirected, so all the power will be transmitted to the desired port.

1310 nm injected at 1310 port
WDM1 -> WDM2
WDM2 -> WDM1
1310 power transmitted
-1.6 dB
-1.5 dB
1310 power reflected out 1550 ingress
-42.9 dB
-41.4 dB
1310 power misdirected to 1550 egress
-26.7 dB
-28.4 dB

1550 nm injected at 1550 port WDM1 -> WDM2
WDM2 -> WDM1
1550 power transmitted
-0.7 dB
-0.6 dB
1550 power reflected out 1310 ingress
-36.7 dB
-37.4 dB
1550 power misdirected to 1310 egress
-19.0 dB
-18.9 dB

Link error-rate testing:

Broadly, two different systems were tested. The first was a simple test comprising only Gigabit Ethernet. The second was a more complicated rig where a 1310 nm OC-12 was multiplexed with a 1550 Gigabit Ethernet signal in a setup most similar to the one planned to be deployed. Both rigs  were tested an a parallel and ant-parallel configuration. No link or data errors were detected in any of the tests, parallel or ant-parallel, using SONET or1310 nm GigE. In all tests, two hosts were connected by a path that passes thru the 1310 path in one direction and the 1550 path in the other. IPERF was used to generate a large amount of traffic over the links.

Simultaneous GigE over 1310 and 1550 nm

Cisco LX and CWDM GBICs were used as 1310 and 1550 nm transmitters, respectively. Each GBICs' signal can be recieved by the other because the GBICs use wideband receivers. The LX and CWDM GBICs were connected (looped) to each other via waves multiplexed thru a single fiber with WDMs on either end. To simulate a path, an air-gap attenuation of 8-9 dBs was added between the two WDMs. This is probably a more demanding test then using an OC-12 on the 1310 because an interferring GigE signal is maximally-bad "noise" for another GigE signal. Over 700 gigabytes were transmitted without error over both wavelengths.
LX and CWDM GBIC WDM test rig
Measuring with the NOYES power-meter, the LX GBIC transmitted at -6.5 dBm and the CWDM GBIC transmitted at 3.5 dBm that was attenuated down to -6.5 dBm with a 10 dB attenuator. Balancing the power of the two signals gives the system the best chance of working correctly - each interfering signal will be attenuated about 25 dB below the data signal. A more powerful signal will leak into the less powerful one most easily in the parallel configuration. In this configuration, enough signal gets through to maintain link even when the more powerful signal is no longer transmitted. The anti-parallel configuration is recommended because it behaves most like two independant patch cables.

1310 nm OC-12 multiplexed with 1550 nm Gigabit Ethernet

This test used a similar setup to the previous test. The 1550 GBIC is still used to transmit to a LX GBIC. Cisco LS1010 OC-12 single-mode PAMs are used to source and sink the 1310 nm wave. While tests were performed, neither the SONET layer nor the GigE layer reported any errors in either the parallel or anti-parallel wave configuration. Over 200 terabytes were transmitted without error.
OC-12 and GE WDM test rig


These WDMs should work effectively for the intended application. Care should be taken to balance signal levels, and the signals should be transmitted towards each other across the WDMs in an anti-parallel configuration.
Address comments or questions about this Web page to the Network Engineering & Telecommunications Section (NETS) by opening a ticket at The NETS is part of the Computational & Information Systems Laboratory (CISL) of the National Center for Atmospheric Research (NCAR). NCAR is managed by the University Corporation for Atmospheric Research (UCAR). This website follows the UCAR General Privacy Policy and the NCAR/UCAR/UCP Terms of Use. The National Center for Atmospheric Research is sponsored by the National Science Foundation (NSF). Any opinions, findings and conclusions or recommendations expressed in this material do not necessarily reflect the views of the National Science Foundation.