About ten years ago,
Quabbin Wire conducted tests to compare the alien crosstalk performance of Cat
5e round four pair cabling versus "flat" cabling constructions. Results
showed poor alien crosstalk performance of the flat cables but confirmed that
1000Base-T Ethernet can be used on Category 5e channels with proper cabling.
This article reviews that data and updates the evaluation of 10 Gig Ethernet using
Category 6A channels.
History and
Background
1000Base-T
(Gigabit Ethernet) and 10GBase-T (10 Gig Ethernet) use
bi-directional signaling on each pair of a four pair copper cable. Because these
protocols are signaling at high data rates and low signal levels, they are particularly
susceptible to even very low noise levels. For this reason, the IEEE
802.3an Ethernet subcommittee identified “Alien Crosstalk” as a critical noise
parameter for successfully implementing Gigabit and 10 Gig Ethernet on copper
balanced pair cabling systems.
Alien crosstalk (AXT) is defined in more detail below;
however, it is generally un-cancelable noise or unwanted signals in cables
within a cable bundle or group of cables. In June 1998, Quabbin Wire & Cable
Co., Inc. performed the tests shown in this report to measure the AXT noise
levels of typical round four pair cabling and some of the alternative “flat”
cabling constructions then being introduced. Both these cable designs met
Category 5e requirements at the time. The AXT test data was later confirmed by
others with the determination that 1000Base-T Ethernet could be run on the
installed base of round UTP Category 5e cabling, but not if using the “flat”
design.
Early in 2003 the IEEE formed the 802.3an
Study Group to evaluate the feasibility of running 10 Gig Ethernet on copper
media. They asked the TIA to define and quantify worst-case AXT values for
Category 5e and 6 cabling. The TIA 42.7 Copper Committee identified the worst
case as the power sum AXT on a pair within the center cable of a seven cable
bundle (six grouped around one) with all seven cables signaling. During almost
three years of modeling and testing, it was determined that 10 Gig Ethernet
could not be economically implemented on a worst-case 100-meter Category 5e or 6
channel. However, an “Augmented” Category 6A copper channel was defined in 2006
that will support 10 Gigabit Ethernet at 100 meters.
Using data developed
by the TIA 42.7 committee, the IEEE 802.3an working group was able to define the
10 Gig signaling method and solve many other technical problems. They issued
their standard in mid-2006, and hardware is now being marketed for signaling on
copper. Quabbin Wire is proud to have participated in the IEEE and TIA 42.7
effort.
As a result of studying alien crosstalk’s effects on
channel performance, Quabbin Wire & Cable Co., Inc. developed stranded patch
cable designs that minimize this noise and optimize channel performance. For
more information on these products contact Quabbin’s sales department, refer to
the news article
posted on this site, or read the White Paper
that illustrates Quabbin’s most
recent AXT test data.
Definition
Alien crosstalk is the data interference induced from a pair in one cable into the pairs of an adjacent victim cable. Since this "noise" is not a signal generated by the victim, transceiver cancellation circuitry cannot be "trained" to cancel it out from the intended data stream. For this reason, limits were developed for alien crosstalk between cables in a bundle so that the noise level budgets of future protocols will not be exceeded.
Test Conditions
One hundred meters each of two lengths of round Cat 5 cable were taped together approximately every two feet. About 10 meters of the cable remained untaped at each end. The cable was on a dry wooden floor. Measurements were taken from .772 to 200 MHz. The cable was allowed to separate naturally between the taped points. The flat cable was an "enhanced" or "Cat 5e" design intended for applications beyond 10 Base T, such as bi-directional signaling. The flat cables were intended to lay one on top of the other, but they typically oriented themselves to be slightly offset due to the fact that they are actually somewhat oval in cross section. The flat cables were taped and installed as the round cables above.
The limit lines illustrated for NEXT are associated with Cat 5 and are 6 db improved per the proposed Cat 5e hybrid cable specification (with 2 cables). The ELFEXT limit is the proposed Cat 5e (63-20 log (f/0.772)) line limit.
The following cross sectional drawings illustrate the two cable constructions.
 |
 |
| "Flat" or "Oval" Cabling | Round Cabling |
Conclusions
The poor alien crosstalk performance of the flat cables is due to a design problem inherent with flat constructions. In round cables, each pair has a different degree of twist. This variable twisting from pair to pair minimizes crosstalk between pairs within a cable. The pairs are also twisted around the cable's axis. This makes it impossible for a pair in one cable to be oriented parallel to a pair in an adjacent cable that has the identical degree of twist.
By contrast, even though flat cable pairs also have a variable degree of twist from pair to pair, they are all parallel to the cable axis. Thus, bundles of flat cable have a high probability that similarly twisted pairs in adjacent cable will be oriented parallel to each other in close proximity, resulting in relatively poor alien crosstalk performance. Although we did not test more than two cables bundled together, we have reason to believe that groups of three or more cables will result in even poorer relative performance.
 |
 |
 |
 |
 |
 |
 |
 |
To comment on this paper email us at engineering@quabbin.com.