The Future: Gigabit Ethernet and Beyond
Gigabit Ethernet Networking
Author: David Cunningham; Bill Lane
This is the chapter where we get to wrap up the book“stick
our necks out”and predict what the future may or may not bring
to the world of Ethernet LANs. This chapter considers three types of “futures”:
Current Ethernet development projects that are likely to result
in supplements to IEEE 802.3-1998, including:
Emerging technologies and non-standards-compliant commercial
products that may or may not be considered for future standardization, but
that are likely to impact future Ethernet LANs and LAN upgrades. These technologies
Commercially available extended-length Gigabit Ethernet links.
Wavelength Division Multiplexing (WDM): A way to support very
high transmission rates with lower-cost laser-based transceivers.
Combining link extenders and WDM with routing switches: An
economical alternative to ATM/SONET and SONET/SDH MANs and WANs.
Some issues relating to what is needed to make the next 10X
jump in the Ethernet transmission rate progression include:
Line coding: Several possible options for use at 10,000 Mbps
Full-duplex operation: Should higher transmission rate links
be limited to full-duplex operation only?
The optical layer: Is a new lower physical layer needed at
the bottom of the Ethernet reference model, particularly for MAN- and WAN-based
We realize that some of the content of this chapter may be speculative
and that we may be pushing the envelope in some areas. But isn't that what
the future is about? The purpose of this chapter is to raise some issues that
we think are important, to emphasize that local area networking is a rapidly
evolving technology, and to get you thinking about what is likely to be possible,
as well as what is currently available.
We will begin with the most certain futures, the projects already officially
under way or nearing completion by the 802.3 standards group.
(IEEE 802.3ab) defines a new media-dependent physical layer (see Figure15.1)
that will support 1,000 Mbps operation over a 100 m link consisting of 4-pair,
Category-5 or better, balanced, unshielded twisted pair (UTP) copper cable.
The purpose of this new physical layer is to provide an upgrade path for current
100BASE-TX networks and to extend the life of existing Category-5 UTP cable,
the most commoninside cable in use. IEEE 802.3ab should be approved as a supplement
to IEEE802.3-1998 sometime during the spring of 1999.
1000BASE-T PHY sublayers and signal transmission procedure is different
from the PHY sublayers and transmission procedures described for 1000BASE-X
in previous chapters. Instead of using two simplex links (two optical fibers
or two wire pairs) to form one full-duplex link, 1000BASE-T sends encoded
signals simultaneously in both directions on the same wire pair, as shown
in the 4-channel link topology of Figure
This means that the signal at the MDI
is the sum of the just-transmitted signal and the signal being received from
the entity at the other end of the link. It also means that thereceiver must
be more than an amplifier and signal detector. The transmitted signal that
is added to the received signal at the MDI must be removed (canceled) before
any amplification and detection of the received signal can take place.
shows a simplified block diagram of the PCS/PMA data circuits. The PMA
hybrid transceiver consists of the following:
A normal Transmit function with four independent transmitters
A special Receive function, with four independent receivers
where the inverse of the transmitted signal is added to the composite signal
from the MDI to cancel the transmitted signal
A clock recovery unit to recover the clock from the received
The actual transmitted signal on each wire pair is a 5-level (+2, +1,
0, -1, -2) pulse amplitude modulated symbol (PAM5). Four symbols,
transmitted simultaneously on the four wire pairs, form a code-group (4D-PAM5)
that represents an 8-bit frame octet. The 4D-PAM5 coding allows an aggregate
1,000 Mbps data rate to be achieved with a transmission rate of only 125 Mbaud
per pair (baud is the transmission rate in symbols/second). Transmitter pulse
shaping and receiver equalization are used to compensatg for the spectral
characteristics and signal distortion of the link.
A continuous stream of Idle symbols (from a restricted set of values [+2,
0, -2]) is sent whenever regular transmission is not in progress to maintain
continuous clock synchronization between the two link partners. During Auto-Negotiation,
one partner is chosen to be master (typically the multiport partner)
and the other to be slave. Subsequent transmission from the master
is then timed by the master's local clock, and transmission from the slave
is synchronized by the clock that is recovered from the received symbol
stream, as shown in Figure
15.4. This form of synchronization is known as loop
Encoding of the outgoing frame octets and decoding of the received
4D-PAM5 code groups occurs in the PCS sublayer. The 1000BASE-T coding procedure
was chosen for several reasons:
Four-Dimensional 8-State Trellis Forward Error Correction
coding (this is the 4D of the 4D-PAM5 code designation) enables the system
to operate as if the signal-to-noise ratio at the analog-to-digital converter
in the receiver had been improved.
Encoding an octet as four 5-level PAM5 symbols allows the
entire octet to be transmitted in one symbol (baud) period, and reduces the
symbol rate to 125 Mbaud on each pair.
Separate scramblers for the master and slave PHYs randomize
the symbol sequence, reduce the spectral lines in the transmitted signal,
and create essentially uncorrelated data symbols between the two opposite
travelling symbol streams on each wire pair. All of these aid in symbol recovery.
While 1000BASE-T is designed
to operate over existing Category-5 cabling, all cables should be tested before
they are used for Gigabit operation. The Gigabit Ethernet Alliance has estimated
that up to 10 percent of the existing Category-5 cables may not be able to
pass two critical performance parameters that were not established when Category-5
cabling initially became available:
Return Loss. Return loss
refers to the magnitude of the signal power returned to a test point as a
result of reflections that are caused by cabling impedance mismatches at the
Far-End Crosstalk (FEXT).
Crosstalk is noise coupled onto the wire pair by signals on adjacent
pairs in the cable (see Figure
15.5). FEXT is noise induced at the far end of the cable from the
transmitter. FEXT can be exacerbated by the use of non-standard connectors
and also by incorrectly installed connectors (such as connections where
the wire pair has been untwisted more than necessary).
problem arose quite innocently. The early Category-5 cable installations were
used for 10BASE-T. However, because 10BASE-T operates at much lower frequencies
than either 100BASE-TX or 1000BASE-T, return loss and FEXT were not a critical
problem (remember that 10BASE-T was originally designed to operate over ordinary
telephone twisted-pair cable).
While the cable was available earlier, Category-5 installation instructions
were not specified and agreed upon until 1995 (see ANSI/TIA/EIA 568-1995),
the same year the first versions of Fast Ethernet were approved. Category-5
connectors were also not available until 1995, and, as a result, Category-5
cables installed before then were typically installed with Category-3 connectors.
Therefore, 1995 is a calendar benchmark and any pre-1995 Category-5 cable
will not have standards-compliant hardware unless new connectors have been
The return loss and FEXT problem was not fully recognized until 100BASE-T2
was being developed (100BASE-T2 was approved in 1997). As things stood in
The Telecommunication Industries Association (TIA) was developing
a technical service bulletin to define new cabling practices, new return loss,
and FEXT measures that will be published as ANSI/TIA/EIA-TSB-95 for an Enhanced
Category-5 cable (Category-5E).
The International Organization for Standardization (ISO) was
adding return loss and FEXT measures to ISO/IEC 11801 for Category-5 cabling
(which ISO will continue to call Category-5 rather than Category-5E).
TIA was developing performance, installation, and testing
specifications for Category-6 cabling with a specified bandwidth of 200 MHz
(Category 5 has a specified bandwidth of 100 MHz). Because Category-6 specifications
are still in development, not all currently available Category-6 connectors
ISO had begun development of performance, installation, and
testing specifications for a Category-7 cable with a 600 MHz bandwidth. The expected completion date for this
category cable is sometime in 2000.
That is all well and good, but what do you do with existing Category-5
cabling? How can you tell whether or not it will support Gigabit operation?
What can you do if it doesn't?
If the current cabling supports 100BASE-TX operation, it is likely to
support 1000BASE-T operation. However, because 100BASE-TX uses only two of
the four wire pairs, the cable should still be tested to determine whether
or not it is suitable for 1000BASE-T operation.
If the current cabling is not being used for 100BASE-TX operation, it
will definitely need testing to determine whether or not it is suitable for
Basic cable testing procedures are defined in ANSI/TIA/EIA-TSB-67, and
a number of suppliers now have new Category-5 test equipment and/or retrofit
upgrades for existing test equipment that will perform the tests specified
If the current cabling does not pass the return loss and FEXT tests,
modifying the link configuration may still bring the cable into compliance
(see the sidebar, “Modifying Category-5 Link Configurations for 1000BASE-T
Modifying Category-5 Link Configurations for1000BASE-TOperation
and 15.7 show
the range of link configurations that may be used for 1000BASE-T. All
1000BASE-T link configurations should be at or between the maximum and
the minimum shown.
15.6. Maximum 1000BASE-T Category-5 UTP Link Configuration
Figure 15.7. Optimized 1000BASE-T Category-5 UTP Link Configuration
The following list provides the corrective actions recommended for links
that do not meet the return loss and FEXT requirements. (There is no implied
order to the action list. We listed those that would likely cause the least
disruption to area activities first. Remember that the link should be retested
after each action is taken.)
If the link has a cross-connect, reconfigure it as an interconnect.
If the interconnect connectors do not meet Category-5E specifications,
replace them with connectors that do.
If the telecom closet equipment cable does not meet Category-5E
specifications, replace it with a cable that does.
If the work area equipment cable does not meet Category-5E
specifications, replace it with a cable that does.
If the work area telecom outlet connector does not meet Category-5E
specifications, replace it with a connector that does.
If the transition point connector does not meet Category-5E
specifications, replace it with a connector that does, or better yet, if sufficient
slack exists in the link, connect the transition point end of the link directly
to the telecom outlet connector.
What if you are not able to correct problems with some of the cables?
Does this mean the cables will have to be replaced? The answer is usually
a resounding no! Remember that most networks will have a mix of 10 Mbps, 100
Mbps, and 1,000 Mbps services. The problem cables will still support 10BASE-T
and 100BASE-T2 with two pairs, and will also support 100BASE-T4 so long as
four pairs are available, even if they won't support either 100BASE-TX or
One of the primary architects of OpenCable, Michael
Adams, explains the key concepts of this initiative in his book
Broadband, Second Edition
by George Abe
Introduces the topics surrounding high-speed networks
to the home. It is written for anyone seeking a broad-based familiarity
with the issues of residential broadband (RBB) including product
developers, engineers, network designers, business people, professionals
in legal and regulatory positions, and industry analysts.