radio, building security, avionics for automated
landing, and many more. A Spectrum Analyzer is needed to
determine the frequency characteristics and the magnitude of
the noise signals picked up by the cabling links. This
laboratory type device can identify the frequencies of the
signals (noise) on the link as well as the relative magnitude
of the signals at the different frequencies. Alien cross talk
is noise induced by other cables in the same pathway. The
detection is the same as those mentioned above. Anytime a UTP
link is tested in a bundle of cables in which some links are
active, the chances are very good that the tester will detect
alien cross talk, especially when the traffic is 100BASE-T. The
tester will report the “external noise detected” message.
Typically, this alien cross talk will not impact the
reliability of the network traffic. Excessive cross talk is
usually reported in the Near End Cross Talk (NEXT) test results
on cable testing devices. NEXT can originate either inside or
outside the link. Crosstalk originating nearest the
transmission source is usually the loudest and has the greatest
amplitude or leaks into the measured pair. To minimize cross
talk, ensure that the cable, at each termination point, is left
untwisted for no more than the allowed 13 mm (0.5 inch).
Crosstalk becomes correspondingly worse as more cable is
untwisted. The cross talk occurring outside the link is
actually cross talk from other adjacent cables, or noise from a
variety of sources. Noise sources include nearby electric
cables and devices, usually with high current loads. These may
include large electric motors, elevators, photocopiers, coffee
makers, fans, heaters, welders, compressors, and so on. Another
less obvious source is radiated emissions from transmitters.
This would include TV, radio, microwave, cell phone towers,
hand-held radios, and anything else that includes a transmitter
more powerful than a cell phone.
Content 3.2
Characteristics of Physical Layer Optimization Problems
3.2.4 Collisions Collision domain problems
affect the local medium, and disrupt communications to Layer 2
or Layer 3 infrastructure devices, local servers or services.
They typically result from the following problems: - Bad
cables
- Marginal or intermittent workstation NICs
- Marginal or intermittent ports on hubs or
switches
- Errors or excessive traffic on the local
collision domain
- Duplex mismatches
- Electrical
noise and other environmental disruptions
Collisions
are normally a more significant problem on shared media than on
switch ports. Average collision counts on shared media should
generally be below 5 percent, though that number is
conservative. Be sure that judgments are based on the average
and not a peak or spike in collisions. If the average
utilization is high (sustained peaks in excess of 60 percent
for shared media, and in excess of 80-90 percent for switched
links) and collision counts are acceptable (average is below 5
percent for shared media, and below 1 percent for switched
links), then the network may simply be saturated. There may be
too many stations transmitting within this collision domain, or
the network architecture may need optimizing for shorter
distances between distant stations. Physical layer or Layer 1
network devices act to increase collision domains, effectively
increasing the scope of a potential problem from bad to worse.
Layer 1 devices such as repeaters and hubs act to regenerate
signals and therefore increase the maximum distance over which
a network can operate. Hubs additionally act to increase the
collision domain. They contain from four to forty-eight ports
for station connectivity. A hub will regenerate any signal
received and resend it out all active ports. Be very careful
when troubleshooting collision problems because the obvious
answer is usually wrong. The addresses found in collision
fragments belong to stations that transmitted legally. Stations
that sent enough of the current frame to have a source address
in a collision fragment usually started transmitting first,
though that depends in part on the monitoring point within the
collision domain. Most of the stations that collide with those
legally transmitted frames are also operating legally. They did
not “hear” anything on the wire, so they began to transmit. If
there is a station that has gone “deaf” and is stepping on
other transmissions because it does not hear them, it will
probably never be discovered because its frame collides with
another transmission and its data is always corrupted.
Troubleshoot the presence of too many collisions, but don’t
examine the fragments closely. Using the corrupted data from
collision fragments will just cause frustration. Late
Collisions
A late collision is counted when a collision
is detected by a device after it has sent the 512th bit of its
frame. No more than a few late collisions should ever occur in
any environment. If a device is incrementing a collision
counter, further investigation is needed as a significant
problem is occurring. If the number of late collisions is
occurring at a steady rate, performance degradation may be
noted.
Any of the following conditions may be causing late
collisions: - Incorrect configuration
- Duplex
mismatch (one host operating at half-duplex while another host
is operating a full-duplex)
- Faulty cabling
- Faulty hub or shared media device
- Faulty NIC or
switch port
- Excessive network traffic beyond the
limitations of the shared media hub or switch port
Note: Collisions and late collisions should never occur on any
router, switch, or NIC port operating a full-duplex. Some Cisco
Catalyst switches, such as the Catalyst 6000/6500, will disable
ports on excessive late collisions, even when operating a
half-duplex. The old attribution that a late collision is
literally the result of a too-long cable is challenged by
calculating the delay introduced by the cable only. While delay
introduced by the cable is a factor, it is much more likely
that other factors such as cable impedance mismatches, signal
attenuation along the cable, too many repeaters, and marginal
interfaces result in situations where a shortened cable
eliminates late collisions.
Content 3.2
Characteristics of Physical Layer Optimization Problems
3.2.5 Other data transmission issues Short
Frames
The most likely cause of a short frame is a
faulty card, or an improperly configured or corrupt NIC driver
file. Jabber
Jabber, is often defined as the
condition in which a network device continually transmits
random, meaningless data onto the network. IEEE 802.3 defines a
jabber as a data packet whose length exceeds the standard.
These packets are called long frames. Cyclically lock the port
out, then check later to see if it is ok. The standard says
that after the jabber timer expires (20,000 to 50,000 bit
times) then the hub should close the port for awhile before
reopening the port to see if the attached device has stopped
transmitting. If jabber is again detected the port may be
closed for another cycle. This may continue forever. In general
jabbers are not a common occurrence. The most likely causes of
jabber are a faulty NIC and/or faulty or corrupt NIC driver
files, bad cabling, or grounding problems. Ghosts
Ghosts are easily created by a variety of causes on coaxial
Ethernet. They may also be caused by something as simple as
installing a second crossover cable between two hubs on half
duplex 10BASE-T. The parallel path sometimes causes very
strange symptoms. The hub will do one of the following two
things: - Lock one port out, sometimes requiring a power
cycle or SNMP management intervention to reopen the port. This
is fortunately the most common result.
- Allow the
error to continue uninterrupted. This is not permitted by the
standard.
The error level often fluctuates between
very little and the available bandwidth for no apparent reason.
In addition to ghosts, this second crossover cable fault