Why should twisted-pair cabling, used in large quantities to carry 4-kHz voice signals in the phone system, suddenly work well at 100 MHz and beyond? The simple answer is that it doesn’t. All twisted-pair cables aren’t the same.
Telephone cables contain pairs of No. 24 (AWG) copper wires with only a few twists per foot and either paper or plastic insulation. The nominal impedance is between 600 and 900 W in the voice-frequency range up to 4 kHz. At 1 MHz, the impedance is nominally 100 W, and voice service testing often is done at 772 kHz. The wire resistance typically is only a small part of the impedance, most of it being contributed by the distributed inductance and capacitance.
Recently, as attention has been focused on various digital subscriber line (xDSL) transmission schemes intended to provide broadband data communications via telephone cabling, the subject of load coils has been highlighted. By itself, a telephone cable severely attenuates higher frequency voice signals because of the low-pass filter formed by its resistance, distributed shunt capacitance, and series inductance. Connecting inductors across the line at specific locations can peak the response within the voice band.
The value of a load coil typically is 66 or 88 mH depending on the telephone-system design. In a commonly used approach, coils are connected every 6,000 ft starting 3,000 ft from the central office. Their effect is to reduce attenuation up to 4 kHz, but they also increase attenuation at high frequencies. This is the reason they must be removed for xDSL, for example, to allow operation over reasonable distances at frequencies above 1 MHz.
Data cables are designed to operate at much higher frequencies than telephone cables. Much of the construction of the two cable types appears to be similar, but the differences are significant.
Cable Construction
Data cables are classified according to the speed of the signal transmission they are guaranteed to support. Category (CAT) 3 is rated up to 16 MHz, CAT 4 to 20 MHz, CAT 5/5e to 100 MHz, proposed CAT 6 to 200 MHz with some tests required at 250 MHz, and proposed CAT 7 to 600 MHz. Actually, because application speeds have increased so quickly, most cabling done recently and continuing today uses CAT 5 or enhanced versions of the standard, CAT 5e. Very little CAT 4 is used.
Speed is affected by attenuation and crosstalk, both of which increase with frequency. The higher number categories, CAT 5, CAT 5e, and the proposed CAT 6, extract more and more performance from the basic twisted-pair medium. For example, higher speed cables generally have more twists per inch than slower ones. However, the specifications only call out the required performance. It’s up to the manufacturers to achieve it.
Attenuation could be reduced by using larger diameter wire. But balanced against cost, the common 100-meter limit on cable length in LAN applications, overall cable diameter, and ease of installation, No. 24 wire is the norm with No. 22 optional. A 100-W characteristic impedance is standard, the stability, consistency, and accuracy of this value affecting data-transmission reliability.
Impedance measurements made on all four pairs in a minimally compliant CAT 5/5e cable are plotted in Figure 1a. Although it’s obvious that the impedance varies well beyond the allowed ±15% limit, the specification permits the measurements to be averaged, so this cable does meet the CAT 5/5e requirements. Figure 1b shows the result of the same test run on high-quality cable. Even within a CAT rating, products can vary significantly in ways that will affect network performance.
Capacitance also contributes to attenuation and is a function of the dielectric constant of the insulation material, the thickness of the insulation, and the number of twists per inch. All of these quantities are interrelated and governed by the need to maintain the 100-W impedance. A higher dielectric constant implies higher capacitance, so materials such as foamed polyethylene and fluorinated ethylene propylene (FEP) or Teflon™ often are used.
FEP is the favorite material for plenum-rated cables, that is, fire-retardant cables that can be run through air-conditioning and heating ducts. FEP has a long history of excellent service in difficult environments, such as when exposed to heat, sunlight, and rain, and exhibits a very low dielectric constant and dissipation factor of 2.1 and 0.0005, respectively.
In addition to the insulation covering the individual wires, an overall protective sheath is used. The sheath material affects performance because the field associated with signals on a twisted pair extends into the sheath. The two wires in a pair are driven differentially, and they are twisted tightly together, but the fields don’t entirely cancel. Twisted-pair cable that contains from one to several pairs, but without shielding, is termed unshielded twisted-pair (UTP) cable. Similarly, shielded twisted-pair cable is known as STP.
Shielding greatly changes the characteristics of cables. Mechanically, shielding complicates cable termination, adds to a cable’s size and weight, and increases the minimum bend radius—you can’t make such tight bends with shielded cable. Electrically, shielding considerably reduces crosstalk among pairs in a cable, a major performance limitation in UTP. Conversely, crosstalk can be compromised by the wire terminations made to connectors in a shielded system.
So-called ScTP is UTP cable with an overall screen. Screening reduces a cable’s susceptibility to interference from external noise sources. It has little effect on crosstalk among twisted pairs within the cable because they are not screened from each other. SSTP is cable comprising a number of shielded twisted pairs within an overall screen and sheath. Proposed CAT 7 specifies SSTP cable.
Coupling from one pair to another has been minimized by cable construction. Using slightly different twist pitches from one pair to another in a cable means that wires in adjacent pairs don’t have the same proximity to each other throughout the length of the cable. This technique is standard practice in data cables and has its origin in the manufacture of telephone cables with as many as 25 pairs.
Besides twisting the two wires in a pair, the four pairs typically making up a data cable are twisted around each other in a specific manner. The objective is to reduce manufacturing variability by establishing a precise relationship among the pairs in a cable.
Belden Wire & Cable has developed a technique of bonding together the insulation of the two wires in a pair. This approach eliminates any separation of the wires that installation and bending might cause, which would disturb the pair’s impedance locally and increase the return loss.
Belden also produces cable in which the four pairs are separately housed in channels within the overall sheath. Other manufacturers use a cross-shaped separator that runs the length of the cable to provide four distinct compartments. Both techniques serve to improve repeatability from batch to batch. More importantly, performance after installation correlates to that on the reel.
Practice…
STP vs. UTP is a debate gaining momentum as data transmission speed increases. Some companies favoring UTP claim that if CAT 7 cable is incorrectly installed, its performance will be worse than CAT 5. Ground loops, current flowing along a shield between grounds at different potentials, can inject noise into the wires that the shields are intended to protect.
One factor contributing to lack of progress in the adoption of CAT 7 is the confusion caused by the manufacturer-specific nature of actual CAT 6 installations. In practice, CAT 6 cables, connectors, patch panels, and related products cannot be mixed with those from another manufacturer without degrading system performance. One effect of the subtle differences among components is to cause impedance mismatches that generate reflections and affect return loss.
Philip Lippel, a technical staff member in Agilent Technologies’ WireScope Operation, said, “Structured cabling systems must be generic. Telecommunications cabling is part of a building’s infrastructure and should be designed to outlast any individual application.
“Newer cabling categories are specified to higher frequencies,” he continued, “but that’s not the only difference. Crosstalk, attenuation, and return loss, for example, also improve with each step up in category. It’s up to the distribution designer and installer to ensure that the user gets reliable, predictable performance. This entails good handling, termination, and layout practices followed by comprehensive field testing to guarantee that the assembled system meets the requirement.”
Usually, you think of a discontinuity such as a connector causing increased return loss. It can just as readily be increased by a local variation in cable impedance, for example, due to a slight untwisting of the wires in a pair. Lower return loss means lower self-generated noise within the network, which may lead to fewer retransmissions and greater efficiency.
Return loss also is a function of the temperature gradient that exists along a cable. There may be as much as a 50°F temperature differential on a cable run that travels through hot ceiling-level trays on its way from an air-conditioned wiring cabinet to the connected apparatus. Especially in marginal systems, changes in temperature can affect the error rate.
Actual twisted-pair impedance relative to its nominal 100-W value can vary within a range according to the category rating. For example, CAT 5 and 5e can vary by ±15%, but fully shielded CAT 7 by only ±3%. In addition to enabling tighter impedance control, shielding improves crosstalk.
To make sure that your new cable performs like that in Figure 1b, find a manufacturer or distributor with comprehensive test capabilities. Anixter is such a distributor, selling cables and accessories from several manufacturers but made to Anixter’s enhanced CAT specifications. Performance is verified by subjecting products to a suite of marginal waveforms gathered from actual customer sites. This testing takes place in the company’s multimillion dollar Chicago laboratory.
…And Theory
When considering a cable’s rejection of common-mode interference, the two conductors in the pair are treated as a single wire. The ratio of the power coupled to the wires relative to that incident on the cable defines the shielding attenuation, as.
This is a measure of the effectiveness of the cable shield. Equation 1 also applies to the attenuation of the common screen in a cable with several twisted pairs.
Normally, a twisted pair carries signals differentially in a balanced mode of operation. Common-mode signals ideally are equal on each conductor and will be rejected at the line receiver. However, because the cable is not perfect, some of the common-mode signal will appear as a differential signal and add to the system noise. The ratio of the common-mode signal to the resulting differential-mode signal is a measure of the common-mode rejection, au.
where: icm = common-mode current
idm = differential-mode current
The capacitance to ground of a twisted pair is increased by a shield, but the added capacitance is very stable and improves the balance of the twisted pair with respect to an interfering common-mode signal. A shield both attenuates interference because of its screening effect and improves the common-mode rejection of the twisted pair.
Near-end crosstalk (NEXT) is a measure of the rejection of signals from one twisted pair by another twisted pair in the same cable at the transmitting end. A more severe test is power-sum NEXT (PSNEXT) that subjects one pair to the signals transmitted on the other three pairs in the same cable.
Older Ethernet LANs used one twisted pair for transmission and a separate one for reception. In these systems, NEXT was the important parameter. Newer systems, such as 1,000Base-T Ethernet, use all four pairs in a cable and transmit and receive simultaneously on each pair. Now, the return loss becomes critical because reflected power adds noise directly to the receive signal.
The authors of a European research paper that compares shielded vs. unshielded twisted-pair performance concluded, “To ensure [meeting] the NEXT limits of the CAT 6 standard, there is no way to avoid a shield around the twisted-pair cable.”1 However, according to Agilent’s Mr. Lippel, “Shielded telecommunications cabling is not very common in the United States. CAT 5e and CAT 6 still have a lot of headroom for today’s common applications.”
Transient Effects
All data cables are designed to exhibit a certain characteristic impedance: usually 100 W for twisted pair, 50 W for most coaxial cables, and 75 W for video cables. But, the impedance is not constant. In addition to the temperature coefficient of the copper wire, there are skin effect and dielectric absorption that increase with higher frequency.
Skin effect changes as the square root of frequency, and dielectric absorption changes linearly with frequency, both increasing a cable’s internal losses. The result is a complex, frequency-dependent pulse response. Silver-plated center conductors often are used in high-frequency coaxial cables to reduce skin effect. But, in the cost-sensitive data-network environment of twisted-pair data cables, this is not an option.
A typical cable pulse response (blue) is shown in Figure 2 (see November 2001 issue of Evaluation Engineering) compared to the response (red) of a simple resistor-capacitor (R-C) network. The curves have been normalized to cross at 63%, the value of the R-C curve after one time constant. The red line is within 0.01 of the final value of 1.0 after five time constants. The blue line shows the very slow so-called dribble-up caused by the cable’s skin effect.
In practice, although a cable may be adequate to transmit the fundamental data rate, higher harmonics will be lost. The result is a reduction in pulse amplitude as well as appreciable rounding of pulse corners, potentially leading to errors in data reception.
These changes in pulse shape become important at higher data rates or over long lengths of cable. They are yet more reasons that reliably achieving very high data rates should not be undertaken lightly.
Reference
- Garbe, H., Kärst, J. P., and Knobloch, A., “Shielded or Unshielded Twisted-Pair for High-Speed Data Transmission?,” Proceedings of the IEEE EMC Symposium, 1998, pp. 112-117.
Additional Reading
Giacoletto, L., “Pulse Operation of Transmission Lines Including Skin-Effect Resistance,” Microwave Journal, February 2000, pg. 150.
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2001 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
November 2001