ee times - getting the most out of your twisted pair cable

5/28/2018 EE Times - Getting the Most Out of Your Twisted Pair Cable

1/2/14 EE Times - Getting the most out of your twisted pair cable

www.eetimes.com/document.asp?doc_id=1274094&print=yes

Davor Glisic, National Semiconductor

Corp.

6/14/2008 01:00 PM EDT

1 Comment

Tweet 0 1

Design How-To

Getting the most out of your twisted pair cable

Moving vast amounts of data between two points quickly, reliably and economically is whatmany designers are facing regularly when building their systems. When it comes to using

copper cables as a transmission media in these data transport systems, twisted pair cables

present many advantages, one of them being low cost. The low cost incentive these cables

offer is very appealing, thus many designers are finding ways to employ them outside of

their traditional applicationdomain- Ethernet. Let's review relative performance bounds of

twisted pair cables, explore signal conditioning solutions (i.e. pre/de-emphasis, equalization)

semiconductor manufacturers offer to push these bounds out, sort out pros and cons of

each solution and understand when you need to consider enhanced but more expensive

cables.

Know its Twists

Setting performance bounds (e.g. how fast and how far one can transmit binary digitalsignals using these cables) for twisted pair cables requires thorough understanding of their

key electrical characteristics. One of them is cable insertion loss or cableattenuation.

Above approximately 1 MHz, its most dominant component is the skin effect loss which is

directly proportional to the square root of the frequency. Per ANSI/TIA/EIA-568-B.2

Category 5e standard, which defines currently the most common twisted pair cable, the

insertion loss is mathematically modeled with the following equation:

Insertion Loss = -{1.967 * SQRT (f) + 0.023 * f +0.05 / SQRT (f)} [dB / 100m]

Withthis equation, you can approximate insertion lossfor any CAT5e cable length at any

frequency (Figure 1). It becomes very handy when you quickly want to assess what kind ofloss your signal can expect.

Share
http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=domain&x=&y=http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=domain&x=&y=https://twitter.com/intent/tweet?original_referer=http%3A%2F%2Fwww.eetimes.com%2Fdocument.asp%3Fdoc_id%3D1274094%26print%3Dyes&text=EE%20Times%20-%20Getting%20the%20most%20out%20of%20your%20twisted%20pair%20cable&tw_p=tweetbutton&url=http%3A%2F%2Fwww.eetimes.com%2Fdocument.asp%3Fdoc_id%3D1274094&via=eetimeshttp://twitter.com/search?q=http%3A%2F%2Fwww.eetimes.com%2Fdocument.asp%3Fdoc_id%3D1274094http://www.eetimes.com/document.asp?doc_id=1274094&http://www.eetimes.com/document.asp?doc_id=1274094&http://www.eetimes.com/http://www.eetimes.com/http://www.eetimes.com/http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=cable&x=&y=http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=domain&x=&y=http://www.eetimes.com/http://twitter.com/search?q=http%3A%2F%2Fwww.eetimes.com%2Fdocument.asp%3Fdoc_id%3D1274094https://twitter.com/intent/tweet?original_referer=http%3A%2F%2Fwww.eetimes.com%2Fdocument.asp%3Fdoc_id%3D1274094%26print%3Dyes&text=EE%20Times%20-%20Getting%20the%20most%20out%20of%20your%20twisted%20pair%20cable&tw_p=tweetbutton&url=http%3A%2F%2Fwww.eetimes.com%2Fdocument.asp%3Fdoc_id%3D1274094&via=eetimeshttp://www.eetimes.com/document.asp?doc_id=1274094&




Figure 1. CAT5e Attenuation per TIA/EIA-568-B.2.

You can use the data from Figure 1 to estimate maximum data rate as a functionof cable

length. The estimates require knowing the criteria your transmitter and receiver mandate for

error free operation. For example, when building RS-485 transmission networks,attenuation of 9 dB at the frequency of 1/tUI in hertz, where tUI is a unit interval at a given

signaling rate, is an accepted industry guideline for determining maximum signaling rate.

For networks that use interfaceICs with reduced voltageswings for higher signaling rates

(i.e. LVDS, CML), attenuation of 6 dB at 1/tUI hertzmay be used as a general guideline

when determining maximum signaling rate for a given cable length. These guidelines

assume dc-balanced data, point-to-point links, zero crosstalk and pair-to-pair skew, and no

external interference.

Figure 2. How Far, How Fast.

Figure 2 illustrates the maximum signaling rate as a function of CAT5e cable length for RS-

485 and LVDS interfaces based on the 9 dB and 6 dB guidelines respectively. In Figure 2,

the flat segment of each curve is determined based on the cable's ohmic losses (typically 9

ohms per 100 meters for 24 AWG twisted pair) and the assumption that the signal driver

requires 100 ohm differential cable termination. The sloped segment of each curve is

determined using the attenuation values given in Figure 1.

Note the dashed portion of the LVDS curve. In theory, LVDS interfaces can transmit sub-
http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=driver&x=&y=http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=hertz&x=&y=http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=voltage&x=&y=http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=interface&x=&y=http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=function&x=&y=




Mbps signals over hundreds of meters of CAT5e, however, LVDS receivers can only handle

+/- 1V of ground noise and as such are not suitable for a long-haul dc-coupled interfaces.

Other important twisted pair cable characteristics are near-end crosstalk (NEXT) and pair-

to-pair skew. The NEXT is critical when considering bidirectional transmission. For

unshielded twisted pair (UTP) cables, the NEXT increases quickly with frequency and leaves

little room for cable attenuation given the allotted signal-to-noise ration, therefore the

maximum cable length for bi-directional transmission is much shorter than it is for

unidirectional transmission. The cables with individually shielded or foiled twisted pairs

(FTP) have much better NEXT performance (Figure 3) and allow bidirectional transmissionat frequencies beyond 1 GHz.

Figure 3. Cables with Individually Shielded Twisted Pairs Have Minimal NEXT beyond 1

GHz.

The pair-to-pair skew is a parameter that should not be overlooked when timing relationship

between signals being transmitted on the same twisted cable is critical. Novice CAT5e users

find themselves surprised when they discover that the pair-to-pair skew can be as high as

45 ns (typically 25 ns) per 100 meters (TIA/EIA-568-B.2). It is a high number of

nanoseconds when dealing with signals that run at hundreds of megabits per second.

Sort Out Available Signal Conditioners

We just reviewed the three major twisted pair cable parameters that paint the cable's

performance boundaries: cable attenuation, near end crosstalk and pair-to-pair skew. While

cable manufacturers provide relatively economical solutions to the NEXT and skew

problems (cables with individually shielded twisted pairs for the first and cables with tightly

controlled pair-to-pair skew for the latter), the semiconductor manufacturers focus on

"correcting" cable attenuation and provide solutions known as signal conditioners. Some of




them come in form of either signal buffers with built in pre-emphasis or de-emphasis

circuits; others are available as fixed, variable or adaptive equalizers. All these solutions

"compensate" for certain amount of cable attenuation within a certain frequency band. How

these solutions accomplish the "compensation" is what differentiates them. Understanding

the basic principles of their operations is critical for making the optimal use of them, so let's

review each solution and see when it makes sense to employ them with twisted pair cables.

Pre-emphasis is a feature of an IC that boosts the magnitudes of high frequency

components of a signal with respect to the magnitudes of low frequency components of the

signal. On the other side, a de-emphasis circuit decreases the magnitudes of low frequencycomponents of a signal with respect to those of high frequency components. The idea

behind is that the frequency response of either of these circuits when combined with the

response of a transmission medium will yield a response which is close to a flat one. A

network with a flat frequency response does not cause inter symbol interference (ISI) which

results in so called ISI jitter, a type of signal jitter caused by cables, printed circuit board

(PCB) traces or any other passive network with a similar frequency response.

For system designers, an important parameter of either pre-emphasis or de-emphasis

circuits is the boost it provides. The higher the boost, the more attenuation the circuit can

compensate for and ultimately transmit over a longer cable. As an example, Figure 4

illustrates an output signal from a driver with four pre-emphasis levels and the signal's timedomain characteristics. The signal is a pattern starting with one bit high (H), followed by one

to N bits H, followed by one bit low (L), followed by one to N bits L, followed by a H and a L.

Note the signal's nominal amplitude of VOD_L and three distinct boost amplitudes

(VOD_H1-3). The boost is simply the ratio of the VOD_H and VOD_L expressed in dB. It is

also important to note that the duration of the boost pulse (tVOD_H) should optimally be

75% to 100% of the unit interval (tUI) with an approximate useful range from 50% to 125%

of the tUI.

Figure 4. Pre-emphasis or De-emphasis Circuit is the Most Useful When the Boost Pulse

Duration (tVOD_H) is between 75% and 100% of the UI.

To illustrate how you can utilize interface devices with integrated pre-emphasis or de-

emphasis circuits to push the performance boundaries of twisted pair cables out, let's

consider a need for transmitting three 1 Gbps LVDS data signals and one 100 MHz LVDS

clock signal over a single CAT5e cable and assume that we have at our disposal a four

channel LVDS buffer with integrated pre-emphasis drivers.

Let's also assume that each pre-emphasis driver provides a 6 dB boost with the boost pulse

duration of 750 picoseconds. Using the 6 dB criteria for LVDS and the CAT5e attenuation

curve, it is easy to determine that the maximum cable length is approximately 7m for a 1

Gbps signal transmission. With the aid of the 6 dB pre-emphasis boost, the maximum

transmission distance may be doubled. In addition, the 750 ps boost pulse duration reveals




that this particular pre-emphasis driver is only useful within the 670 (0.5/ 750 ps) to 1670

(1.25 / 750 ps) Mbps range, therefore, the 100 MHz clock signal does not benefit much from

the pre-emphasis driver.

However, the CAT5e cable attenuation at 100 MHz is only about 2.2 dB per 10m, therefore

the clock signal can reach enough distance before it runs out of amplitude. Also, recall that

the CAT5e pair-to-pair skew is typically 25 ns / 100m, therefore 7m cable has up to 1.75 ns

of pair-to-pair skew which may still be acceptable portion of the 10 ns clock period.

Doubling the cable length doubles the skew which now becomes a large portion of the clock

period.

While the pre-emphasis and de-emphasis circuit design constraints limit the maximum boost

to about 16 dB, commercially available cable equalizers can provide equalization boost as

high as 40 dB. The cable equalizers are typically devices with integrated peaking filters

whose frequency response curve from the center frequency, fC, to some lower frequency

on the left from the fC is closely matched to the inverse of cable's frequency response or

attenuation curve.

As an example, Figure 5 shows a frequency response of a peaking filter with a center

frequency at 1 GHz and an inverse of a loss characteristic of a 10m CAT5e cable. Both

responses are closely matched up to about 600 MHz. At this frequency, the equalizationgain is approximately 6 dB. An equalizer with 6 dB of gain doubles the transmission distance

of the driver without any signal conditioning.

Figure 5. Response of a Peaking Equalizer with the Center Frequency at 1 GHz Closely

Matches the Inverse of a 10 m CAT5e Response I.

Equalizers can be fixed, variable or adaptive. Fixed equalizers have fixed frequency

response and may be employed in applications where transmission media have known fixed

length. Variable equalizers may have several equalization boost settings, providing more

flexibility with the interconnect length. The most flexible are adaptive equalizers that

automatically determine the transmission media loss and apply optimal equalization boost.

When using adaptive equalizers, it is important to realize that most of them are designed

around specific cables, bit rates and assume specific signal characteristics at the transmit

side of the cable (e.g. amplitude, rise time, etc.). Knowing this information, you can quickly

determine whether the equalizer is a fit for your transmission medium.




EMAIL THIS PRINT COMMENT

When using equalizers, especially the ones with high gain capabilities, one needs to be

very careful with the system noise. High gain, wide bandwidth equalizers can compensate

for interconnect losses that are as high as 40 dB. However, a signal that is attenuated by

40 dB is extremely susceptible to noise. Maintaining 10-15 dB of signal-to-noise ratio means

that you need to keep noise at 50-55 dB levels or better around the equalizer and the

transmission lines connecting to the equalizer inputs. This requirement may easily be

violated if not careful enough.

To illustrate the gains one can realize with high gain adaptive equalizers, let's now consider

a need for transmitting a single 500 Mbps LVDS data signals over a single twisted pair of aCAT5e cable and assume that we have at our disposal an equalizer that provides 30 dB at

1/tUI hertz. Using the CAT5e attenuation curve, it is easy to determine that the cable length

that attenuates a sinusoidal signal by 30 dB at 1/tUI hertz is about 55m. This means that an

equalizer with 30 dB boost at 500 MHz can enable about 55m longer transmission distance

over CAT5e than what a circuit without the equalizer can. A real world example that shows

how to realize long transmission distances over CAT5e cable is described in National

Semiconductor application note AN-1826 (http://www.national.com/an/AN/AN-1826.pdf).

Also, recall that the CAT5e pair-to-pair skew is typically 25 ns / 100m, therefore

transmission of two or more synchronous500 Mbps signals would require either a cable

with much tighter pair-to-pair skew or some type of skew compensation mechanism.

Conclusion

There are three major twister pair cable parameters that paint the cable's performance

boundaries: cable attenuation, near-end crosstalk and pair-to-pair skew. If the near-end

crosstalk and pair-to-pair skew are the limiting factors for achieving necessary transmission

distances, you need to look for enhanced cabling solutions. If it's cable attenuation, there

are multiple commercially available semiconductor solutions available. Knowing the

principles of operation of the signal conditioning solutions and how well they match twisted

pair cable attenuation characteristics, the transmission distance up to 100 meters may be

achieved.

About the Author

Davor Glisic is Senior Applications Engineer at National Semiconductor Corp.

Copyright 2014 UBM Electronics, A UBM company, All rights reserved. Privacy Policy | Terms of Service
http://legal.us.ubm.com/terms-of-service/http://legal.us.ubm.com/privacy-notice-highlights/http://legal.us.ubm.com/copyright-notice/http://www.industrialcontroldesignline.com/encyclopedia/defineterm.jhtml?term=synchronous&x=&y=http://www.eetimes.com/document.asp?doc_id=1274094&print=yeshttp://www.eetimes.com/email.asp

ee times - getting the most out of your twisted pair cable

Documents