Twisted-pair distributed

transmission-line parameters

For more accurate modeling of twisted-pair transmission-line distributed parameters, you must take into account several important factors. Material and structural make-ups, twist pitch, skin effect, and proximity effect all playa role. By T Kien Abstract Both unshielded-twisted-pair (STP) your of applications. calculation missing, Papers transmission Ethernet Despite office's (UTP) and shieldedare widely usage of to Truong, The Boeing Co importance. called Using MathCAD, behaviors we will of the explore many line. SPICE terms With and of the the so-

frequency-dependent "constants"

of a lossy transmission most accurate of the You model fidelity. the standard

Concerning models that more don't

simulation, allow for can high using

twisted-pair used from avionics both accurate types

lines universal lines,

for frequency-dependent modeling. the attenuation

to safety-critical references

are necessary accurate you with

calculation

distributed

transmission especially on or coax

parameters, dispersion frequency line than for topic

of the twisted-pair transmission cable,

parameters lines often can pair. and

in the transmission

line as a function can compute model line will Circuit

are mostly applications. address readily For pair cannot (hence, In this find

for high-frequency and anyone on twisted the

response

distributed model.

rather be the

microstrip find the

the lumped the twisted-pair

ladder

modeling

the equations practical most the parallel when we

for these types, guidance wire model in bus

but you twisted

transmission article.

can rarely using

of an upcoming

example, explain

textbooks change

configuration

impedance transceiver).

Transmission-line The function length) shunt These and (wave characteristic line the of

parameters impedance, is a complex distributed R, C, and of ZO, quantity .It (per of a is a unit L, G. line loss,

reflection) article, more .These

the wires the

are separated parameters account that not transmission series

for magnetic reflect geometry only

coupling accurately

(current-mode analyze the physical take into

twisted-pair make-up, of twist, wires, the

parameters series shunt the loss). after

parameters dimension,

resistance

inductance conductance

the measured effect

structural

capacitance parameters material propagation,

and material the proximity dielectric

constants

but also the pitch and the effects

are modeled energy

the geometry conductor

of current-carrying

make-up

transmission

dispersion,

of frequency

storage,

on skin depth resistance. Introduction Most s~ch dlnates after the transmission-line as parallel or Most parallel with ~Ire textbooks usIng cable wire a and can provide the coorpolar pair air These at high the signal the

dielectric

loss, and radiation jwL)/ resistance common for and is line in example, true (a+

Zo = ~(R+ With ohms, purely This both a

jwC) and

= Ro +jXo

c,

reactance

,:~{'[~fi~ specified;t{r~,.;b for of a given is

derivation

of bus i~peda~ce coaxIal

for simple usIng replace dielectric.

~eometry, the

misconception is that independent only which

the .CartesIan the twisted

ZO=12O.0.,

the impedance frequency. a or for R=G=O

resistive situation

coordinates. environment models frequency integrity

analyses

model

lossless for a R/L=G/C

transmission distortionless

uniform

transmission

line in which

are unsuitable particularly or for applications in which at every point along

when Zo becomes

r-;-;-;;; .. '\I L / C , WhICh IS real.

the bus is of high

In general, we have a lossy twisted-pair transmission line for which the parameters R, L, C, and G are nonzero and are themselves functions of frequency. What we need is the characterization of a bus cable that shows the variation of its magnitude and phase with frequency. It is only if the impedance remains fairly constant over our frequency range of interest that we can treat the characteristic impedance as a constant, We then can specify the load impedance to match the transmission line for that particular range of frequency.

In general, the permittivity of a dielectric is a complex number and changes with frequency: E=ErEO=E'+jE", However, between 1 and 100 MHz, the permittivity of PTFE (polytetrafluorethylen, or Teflon) is almost constant at 2.1, whereas its dissipation factor is 0.0002 (Reference 1). The same is true for the permeability of the conductor, which is a complex number and changes with frequency: 1.1=l.1rI.10=1.1'+jl.1", However, I.1r can change drastically at the low (kilohertz) level of frequency. At the megahertz level, the relative permeability of both copper (0,999991) and silver (0.99998) is practically unity, and thus we will use 1.10in the equations (Reference 1).

UTP impedance calculation Below are the constants and specification for a given 1200 UTP used in this analysis (all units in mks). Wire length: Wire Spocing:
C on d to uc or ' t lome er:

I := 50 D := 2.11' 10-3
d := 112 .' 10 -3

If you use the virgin dielectric constant, er=2,1, for PTFE, you would miscalculate the impedance
using the simple parallel

geometry that most textbooks lIst.

wire-over-ground-plane .

Virgin PTFE dielectric constont:

e r := 2.1 := 1.56 p e 0 := 8.854 ,10-12 11:= 4 ' It , 10-7 (JI := 5.8 .107 (J2 := 6.17 ,107 3 4 7 f := 10 , 10 ..10 fo := 6.106 p f := 2.10 T := 20 , .. The calculatIon for the helIcal pItch angle of twist (Referenc~ 2) plus the b~s length (before twisting) and velocIty of propagatIon are: In this example, the conductor for the twisted pair consists of 19 silver-plated strands of copper and multiple layers of dielectric with different material

Measured effective UTPpermitivityet Air permitivity:

.. D. Iect rlc le b I' permlo 'IIty:

constants as the fi gure below illustrates:

Copperconductivity: Silver (plating) conductivity: Frequency: Nominalfrequency: , .-4 Dielectricloss tangent: Numberof twists (/m):

All equations, calculations, and plotting in this. analysis use MathCAD software which is a powerful and easy-to-use math w~rksheet. After establishing the equations, you can substitute the transmission-Iine values with your own twistedpair parameters and watch all the calculated results and plots change instantly.

9 := atan{T'1t'O)

9=

7,55odeg

10:=T,I,1t'O,~ jl-r~
v
,-

10=

50,44

P'-F:;:

= 0,8 As the bus-core figure illustrates, the diameter of each strand is about one-fifth of the overall conductor diameter. The cross-sectional area of each strand is therefore 1125the overall conductor area. For this lossy twisted-pair line, you calculate the skin depth and dc resistance as follows (taking into account the proximity factor and the conductor area ratio as a good approximation):

Using the twist angle as a factor to compensate for the equivalent dielectric constant and with q as the twist correction factor, the expression for the equivalent dielectric constant, eeq=I +q( er-1), is based on the facts that eeq=l for all air (er=1) and eeq=er for all dielectric (q=l). The parallel equivalent of these layers should match the measured overall dielectric constant of 1.56. We can calculate the impedance of the twisted pair (still assuming losslessness)as follows:

~{ f) := ~

-3 q := 0,45 + 1,10

(
' 9'-

180 1t

2 q = 0,51

Rdc:={1,15)'-'

(19)
25 r(d\2

e eq := 1 + q , ( e r -1 )

e eq = 1,56

l1t' \ 2)

(d )2

,a 2 J

C eq :=

(1t'eo'eeq) ~

-11 C eq = 3,47-10

Rdc=0,029 At high frequency, the sk,in ef~ect forces mo~t of the propagation of electrical sIgnal to the sIlver

,- ( 11' 0' e eq e eq ,C eq
~

L eq

= 4 99 -10-7 ,

plating. Youand better corrosion silver for better commonly use resistance. conductivity

Z eq := C eq eq Unlike

The setup for the equivalent dielectric constant as a function of the twist pitch is as follows: Z eq = 119,83 ~ := 0 , 5 " 60 bus has two e r := 2,1 { ~) := 0,45 + 1,10-3,{ ~) 2 q e eq { ~) := 1 + q { ~) , (e r -1

coaxial cable, the twisted-pair

adjacent conductors that give ris~ t~ th~ pro~imity effect in which the current dIstributIon IS not uniform as the next figure illustrates. Lumping all of the wire strands together, the proximity correction factor for two wires that are right next to each other is P=1.15 (Reference 3),

This setup results in the following

plot:

y=a+j13=.J(R+ jwL)o(G+ The MathCAD constant is:

jwC).

20

expression for the propagation IX!) 1

y(f)

:= (1.RdC.d.~

+ j .2.7t.f.Le~

.J( (2-1t.foPf.Ceq + j .2.1t.f.Ce~ ) Depending on whether the cases are lossless, low loss, or distortionless, the attenuation and phase factors can be approximately constant or linear with frequency. The phase velocity , or velocity of propagation (vp=ro/13), is constant only if13 is
approximately following plot a linear function shows the of frequen.cy. attenuatIon ~e In

You can calculate the effect of the twist the impedance as follows: JD acos,\~
Z(" ) :=. 1t

)
~ ( o -.o~ II:\ \ \ e a e eq( " ) )

decibels/meter: 140 a(f) := Re(y(f.

(~
In( 10)

)
Z(~)

12 10 8

0.0155 0.010 a(!) 0.00

60 0 ~

0 0 5.1rf> f 1.107 The impedance of the twisted pair is fairly constant with a fixed twist angle. Nonuniformity in conductor diameters and dielectric thickness, together with variation in the twist angle, can significantly affect the bus impedance. Intuitively, you can see why the impedance increases when the twisted wires are separated for current-mode transceiver coupling: As.the wires are separated to form an eye, the capacItance decreases, and the inductance increases. The following plot shows the impedance variance with O.OOl-in. manufacturing tolerances on the

The phase shift in degrees/meter is: 13(f) := Im(y (f .~ 1t

c~nductor diameter, t:

diameter,

d, and the overall

dielectric

):;f:J;;' " Using

the

theoretical

propagation

,,~~, constant,-,?,C and input

"

calculate the characteristic impedance impedance as functions of frequency:

130
Y2(f) := (R(f) +j '2'7[-f'L)-(G(f)'+j

""" 'z
," '2-7[ ,pC)"
" ",

Zd1(t)

125

e-E7-e Z d2( t )
+-t-+

120

( f ) ,,-

R(f) G( f)

+J

'2-7['f'L, , + J -2 '7[ 'f'C

'"",CC""'""" """""'C',c' c"'C',"C'C',""",c

,, ',:",,;",,:'

',""

0,0021 115

0,00215
t

0,0022 a 2(f) := Re y2(f

() 20 :
( ) -'c'
In( 10)

'

," "c ""c'

W1th thlS plot m mmd, you can see that the bus terminator need not exactly match the specified bus impedance because the impedance likely varies by a small percent- Because the resistive termination swamps any small presence of reactive load, it has to match only to a nominal busimpedance value in the middle of the above impedance band. Even at 5% mismatch, the reflection coefficient would be (126-120)/ (126+ 120)=2.4%, which results in reflected power of only 0.12%.

(ZL-COS~Y2(f)-I)
Zinif) := Zo(f)-

+ Zo(f)'Sin~y2(f).I)

. Zo(f)-cos~y2(f)-I) + ZL-sin~y2(f).I)

Although MathCAD can calculate all the complex math functions, it is useful to find an easy way to calculate some real numbers for simple comparison- You can calculate the bus impedance, Z, and propagation constant, y, for a particular frequency from the distributed RLCG parameters as follows: r = .J(R + jOJL)(G + jaJC)

Armed with this information, you can first analyze the input impedance as a function of frequency at a bus length of 50m and a 5% mismatched load at 1260.

-OJL)( , "..1'"'\ (1+

JLV'-'J

jOJL

)( 1 ~

+ jOJC )

I := 50 Z L := 126 f := 1 '106 , 1.1 '106 " 10 .106 R(f) G(f) := 0,06 + 4,69 -10-4,[r := 6,17 .10-12,f

For high-quality, low-Ioss cable at high frequencies (more than a few megahertz), in which oJL> > R and aJC> > G, you can use the following simplification for the characteristic impedance: Z= ~ V(G+ jaJC) =f C

L := 4,99 '10-7 C := 3-47 .10 -II

Also, by using the approximation (l+a)(l +b)=(l +a+b) and Jl+-;; = 1 + c /2

for a,b,c I, you can manually evaluate the propagation constant as follows:

STP bus As for STP, the shield now contains the electric field, and this distribution represents additional shunt capacitance from each twisted wire to the shield. For the differential mode of transmission, the equivalent effect is increased distributed capacitance, C; decreased impedance; and longer propagation delay. The extra dielectric insulation between the shield and the already-insulated UTP also increases the distributed conductance, G, which represents dielectric loss. Note that the shield has no effect on the magnetic field. With differential mode of propagation, the shield carries no net current (except for commonmode noise current that results from the transmission-line imbalance). The parameters for STP are therefore similar to those of UTP, except for the increased capacitance and conductance. The dc resistance, Rdc, and skindepth resistance, Rac, also change due to the proximity effect (change in surface current distribution) from the shield.

References: I. "Reference Data for Radio Howard Sams & Co, 1956. Engineers,"

2. Lefferson, Peter, "Twisted Transmission Line," 1971.

Magnet

Wire

3. Magnuson, Phillip, "Transmission Line & Wave Propagation," 1981. 4. Brown et al, "Lines, Waves, and Antennas," 1973.

Author's Biography T Kien Truong has been an engineer with the Boeing Co (Seattle) since 1985. He is a graduate of Iowa State Univ- -: ersity (Ames, IA). His professional experience includes working with antenna and RF , fiber-optic sensors, and faulttolerant computer design. In his free time, Kien enjoys the freedom of telemark skiing and sailing. You can reach him at Boeing Commercial Airplane Group, PO Box 3707, M/S 19-HM, Seattle, W A 98124.

Summary Using MathCAD, this article has constructed the frequency-dependent equations of the many socalled constants of a lossy transmission line. You need to be careful to find out all of the assumptions; be careful not to take a given component constant at its surface value. Manufacturers' specifications usually associate with a specific test condition. It is common to find published data measured only at one frequency (I MHz typical). This article plots the changes in the impedance, attenuation, and phase as functions of frequency, twist pitch, load mismatch, material make-up, and dimensional tolerance of the twisted pair. With accurate calculations for the distributed parameters, you can build an accurate model for the twisted pair that reflects real transmission-line behavior for a range of frequencies.