Professional Documents
Culture Documents
Crosstalk - Correção Por Comprimento PDF
Crosstalk - Correção Por Comprimento PDF
De Montfort University
Leicester, U.K.
+44-116-257-7056 apd@dmu.ac.uk
propagation constant to be determined. For connectors, however, this
is not feasible, as the delay has been specified, unfortunately, with
such a large margin, that the calculated phase reaches values which
are too high compared to the electrical length of the connector. This
issue should be revised seriously by IEC TC48.
Abstract
This is the first of two related papers that focus on the current status
of a comprehensive channel model, as developed by ISO/IEC
JTC1/SC25 WG3 and the subgroup MTG (Modeling Task Group).
Consequently, the papers deal with data grade channels which
contain four twisted pairs (hence 4 Shannon channels) as used for
Ethernet applications according to IEEE 802.3. Such channels may
be configured with different cables and cable types, interfaced with
(mostly) RJ45 mated connectors. Though these connectors are
considered as very short transmission lines [1], they behave in the
cascade as a lumped element transmission-line, i.e. the reflections
upon impedance mismatches at the entrance and exit of the mated
connector are hardly differentiable in the frequency domain.
Therefore, these papers concentrate on the effects of the cables and,
predominantly, the problems encountered upon the concatenation of
cables.
The phase for the reflection coefficient, based on the return loss,
which is based on the round trip, is twice that of the propagation
constant. This relies on the general assumption that the cable is
essentially homogeneous over its length: something that cannot be
achieved even in the best cable designs. Future work needs to
consider impedance or return loss roughness as a function of the
length of the cable, which will then need to be related to the distance
domain instead of the frequency domain.
For crosstalk, i.e. for NEXT and FEXT the phase problem is even
more acute. However in general, there exists a correlation between
the return loss and the crosstalk, as shown in [3]. In another
important paper [4] Friesen compares NEXT between two pairs in
the frequency and distance domain on a long cable and then bisects
the cable and remeasures the NEXT. He finds good agreement, as
we found also in [2], with minor discrepancies attributable to the
necessary deformation of the cable geometry at the ends for
measurement and connectorization.
2. Background
We use the following specification limits according to the IEC
61156 series of documents, here for Cat. 6A cables. This paper only
considers the concatenation of cables. We use a frequency range up
to 1000 MHz in our calculations, which is in excess of Cat. 6A cable
specifications.
1. Introduction
The concatenation of cable component measurements is very
accurate using the complex measured Sparameters and transforming
them into Tparameters [2]. It is also feasible to use Kparameters
neglecting of the errors introduced by the junctions, which
inherently create an impedance irregularity, not being picked up in
this case. One difficulty resides in the mutual conversion of S; T
and K parameters for multiport networks. These problems are
nevertheless solvable.
RL =
= 1.82 f + 0.0091 f +
1 f 1000 (500 )
20 + 5 log( f ) (1) f 10
0.23
532
25
25 7 log( f )
17.3
(1)
10 f 20
20 f 251.8
f 251.8
There are no indications about the phase for any of these values.
Therefore we use the propagation delay to determine the phase. The
propagation delay is specified for all cable categories to the same
requirements, i.e. 570 nsec/100m A rationalization allows the
determination the velocity of propagation and there from the phase:
v=
1011 f
534 f + 36
[m / sec]
(2)
2 108 f
v
[radian / m]
Using the equations Eq (1) and Eq (2), we can easily derive the
propagation constant:
l = ( + j ) l
[ Neper; radian ]
0.23
1.82 f + 0.0091 f +
2 108 f
f
l
=
+ j
(3)
S l = 10
l
20
[ cos ( l ) + j sin ( l ) ]
(4)
3.1.2. NEXT
Initially, length correction according to the IEC length correction
formula is applied:
1 10
l
5
100
1 10 5
[dB @ l]
(5)
This is the length correction for the absolute value of NEXT, not taking
the phase into account. Here the NEXT has to be uncorrelated with
length. The relevant S-parameter can then be obtained from Eq (5):
SNEXTl = 10
NEXTl
20
cos ( 2 l ) + j sin (2 l )
(6)
3.1.3. ACRF
ACR-F is generally a derived value (though it could be also
measured directly using a vector network analyzer without an
incorporated S-parameter test set). There are two approaches to
obtain a length correction following the IEC length correction:
l
ACR Fl = ACR F100 10 log10
100
FEXTl = ACR Fl l
(7)
or :
FEXTl = FEXT100 10 log10
l
l
+ 100
1
100
100
533
SFEXTl = 10
FEXTl
20
cos ( l ) + j sin ( l )
(8)
Here is assumed, as in section 3.1.1 above, that the pairs have the
same propagation phase. A detailed analysis of the potential impact
of the inhomogeneous distribution of coupling coefficients is still
outstanding.
1 j
Zo = Z asymtote 1 + 0.055
f
[ ]
(9)
Figure 2: Homogeneous cable return loss for an
asymptotic impedance of 100
RL
20
[ cos ( 2 ) + j sin ( 2 ) ]
cos h ( l
K l = sin h l
Zo
( )
2.) RL = 20 log10
SK l =
Zo2
(10)
( )
( )
Zo sin h l
K11 K12
=
cos h l
K 21 K11
Zo ZL2
2
2
Zo + 2 ZL Zo cot h( l ) + ZL
(11)
Zo2 ZL2
+ 2 ZL Zo cot h( l ) + ZL2
Towards this we use compute first the return loss based on the
second equation of Eq (11).
534
Zo ZL
RL mean = 20 log10
Zo + ZL
Zo ZL
RL Envelope = 20 log10
1 e 2 l
Zo + ZL
(12)
The sign in the last bracket indicates the lower and upper envelope,
respectively. The envelopes are indicated in the Fig. 2 to Fig. 4.
Nl m
for l 50 m
(13)
for l > 50 m
3
2
535
536
Figure 14: Return loss limit for 100 m cable and return
loss obtained by concatenating two 50 m cables with the
same return loss as the 100 m cable
74.461
NEXTL13
FEXTCa
60
FEXTL14
ACRFL14
ACRF
40
20
5.382
0
1
fmin
10
100
fk
3
1 . 10
fmax
Frequency [ MHz ]
NEXTCa
80
537
SRL l = 10
N RL100
l
20
(cos(2 l ) + j sin(2 l ))
(14)
30
Return Loss [ dB ]
25
RLL1
RLcab
20
k
k
15
10
9.215
10
1.0
fk
Frequency [ MHz ]
100
3
1 . 10
fmax
Figure 17: Return loss limit for 100 m cable and return
loss obtained by concatenating four 25 m cables with the
same return loss as the 100 m cable
538
6. Conclusion
This paper has demonstrated the benefits of considering length
correction in the concatenation of cables, In terms of future
work recommended from this research, our preliminary results
indicate clearly that we have to revise the length correction at
least for the crosstalk parameters. As these are loosely related
to the reflection coefficient or return loss, we have to consider
both types of parameters jointly. As mentioned above, we have
to come up with a length correction for these parameters based
on a length domain correction.
7. References
[1] J.-H. Walling and D. C. Hess, A contribution to the
determination of the electrical length of mated data grade
connectivity based on channel measurements and simulations,
Proc. IWCS 55(2006), p. 147 - 152.
[2] J. Castro and J. H. Walling, Application of generalized
scattering matrices for the modeling of twistedpair cables,
Proc. IWCS 57(2008), same volume.
[3] H. W. Friesen, Roughness noise resulting from given SRL
levels, Handout for IEC 46CWG7 Session, AT&T Bell
Labs, Norcross, GA, May 1920, 1994.
[4] H.W. Friesen, Experimental verification of near-end crosstalk
equation for balanced telephone cable pairs, IEEE National
Telecommunication Conference 1973, Atlanta, GA, CVol. I, p.
8C-18C-11.
5. Discussion of results
The use of the return loss limit for 100 m cables, applied to 50
and 25 m long cables, concatenated such as to obtain again 100
m long cables yields results which demonstrate a high level of
error. Therefore, this approach has been omitted.
7. Authors
539
Engineering
Division
at
De
Montfort University Leicester, UK
and has particular research interests
in CEM Validation, communications
cabling
and
technology
management. He has published over
100 papers in journals and
international symposia.
Dr Duffy is a Fellow of the
Institution of Engineering and
Technology (IET) and a member of
the Chartered Management Institute
(CMI). He is active in the IEEE
standards activity on the validation of CEM. He is a member of
the International Compumag society and the Applied
Computational Electromagnetics Society.
540