RE: [SI-LIST] : Twisted Pair Impedance

About this list Date view Thread view Subject view Author view

From: Larry Miller ([email protected])
Date: Tue Nov 21 2000 - 16:11:44 PST

I believe I can speak to that a bit, having been on the IEEE 802.3ab
1000-BASE-X (Gig Copper) committee.

1. The line data rate for Gigabit Copper is the same as for 100BASE-TX.
However, all 4 pairs are used simultaneously to both send and receive. A
coding scheme is used to get a 1 Gb/s bi-directional data system. The energy
spectrum of 1000BASE-T is virtually identical to that of 100BASE-TX. (125
MHz signaling rates.)

2. The delay amongst the various pairs must be very tightly controlled
because the transceivers' receivers have to not only subtract their own
transmitted signals from what they hear, but also the reflections of direct
and reflected signals from the other 3 pairs in the cable. This requires a
very sophisticated DSP echo-canceling and crosstalk-canceling circuit on
each receiver. The amount of cancellation required is so high that very good
time resolution is required; a 128 or higher number of stages of DSP
filtering is used. Because of the preciseness of cancellation required, a
limited range of skew between channels can be tolerated. Handling a lot of
skew with very high resolution would require a huge number of filter stages
in an already exceedingly complex circuit. Therefore the skew between
channels must be held to comparatively small amounts (hence the tight
tolerance). The commercial transceivers on the market actually have
registers that can tell you the relative path delays (skew) between pairs as
measured by the transceivers when they become "trained" to the particular
cable and connectors they are using for that session.

3. Category 6 and 7 cables (at least Cat 7) are shielded because bundling
unshielded wires together as is commonly done introduces enough "alien"
(uncorrelated) crosstalk from other cables to hopelessly confuse the echo

4. The slightly different lays of pairs in the cable are to avoid
periodicity between pairs that would produce resonances or impedance lumps
due to happenstance cable wire alignments. This has been true for at least
Cat 5 as well.

5. You CANNOT have several inches of untwisted cable in a connector area.
The TIA spec is for 1/2" maximum if I remember correctly, and indeed for Gig
Copper (which was designed to operate over single Cat-5 cables) it is the
connector areas, not the cable runs, that are the problem areas for
confusing the echo and crosstalk cancellers.

Hope this helps,

Larry Miller

> -----Original Message-----
> From: Barry Ward [SMTP:[email protected]]
> Sent: Tuesday, November 21, 2000 10:17 AM
> Subject: [SI-LIST] : Twisted Pair Impedance
> I have read about the newer Cat 6 & 7 twisted pairs with their gigahertz
> data rates. From transmission line theory I understand that the distance
> between the conductors and the dielectric constant of the intervening
> material are the primary factors determining the capacitance and thus the
> impedance. But the new Cat 6 & 7 cables have very tight specs for the
> lay, or twisting, of the pairs. I am forced to believe that the lay and
> its consistency is therefore also very important. I assume some of this
> is for crosstalk reasons (the longer the distance between any repetition
> of the lay the less the x-talk), but is some of it also for impedance
> reasons? What is the math behind any of this (assuming a single pair
> surrounded by air for an infinite distance). Approximately how does this
> math change when multiple pairs are cabled together?
> Any answers would be greatly appreciated. I work for a cable manufacturer
> where we build cables with twisted pairs. We work in relatively low
> frequencies (up to 10 MHz), and the lay does not seem to be too critical.
> We untwist the pairs for a couple of inches at the ends of the cable to
> terminate the wires at connectors; this does not seem to make the cable
> unusable.
> Again, thanks for any information.
> Barry Ward
> GeoSpace LP
> [email protected] <mailto:[email protected]>
> 713-939-7093

**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected] In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at

About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:30:13 PDT