Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [10GBASE-T] channel model


Perhaps, we can list down the main system modeling requirements for
feasibility study, discuss and agree on these. Then one could look
into the respective characterization requirements on the basis of what
is realistically possible to achieve...

Modeling requirements:

1. Baud rates up to 1.25 GHz max. - various rate choices may need to
be studied.

2. Ease of impulse response generation for a given baud rate and its multiples
for the Channel, Echo, NEXT and FEXT using preferably a unique data set.

3. Model Channel behavior up to 1 GHz insuring a causal system.

4. Proof of feasibility at environmental conditions: 50 C (?)

5. Relaxed impairments accuracy at very high and very low frequencies
(TBD - say above 500 MHz and below 1 MHz), as long as behavior
remains "physical" - motivation:

a) signal psd and channel magnitude would be small at high freq.

b) system will need to have some smoothing and anti-out-of-band
noise LPFs, helping condition in a)

c) there will be in the end-to-end system a high-pass of at least a 2nd
order, hence making the modeling accuracy at low frequencies and DC
practically irrelevant

d) there should be allowed certain characteristic variability at the band
edges to account for practical realizations

e) noise floor

The channel could be reliably modeled up to 1 GHz using continuous
time analytical approximation - even if it is based on the frequency-limited
(~500 MHz) data  set. One notable benefit of the analytical channel model
is that it could be readily sampled at a  required baud rate to produce a
causal impulse response.

The Echo, NEXT and FEXT models could be derived based on the frequency
characterization (~500 MHz) performed with sufficient resolution, such that a
simple interpolation would allow for impulse response generation at different
sampling rates. (Similar approach was taken in 1000BASE-T.)


William Jones wrote:

The issues that I see in determining the frequency range of the channel characterization are:

1) The ability to make accurate high frequency measurements and to validate the results among multiple independent cable vendors.  This suggests a lower frequency range. 

2) The frequency range must be the same for all types of measurements (line as well as crosstalk measurements).  So, although accurate higher frequency channel measurements may be possible, our experience has been that FEXT is the most challenging.  This can limit the frequency range more.

3) Flexibility to perform system tradeoffs suggests a higher frequency range.

Based on these considerations, I propose the characterization of all measurements (line and crosstalk) be 
from DC to 500 MHz.


Chair Modeling Ad Hoc

William W. Jones, Ph.D.
Director of Systems Engineering
SolarFlare Communications, Inc.
949-581-6830, ext. 2550
mobile: 619-405-2445
fax: 949-581-4695

-----Original Message-----
From: []
Sent: Wednesday, January 29, 2003 9:32 AM
To: Larry Cohen; William Jones;;
Subject: Re: [10GBASE-T] channel model


Thanks for getting this started.

Your contribution addresses (at least) three significant elements of the 
channel model development: 

1. frequency range of channel characterization
2. the use of temperature dependent transmission models
3.proposed model limits (with supporting measurement data)

We should try to reach consensus on the elements.

1. The modeling ad hoc should reach a consensus on the frequency range of 
channel characterization.
Note:We should limit the channel characterization to account for relevant 
signal spectrum and 
minimize excess bandwidth characterization. 

2. The 10GBASE-T study group should reach consensus on temperature 
dependencies (in the objectives?).

3. I'm in the process of developing a straw proposal for the cabling ad hoc 
to address 
the generation of measurement  data.  Note: We need the resolution of the 
frequency range of channel 
charcterization (1).


Chris DiMinico
Chair Cabling Ad Hoc
MC Communications

In a message dated 1/29/03 1:33:04 AM Eastern Standard Time, writes:

<< Attached is a proposed 100 meter Cat 5e channel model for 10GBaseT study 
(CHANMOD.TXT).  This model is defined by complex-valued (X + jY format) 
insertion gain in the at 500 kHz intervals from DC to 1 GHz.   
 The proposed model is derived from measurement (at 20 C) of a 100 meter 
channel (with Cat 5e patch cords and 4 Cat 5e RJ45 connector interfaces).  
The baseline 100 meter channel model was temperature corrected to 50 C. (122 
F.) using the procedure defined in ASTM D4566 Section 26.4.  The above 
mentioned temperature correction method modifies the magnitude but not the 
phase of the insertion loss.  Experimental measurements confirm that 
temperature effects on the phase are minimal so the approximation is 
sufficiently accurate.  
 The most recent proposed model is an extrapolation of the ISO 11801 Class D 
channel loss limit.  The ISO 11801 Class D limit is designed to accommodate 
expected worst-case measurements below 100 MHz and consequently includes some 
additional margin to meet this requirement.  Since this limit is only 
specified below 100 MHz, extrapolation of the limit through smooth curve 
fitting becomes less accurate as it is extended farther beyond its defined 
boundaries.  The new proposed model is based upon measurements beyond 500 MHz 
and thus validated to at least 500 MHz (and somewhat beyond 500 MHz).  As 
shown in the attached graphs (CHANMOD.DOC), the insertion loss for the 
proposed model differs from the ISO 11801 Class D 2002 channel limit by less 
than 0.2 dB from DC to 50 MHz and less than 0.5 dB from 50 to 100 MHz.  While 
it does not represent an absolute worst-case channel, it represents a 
reasonable model for a feasibility study simulating a maximum length channel 
under worst-case temperature conditions.