June 1, 2005 IEEE802.3ap Channel Ad Hoc Meeting

Agenda:
	Mellitz: Informative Channel Spec
	D'Ambrosia: ACR and crosstalk 
	
Reminders:
	Channel Ad Hoc today, and 6/15 10am PST
	Deadline to submit presentation time is June 15 @ midnight EDT
	Deadline for draft comments is June 15 @ midnight
	802.3ap Interim meeting is June 21-23 in Minneapolis, MN

Carry Over Items:	
	none

Mellitz - Informative Spec Value Recommendations
Charles - equation 69-3 on page 7, what is delta2 min, delta1 min
Rich - those are confusing...the measured and spec values need to be clarified.
Steve A - how does slide 5 relate to slide 8?
Rich - slide 8 gives measured values below AF2 and AF1
John - the red line on slide 8 is goergen limit. The purple green and blue lines 
on slide 5 are maximum ripple limits.
Rich - 60% of the channels would dip below the line, but no remain there very long.
John - there are several models that must be met simulataneously
Rich - since we didn't build a model to simulate, we are trying to put limits on 
the complex real channel characteristics.
Petre - What about the second
Charles - Rich, the min and max are delta values, they should be delta plus or delta minus.
Rich - the measurement and the spec value need to be clearly defined.
John - delta1 relates to F1, talking above and below confuses things. The history of the 
the thinking was that the big ripples are between F1 and F2, if you lump those ripples 
in the F2 bucket, the F2 bucket is excessively constrained.
Charles - some folks were assuming deltamin and deltamax were the same for KR, KX, KX4
John - The real limit depends on slide 5 
Annirudda - on page 8 there are four parameters, delta1min/max, delta2min/max correct?
Charles - the deltas in the figure are confusing, the equations on page 7 are correct
John - all the values for the deltas are at the F1
and F2 specified.
Annirudda - do we apply the delta values on table 13 to the f1 and f2 for KR,KX,KX4. 
Based on the informative line. The goal would be to match page 11, 12, fit line.
Brian - The objective in the informative channel model
gives channel builders insight into what should work.
The decision for these lines should be based on what
worked or not with 10G simulations.
Xiao Ming - how much confidence can we get with this
approach?
Brian - It is usefull if it's descriminating real 
good channels from real bad channels
Annirudda - based on the simulations we have seen
all channels except for the T channels are solveable
Brian - on your 11,12,13 plots, you may want to put channels that joe couldn't 
solve in red
Rich - You need >2nd order effects, and it's too complicated
to represent on insertion loss plots
John - We need to replot the limit lines for KX, KX4 and KR
Charles - You need different fit lines for KX, KX4, and KR
John - We were tasked to come up with a single
equation across KX, KX4 and KR.
Rich - i can see how the fit line would be different
Adam - the original LMS fit was 1-6GHz
Charles - one could interpret the motion to create
too steep of a curve for the other ports
Rich - what do you think we should do?
Charles - separate limit lines; reduce the aggressive freq
ranges for KX and KX4, since KR just goes to nyquist;
the best lines in the sand is Rich's presentation
to show how the channels are ranked.
John - the channels may fail for reasons other
than the channel spec defines
Rich - abler did sims w/wo xtalk, 
Charles - we would like to reject channels that fail
simulation. they have to only fail one test. we would
like to pass the ones that don't fail.

John - ACR update
Charles - you are saying there is SDD21 and XT we
shouldn't ignore the channel reflection coefficient.
John - when i parse through the data i can rationalize
the data in my head.
Rich - as far as the return loss. I reported the 
channels.
John - slide 6 shows tyco case 2 outperforming In3
and In5 w/wo crosstalk.
Rich - abler data includes package effects
Charles - suggestion, including the return loss with 
a line, or look at transimpedance characteristic. The 
third approach would be to look at the reflection
as crosstalk is changed.
NEC - why are we looking at insertion loss below nyquist?
Charles - there is impact of ripples below nyquist.
Brian - I think we should focus on KR, and avoid the time 
sink associated with KX and KX4 specs.

Straw Polls

1. Define deltamin/max, IL1, IL2 independantly for KX, KX4, KR
	Y = 2
	N = 15
	A = 4

2. Define limit lines for insertion loss and ripple
	1. single slope line with segments for port type = 12
	2. different slopes for each port type with same attenuation end points. = 0
	Abstain = 5
(there were several comments on the confusion associated
with the choices on this straw poll)

-Shannon