Mike A. opens outlining the meeting goals. Continues going over slides 2-5 of his presentation ( http://www.ieee802.org/3/ap/public/signal_adhoc/altmann_s1_0205.pdf ).

Joe A. goes over his presentation (http://www.ieee802.org/3/ap/public/signal_adhoc/abler_s1_0205.pdf ) detailing the impact on simulation results due to the package and IC model. Joe’s conclusions are that a package model should be included but that we should dispense with using an addition IC model.

[Shannon] I agree with Joe’s conclusions that a package mode is needed.

[Matt H] How was the coupling cap handled?

[Joe A] By cascading the s-parameters models of package-cap-backplane

[Matt H] Shannon, was there any progress in generating AC coupling cap models with vias?

[Shannon] No.

[Mike A.] For now, the s-parameters files we have is … all we have so it will be difficult to consider additional refinements.

[Mike A.] Joe, which package model would you recommend?

[Joe A.] I do not feel strongly either way but, if I had to pick one, I would pick the inductor model.

[Mike A.] Why?

[Joe A.] The inductance is going to be a problem at high frequency and the caps used in the network seem reasonable. Not a strong bias one way or the other, though.

[Shannon] It looks good to me.

[Matt H] Note that the slide in your presentation (i.e. Joe’s presentation) labels incorrectly the two models (problem in original slide from Richard Mellitz).

[Joe A.] In which case, I will reverse my opinion and pick the cap-like package model.

[Mike A.] I think I agree with that. 1.1nH seems high. Shannon, what were the conclusions form your measurements?

[Shannon] These models seem to be the worst case examples of two different substrates.

[Fulvio] Are the results in slide #5 mislabeled as well?

[Joe A.] No, these reflect the actual package model filenames.

[Mike A.] Given that, what is the comment about the IC model?

[Joe A.] I do not see the need to add additional capacitance to the package mode to model the IC. As far as the AC coupling cap, I can go either way. In my view, the available model, without any via, is not adding significant impairment. I would go without it but would urge a common position.

[Mike A.] Let’s go back to my presentation, slide #7, where I have listed some straw polls.

Straw Poll #1: For our sim in March, should our sim use a package model?

YES by acclamation (12 participants)

Straw Poll #2: From Joe’s presentation, which package model should we use?

CAP-like: 11 INDUCTOR-like: 0 BOTH: 0 Abstain: 1

Straw Poll #3: Should we use a model for TP4-Tp5 for the purpose of presenting results for the March Plenary? As noted by Joe Abler, this includes cascading with all the crosstalk channels (up to 8).

YES: 4 NO: 7 Abstain: 1

[Mike A.] I would like to interpret this result: anybody can add a TP4-TP5 model but, for the purpose of the March meeting, I must echo Joe’s position that results will have the most impact if they align/are consistent with everyone else’s results.

[Joe A.] If anyone notices a significant difference with or without the TP4-Tp5 model, please pass the info on the reflector so anyone else will be able to look at it.

[Mike A.] This covers the agenda for today. On final point I want to raise, is entirely simulator dependent. If you use ADS, or a Spice like simulator, the cascading of network described by s-parameters is taken care for you. If you are using a Matlab-like platform this is not the case and care should be taken on how this step is executed. In March it would be useful to detail how these elements are cascaded. Is there any other question before we adjourn?

[Shannon] Joe, in your January presentation you showed that the Tp4-Tp5 model did have an effect.

[Joe A.] The results were mixed and the differences were minimal.

Meeting was adjourned.

Attendance:

Joe	Abler
Arne	Alping
Michael	Altmann
Rob	Brink
Matt	Hendrick
Mike	Lerer
Cathy	Liu
Charles	Moore
Petre	Popescu
Shannon	Sawyer
Fulvio	Spagna
Schelto	van Doorn