Signaling ad-hoc conference call (minutes), Monday, August 23, 2004

Attendance:

Joe Abler [IBM]
Michael Altmann [Intel]
Stephen Anderson [Xilinx]
Michael Beck [Alcatel]
Harmeet Bhugra [IDT]
Joe Caroselli [LSI Logic]
John D'Ambrosia [Tyco]
Justin Gaither [Xilinx]
Adam Healey [Agere Systems]
Bill Hoppin [Synopsys]
Mike Lerer [Rapid Prototypes]
Cathy Liu [LSI Logic]
Mary Mandich [Lucent]
Charles Moore [Agilent]
Roland Moubarak [TI]
Tom Palkert [Xilinx]
Petre Popescu [Quake]
Jeffrey Sinsky [Lucent]
Fulvio Spagna [Intel]
Schelto van Doorn [Intel]
Brian Von Herzen [Rapid Prototypes, Inc.]


  a.. Charles Moore volunteers to submit a proposal based on "receiver interference" technique. 
  b.. John D'ambrosia will put together a a presentation of "interesting" channels to be used for simulation 
  c.. Bill Hopin will also put together a presentation on simulation methodology 
  d.. Fulvio raised the question of how to model crosstalk and whether simulation results should be based on ideal timing or include the effects of a CDR 
  e.. Mike agreed that results comparison should consider crosstalk and timing effects 
  f.. Charles expanded a little on his "receiver interference" technique to point out that the method would not explicitly consider individual aggressors but rather lump everything into a sinusoidal disturbance 
  g.. Fulvio volunteered to help Charles on the timing modeling 
  h.. As the subject of exchanging proposal etc. through the Reflector, Adam pointed out that attachment should be limited to below 100K. Anything larger should be sent to him directly for posting on the web 
  i.. It looks like Charles, Fulvio John and Mike will work together on a methodology proposal 
  j.. Ali requested Mike to issue a list detailing what is needed for a methodology proposal 
  k.. Mike then raised the issue of the Channel model pointing out that the Channel ad-hoc will issue templates but that it is not clear on how those should be used. For example, should we simulate with a real channel approximating the template? 
  l.. John pointed out that "worst" and "best'  are relative terms as differences can be emphasized because of different signaling/equalization choices. So how many curves do we want to pick? 
  m.. Charles proposed 4 curves: worst case template, a short line and a couple of cases particularly "perverse" because of reflections etc. 
  n.. Jeff agrees on the relativity of the terms "best" and "worst" and proposes to come up with a set of parameters which would allow to cover all bases 
  o.. Mike suggests that if we look at channel topology  maybe we could limit the number of combinations. For example if crosstalk is connector dominated it may not change with the length of the backplane 
  p.. Petre argues that the aggressor/target relative length is a factor that should not be forgotten 
  q.. Mike expresses the concern that if crosstalk is allowed to vary then each insertion loss curve may be accompanied by a family of crosstalk curves running the risk of making the whole problem intractable 
  r.. John reiterated his willingness of putting together a presentation illustrating the various channels scenarios so we can close on what exactly are we going to simulate with. Those interested in particular features (dips, variations, reflections etc) should e-mail him. 
  s.. Steve asked if his synthesized model was going to be used for the simulation work and the overall response was positive. 
  t.. The discussion then moved on whether we should have a worst case model worst than the template. The discussion quickly became a discussion on how can we establish how much margin a certain solution may have. This subject was table till after next meeting 
  u.. After some more discussion on the characterizing channels in the frequency and time domain, a consensus seemed to emerge that we should have both. 
  v.. In closing Adam reminded everyone to send him presentations one day in advance of next meeting.