(attaching a text version of the minutes for ease of archiving)
======================================================================
IBIS INTERCONNECT TASK GROUP
http://www.ibis.org/interconnect_wip/
Mailing list:
ibis-interconnect@xxxxxxxxxxxxx<mailto:ibis-interconnect@xxxxxxxxxxxxx>
Archives at //www.freelists.org/archive/ibis-interconn/
======================================================================
Attendees from March 23 Meeting (* means attended at least using audio)
ANSYS Curtis
Clark*
Cadence Design Systems Bradley Brim*
Cisco David
Siadat
Intel Corp. Michael
Mirmak*
Keysight Technologies Radek Biernacki*, Ming
Yan*
Mentor Graphics Arpad Muranyi*
Micron Technology Justin Butterfield*,
Randy Wolff*
SAE ITC Maureen
Lemankiewicz, Logen Johnson*
Signal Integrity Software Walter Katz*, Mike
LaBonte*
Teraspeed Labs Bob Ross*
University of Aveiro in Portugal Wael Dghais
Michael Mirmak convened the meeting. No patents were declared. Michael
called for comments on the minutes of the March 16 meeting. Bob Ross stated
that he had read the minutes but suggested others may not have had time to
review them. Brad Brim moved to defer consideration of the minutes. Bob
seconded the motion. No objections were raised and the motion carried.
Michael thanked Mike LaBonte for helping with the minutes.
No opens were raised.
Arpad Muranyi showed a set of slides, developed after discussions with Brad to
clarify specific issues and questions. Arpad called attention to a sentence in
Draft 30: "The Number of Terminals entry in the Interconnect Model shall be an
integer equal to N+1" A later sentence states that "Terminal N+1 shall be
connected to a Pin Pad, or Buffer terminal which is in turn connected to a Pin
with a signal_name of POWER or GND."
The key question to answer is: how many ports (N) are used for a given number
of pins+pads for a package model?
Arpad focused on an additional quote from Draft 30: "Number of terminals
subparameter... defines the number of Terminals associated with the
Interconnect Model." This is ambiguous. Arpad showed several images,
summarizing what he heard in recent discussions:
1) 3 pins, 3 pads; 3 connectivity options
a. Number of Terminals=7, ports(N)=6
b. Number of Terminals=6, ports=5; Vss pin as reference
c. Number of Terminals=5, ports=4; Vss pin and pad are shorted, and used
as reference
Brad suggested that measurements or computation of S-parameters means Option C
- Data and Vss on the pad side with two ports, etc. Michael asked whether the
Vss pin and pad are still shorted. Brad replied that this depends how you
netlist. You can use 6 terminals and 4 ports. You can netlist in 3 ways; 5
terminals, 8 terminals, 4 port with global ground. Additionally, the wrapper
could have 6 terminals, but the S-element would have 8 terminals with two
terminals repeated twice.
Radek Biernacki suggested that a solution having 6 terminals with 4 ports would
not be variable. Brad agreed, except that one would netlist as 1+ref1, 2+ref1,
3+ref2 and 4+ref2.
Radek replied that it's possible to have 4 ports or 5 ports. Brad asked how
one would compute or netlist it.
Arpad's second example:
2) 2 pins, 2 pads; 2 options
a. Terminals=5, ports=4; where is reference?
b. Terminals=4, ports=3, Data_X pin used as reference
Arpad's second example was classified by Brad as not applicable as there is no
reference. No power nodes or ground nodes are shown for the TX buffer in the
drawing; there is an implicit extra node. Radek stated that it's an incomplete
description. A differential port is two single-ended ports with the same
reference terminal for both ports.
Arpad stated that a shortcut can't cover all these variations.
Walter Katz stated that we are defining an interconnect model; the Draft text
does not state how to connect anything. In short, an S4P has 4 ports with 5
terminals, nothing more. There is always a pin for referencing in reality.
Radek agreed, stating that Walter's approach is consistent and unambiguous: N
ports, N+1 terminals.
Walter suggested that Arpad's second diagram would be described through an S4P
for signal integrity. Bob added that the extra terminal, not shown, is the
global ground. S4Ps are common for differential.
Radek asked Randy Wolff about his usage model. Randy replied that in his
implementation, one must specify a reference for every one of your ports.
Brad asked whether we want to accept external nodes that are not specified.
This relates to "GND" reference clean-up discussion in other Task Groups. If
we choose for this issue to create another unspecified node, we will have
"gotten worse when we are trying to get better". Mike LaBonte asked whether
this is unspecified by the model-maker, as opposed to unspecified in the data.
Brad suggested that Arpad's option 1 choice implies extra, unspecified node.
Brad asked how, given six nodes, even at DC, one can make a unique 6-port
S-parameter. You have to have a seventh node. T
Michael asked whether we could simply assume Arpad's option 1 with universal,
ever-present GND. Bob agreed, there could be an unstated reference plane.
Radek stated that having this node not a violation of current out/in summation
requirements. But data is underdetermined.
Radek agreed with the current text description, but wants removal of the
required connection to a POWER or GND pin. Bob suggested that one could have a
dummy extra GND terminal or pin.
Brad noted that everyone wants a general solution; he added that Walter wants
to avoid having 2 extra files. He suggested that one can emulate netlisting
with N, N+1 or 2N terminals, instead of having a specific set of rules. An
alternate approach would be to effectively have no rules about anything related
to connectivity. Though this would permit model=makers to make mistakes but
would be simple and general to implement.
Brad accepted the AR to write up explanatory text to cover this approach, for
discussion in the next meeting.
Arpad moved to adjourn. Brad seconded the motion. The meeting adjourned.
Attachment:
image002.jpg
Description: image002.jpg
