Scott, You should limit yourself to extending my proposed format, you can do it based on BIRD 125, or BIRD 145, or come up with a totally new format. Walter From: Walter Katz [mailto:wkatz@xxxxxxxxxx] Sent: Wednesday, September 18, 2013 5:38 PM To: 'Scott McMorrow' Cc: 'IBIS-ATM'; 'Randy Wolff' Subject: RE: [ibis-macro] The EMD like syntax of package and on-die interconnect models Scott, The example are the kinds of models that we receive from IC Vendors. Please use the syntax I have proposed, and demonstrate how you would describe the terminals of a model that you think IC Vendors should be delivering. When I say using my syntax, I mean that there are branches in package models of the form (<terminal number> (Pin (<keyword> <Value>) (<keyword> <value>) ..) (<terminal number> (Pad (<keyword> <Value>) (<keyword> <value>) ..) And the branches in on-die models of the form: (<terminal number> (Pad (<keyword> <Value>) (<keyword> <value>) ..) (<terminal number> (Buffer (<keyword> <Value>) (<keyword> <value>) ..) The <terminal number> is also the port or node of the sNp or the .subckt. Please do not get hung up by the parameter tree syntax, we could implement this same functionality with an IBIS like keyword syntax, e.g. (sdram (DQ0 | s2p for this specific pin (C2) (Tstonefile_File (Corner DQ0_typ.s2p DQ0_min.s2p DQ0_max.s2p)) (Terminals (1 (Pin (Pin_Name C2))) (2 (Pad (Pad_Name C2))) ) ) Could have been implemented as [IBIS-ISS Package Models] sdram [Package Model] DQ0 Tstonefile DQ0_typ.s2p DQ0_min.s2p DQ0_max.s2p [Begin Terminals] Terminal 1 Terminal_Type Pin Pin_Name C2 Terminal 2 Terminal_Type Pad Pin_Name C2 [End Package Model] So propose a specific syntax for your approach, and document exactly how an IC Vendor should generate such a model, and how an EDA tool should use such a model. Walter From: Scott McMorrow [mailto:scott@xxxxxxxxxxxxx] Sent: Wednesday, September 18, 2013 5:24 PM To: Walter Katz Cc: IBIS-ATM; Randy Wolff Subject: Re: [ibis-macro] The EMD like syntax of package and on-die interconnect models There is a not to subtle difference between supplying models that accurately represent the part, and supplying models that represent a conservative model. Supplying a model with a part is an inferred warrantee that could affect the production yield of a part, so IC Vendors (especially high volume ones) will be conservative on their models, often specifying the models using typ/min/max models from a standard as opposed to actual accurate models. This is a false choice. There is no inferred or implied warranty. A disclaimer usually is attached to a model. "Do you object that our package solution support both detailed aggressor information as you support and bounded aggressor information?" Your approach is a subset of mine. With my approach you can created a bounded model. It's very simple. A list of Drivers. A list of Receivers. A List of Receivers available for crosstalk analysis (which is a subset of the list of receivers), and a method to indicate those drivers and receivers that are in proximity to the analyzed receiver. The list of proximate I/O cells is optional for models encapsulated within a package, since by definition they are proximate. On Wed, Sep 18, 2013 at 5:15 PM, Scott McMorrow <scott@xxxxxxxxxxxxx> wrote: Walter A generic model can specify Drivers, receivers, and proximity to the target receiver. If you want I can go through each one of your examples and show that it can be done in this way. Or ... do you already have models that you are forcing the specification around? On Wed, Sep 18, 2013 at 5:05 PM, Walter Katz <wkatz@xxxxxxxxxx> wrote: Scott, Scott, with all due respect, your approach is most accurate, and I expect you will try to convince IC Vendors to supply the models in your format; our packaging solution must support such models. You should not be surprised that many IC Vendors (and particularly those that have very few letters in their names) will not supply such model in your format, but that they supply models that are either generic to a bus with generic worst case aggressors, or are specific to a pin for the victim through path with generic worst case aggressors; our package solution must support this as well. There is a not to subtle difference between supplying models that accurately represent the part, and supplying models that represent a conservative model. Supplying a model with a part is an inferred warrantee that could affect the production yield of a part, so IC Vendors (especially high volume ones) will be conservative on their models, often specifying the models using typ/min/max models from a standard as opposed to actual accurate models. Do you object that our package solution support both detailed aggressor information as you support and bounded aggressor information? Walter From: Scott McMorrow [mailto:scott@xxxxxxxxxxxxx] Sent: Wednesday, September 18, 2013 4:22 PM To: Walter Katz Cc: IBIS-ATM; Randy Wolff Subject: Re: [ibis-macro] The EMD like syntax of package and on-die interconnect models Walter, with all due respect, all one needs to know are three pieces of information: 1) where are all receivers and drivers 2) what receiver are we currently measuring 3) what drivers in near end proximity to the receiver. with those three pieces of information, all crosstalk methodologies can be implemented. One does not need special cases to describe each potential methodology. Scott On Wed, Sep 18, 2013 at 4:02 PM, Walter Katz <wkatz@xxxxxxxxxx> wrote: Scott, There is more than one way to skin a cat. There are crosstalk methodologies that do not care about victim and aggressors. There are crosstalk methodologies that do care which terminals are victims and aggressors. An IC vendor can and will choose the way they are going to construct and distribute these models. We need to be able to support both methodologies. Walter From: Scott McMorrow [mailto:scott@xxxxxxxxxxxxx] Sent: Wednesday, September 18, 2013 2:38 PM To: Walter Katz Cc: IBIS-ATM; Randy Wolff Subject: Re: [ibis-macro] The EMD like syntax of package and on-die interconnect models i disagree with victim aggressor constructs. All that needs to be known is the location of the drivers and receivers, and grouping of such for a general solution. On Wed, Sep 18, 2013 at 2:35 PM, Walter Katz <wkatz@xxxxxxxxxx> wrote: All, I am enclosing a presentation which demonstrates the simple language used to describe both the package and on-die interconnect models using the EMD like methodology applied to all of the types of models that IC Vendors are requesting to deliver. Walter Walter Katz wkatz@xxxxxxxxxx Phone 303.449-2308 Mobile 303.335-6156 -- Scott McMorrow Teraspeed Consulting Group LLC 16 Stormy Brook Rd Falmouth, ME 04105 (401) 284-1827 Business http://www.teraspeed.com TeraspeedR is the registered service mark of Teraspeed Consulting Group LLC -- Scott McMorrow Teraspeed Consulting Group LLC 16 Stormy Brook Rd Falmouth, ME 04105 (401) 284-1827 Business http://www.teraspeed.com TeraspeedR is the registered service mark of Teraspeed Consulting Group LLC -- Scott McMorrow Teraspeed Consulting Group LLC 16 Stormy Brook Rd Falmouth, ME 04105 (401) 284-1827 Business http://www.teraspeed.com TeraspeedR is the registered service mark of Teraspeed Consulting Group LLC -- Scott McMorrow Teraspeed Consulting Group LLC 16 Stormy Brook Rd Falmouth, ME 04105 (401) 284-1827 Business http://www.teraspeed.com TeraspeedR is the registered service mark of Teraspeed Consulting Group LLC