All,
And I made these changes to bird181.2_draft1
Walter
From: ibis-editorial-bounce@xxxxxxxxxxxxx
[mailto:ibis-editorial-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz
Sent: Friday, April 14, 2017 10:40 AM
To: ibis-editorial@xxxxxxxxxxxxx; 'IBIS-Interconnect'
<ibis-interconn@xxxxxxxxxxxxx>
Subject: [ibis-editorial] Simple fix to BIRD 181.1 RE: Buf, Pad, Pin BIRD
189 and legacy IBIS package models.
All,
I re-read BIRD 181.1 thoroughly and found a simple, but what should be an
obvious error in this section.
The fundamental problem in this and similar sections:
Keywords: [Pulldown], [Pullup], [GND Clamp], [POWER Clamp]
Required: Yes, if they exist in the model
Description: The data points under these keywords define the I-V tables
of the pulldown and pullup structures of an output buffer and the I-V tables
of the clamping diodes connected to the GND and the POWER IBIS model pins,
respectively. Currents are considered positive when their direction is into
the component.
So, what is these simulator I-V tables connected to in the model? They are
connected to
"the GND and the POWER IBIS model terminals, respectively"
Walter
From: Walter Katz [mailto:wkatz@xxxxxxxxxx] ;
Sent: Thursday, April 13, 2017 10:29 AM
To: ibis-editorial@xxxxxxxxxxxxx <mailto:ibis-editorial@xxxxxxxxxxxxx> ;
IBIS-Interconnect <ibis-interconn@xxxxxxxxxxxxx
<mailto:ibis-interconn@xxxxxxxxxxxxx> >
Subject: Buf, Pad, Pin BIRD 189 and legacy IBIS package models.
All,
If we choose to include BIRD 189 and not BIRD 181 in IBIS 7.0 we could
include the following (either in a BIRD, or in a statement that accompanies
IBIS 7.0). The following is a firm grasp of the obvious. It is technically
correct.
In IBIS 6.1 and all previous revisions of IBIS, the I/O buffer terminals
where assumed to be at the die pads (die package interface). The on-die
interconnect between the die pad and the buffer terminals has always been
assumed to be included in the IBIS I/O buffer model, either in the IV curves
and/or in C_comp. At the high data rates of modern I/O buffers this is no
longer sufficiently accurate. BIRD 189 addresses this by allowing the
package interconnect model to be defined either between the pins of the
component and the I/O buffer signal and rail terminals, or to be split
between the pins of the component and terminals at the die-pad/package
interface and between the die-pad/package interface and the I/O buffer
terminals.
BIRD 189 package models that are defined between the pins and the I/O buffer
terminal models may or may not include the interconnect between the die-pads
and the I/O buffer terminals. If the package models do not include on-die
interconnect, then I/O buffer models need to be constructed as before, and
include the on-die interconnect in their IV curves and/or C_comp. If the
package models do include on-die interconnect, then I/O buffer models need
to be constructed (most likely through a SPICE-to-IBIS or similar process),
and not include the on-die parasitics that are included in the package
models.
There are many places in the current IBIS specification (6.1) that name the
I/O buffer terminals. These names did not anticipate the requirement that
these terminals could either be at the die-pad package interface, or
internal to the die and not at the die-pad interface. EDA tools shall assume
that all IBIS data is defined at the I/O buffer terminals, and that these
terminals are connected to BIRD 189 "Buf" terminals, and that there is no
double counting of on-die interconnect included in the I/O Buffer model and
the package interconnect.
Using legacy package models that assume the on-die interconnect is included
in the I/O buffer models when BIRD 189 package models will leave out any
on-die interconnect that is included in the BIRD 189 package models.
Walter
Walter Katz
wkatz@xxxxxxxxxx <mailto:wkatz@xxxxxxxxxx>
Phone 303.449-2308
Mobile 303.335-6156
Attachment:
bird181.2_draft1_wmk.docx
Description: application/vnd.openxmlformats-officedocument.wordprocessingml.document
