Scott,
Assuming “You” is the IC vendor, he has the following choices:
1. Lump the RDL into the package interconnect.
2. Lump the RDL into the on-die interconnect.
3. Treat the silicon and the RDL as a package.
a. RDL being the package model
b. Then mount this “RDL Package” into a substrate using EBD (or EMD the
“BIRD 189” version of EBD)
Walter
From: Scott McMorrow [mailto:Scott@xxxxxxxxxxxxx]
Sent: Tuesday, March 14, 2017 2:24 PM
To: Walter Katz <wkatz@xxxxxxxxxx>; Arpad_Muranyi@xxxxxxxxxx;
ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Re: Comments on BIRD158
Thanks for the overview. I get where you are going. I assume for the
purposes of on-die interconnect, you would most like lump in the RDL layer,
since it is most closely related to the silicon.
Scott McMorrow, CTO Signal Integrity Group
Samtec
Office 401-284-1827 | +1-800-726-8329
<http://www.samtec.com/> www.samtec.com
From: Walter Katz [mailto:wkatz@xxxxxxxxxx]
Sent: Tuesday, March 14, 2017 2:12 PM
To: Scott McMorrow <Scott@xxxxxxxxxxxxx <mailto:Scott@xxxxxxxxxxxxx> >;
Arpad_Muranyi@xxxxxxxxxx <mailto:Arpad_Muranyi@xxxxxxxxxx> ;
ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: RE: [ibis-macro] Re: Comments on BIRD158
Scott,
The short answer is No, Yes, Maybe, and All of the Above. It is undefined
for BIRD 158, for [External Model] and for [Model]. What we end up doing is
reading the documentation that comes with the model and insert the
appropriate interconnect model between the “Buffer Model”, and the pin of
the component.
Any of the existing IBIS Packaging models do not have sufficient accuracy to
use in SerDes (and DDR) interfaces. We are trying to solve this problem
(BIRD 189) by first allowing broad band interconnect models. These models
can be I/O, PDN, coupled and uncoupled. We also decided to allow
interconnect models from pin to buffer, and to optionally split them up from
pin to pad (package) and pad to buffer (on-die). Bottom line is that we
wanted this new interconnect standard to interface how interconnect models
are currently being created (not for IBIS but for current modeling
requirements).
I think as a practical matter, most I/O buffer models delivered as
Touchstone files include the buffer and on-die interconnect to the pad. I
also think that most Classical IBIS [Models] include the parasitics to the
chip package interface, and I think the standard implies but does not
require this. Some of us on the committee have a different opinion on this
point.
One of the reasons for enabling partitioning the component interconnect
between on-die and package was power deliver. Different die and be used with
the same package, and different packages can be used with the same die. It
could be convenient, and reduce library maintenance by supplying independent
on-die and package models and having the EDA tool marry the two, rather than
delivering combined package and on-die models for each die/package
configuration.
What we are currently limiting ourselves to is packages with a single die,
and the die is mounted to the substrate using flip chip or wire bond, or
similar connection methods. It is our intention to support more complex
structures such as multichip, stacked die, interposers, … when we extend EBD
to support the same interconnect model interface as in BIRD 189.
I hope this clarifies the situation.
Walter
From: ibis-macro-bounce@xxxxxxxxxxxxx
<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Scott McMorrow
Sent: Tuesday, March 14, 2017 1:28 PM
To: Arpad_Muranyi@xxxxxxxxxx <mailto:Arpad_Muranyi@xxxxxxxxxx> ;
ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: Comments on BIRD158
On-die interconnect?
When this is referenced, what does it mean?
1) Interconnect physically on the active substrate?
2) Silicon, Glass, or other interposer between the substrate and the
package?
3) The redistribution layer?
4) All of the above?
Scott McMorrow, CTO Signal Integrity Group
Samtec
Office 401-284-1827 | +1-800-726-8329
<http://www.samtec.com/> www.samtec.com
From: ibis-macro-bounce@xxxxxxxxxxxxx
<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, March 14, 2017 1:23 PM
To: ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: Comments on BIRD158
Walter,
You are raising an interesting point. I think you are correct
in some ways, but not completely.
A [Model] can only have I-V and V-t curves. How would you describe
a package or on-die interconnect with I-V and V-t curves? I think
that is pretty impossible… So for [Model] I would think that it is
pretty obvious that it is buffer and buffer only, consequently its
connection point is at “buffer terminal”. In legacy IBIS, where we
don’t have on-die interconnect this is the same as pad.
The story gets more complicate when we have [External Model] inside
a [Model] because [External Model] is allowed to have SPICE (IBIS-ISS)
subcircuits, which ***could*** be used to describe interconnects.
However, the spec states (pg 91) that:
“Same as [External Circuit], except limited to the connection format and
usage of the [Model] keyword”,
which to me means that the way it is connected is identical to the
way [Model] is connected, and with the above argument we established
already that [Model] can’t be connected any other way than at the
“model terminal” points.
So I don’t think we need to worry that [Model] or [External Model]
would include on-die or package modeling. But, it would never hurt to
make that extremely clear, perhaps by adding a statement on that if
necessary.
Thanks,
Arpad
======================================================================
From: Walter Katz [mailto:wkatz@xxxxxxxxxx]
Sent: Tuesday, March 14, 2017 12:00 PM
To: Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx
<mailto:Arpad_Muranyi@xxxxxxxxxx> >; ibis-macro@xxxxxxxxxxxxx
<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: RE: [ibis-macro] Re: Comments on BIRD158
Bob, Arpad,
Do we not have the same issue for [Model] and [External Model]? It will be
possible for one [Model] to have legacy IV/VT/C_comp/C_comp_model
(C_comp_model is the proposed C_comp model BIRD), an [External Model], and a
Ts4file. I would expect that each one of these could have a different
“?_includes”. Should we not have:
In the .ami reserved parameters section
(Ts4file_Includes <"buffer"|"pad"|"package">)
In the .ibs [External Model] section
External_model_includes <buffer|pad|package>
In the .ibs [Model] section
Model_includes <buffer|pad|package>
I believe in all cases the default would be “pad”.
Walter
From: ibis-macro-bounce@xxxxxxxxxxxxx
<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Tuesday, March 14, 2017 1:53 AM
To: ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: Comments on BIRD158
Bob,
I like your idea of using a single parameter with a list of
options for the various cases. (I actually thought of something
similar as I was writing my email, I just didn’t bake it long
enough to be able to verbalize it)…
Regarding your two reasons for being hesitant to endorse it, I
don’t quite understand your reasons.
#1) I don’t see this as a constraint, because we can add more
entries to the list as new cases arise (although in this situation
I don’t expect a whole lot of now boundaries or connectivities
to be invented).
#2) This is absolutely useful and needed information for the EDA
vendors. How would they know how to generate a netlist correctly
without this information, especially if the .ibs file contains
“extra” information which pose multiple options to the EDA tool
regarding how to connect the buffer models? For example, if the
.ibs file contains an on-die interconnect model, a package model,
and, the buffer model includes everything all the way to the pin/ball,
we need to tell the EDA tool what is in the buffer model or where
its boundaries are so that it could generate the netlist correctly
without double counting for the on-die interconnect and the package
models…
Thanks,
Arpad
=======================================================================
From: radek_biernacki@xxxxxxxxxxxx <mailto:radek_biernacki@xxxxxxxxxxxx>
[mailto:radek_biernacki@xxxxxxxxxxxx]
Sent: Monday, March 13, 2017 5:50 PM
To: dmarc-noreply@xxxxxxxxxxxxx <mailto:dmarc-noreply@xxxxxxxxxxxxx> ;
Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx <mailto:Arpad_Muranyi@xxxxxxxxxx> >
Cc: ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: RE: [ibis-macro] Re: Comments on BIRD158
Hi Bob,
Not a bad suggestion (though you are hesitant). I think the point is that
you do not constrain the use of the data, nor impose a requirement on the
model maker, but rather facilitate a way for the model maker to inform the
world what the data in the file was measured on.
Regards,
Radek
From: ibis-macro-bounce@xxxxxxxxxxxxx
<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Bob Miller
Sent: Monday, March 13, 2017 1:37 PM
To: Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx
<mailto:Arpad_Muranyi@xxxxxxxxxx> >
Cc: IBIS-ATM <ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx> >
Subject: [ibis-macro] Re: Comments on BIRD158
I'll put my oar in <wink>...
If we decide that the model maker, via reserved parameter(s) must tell the
user in a formal way that the Ts4file includes stuff beyond the buffer
termination impedance, I think it can be done with one parameter, Type List,
Format String. E.g. "(Ts4file_Includes <"buffer"|"pad"|"package">)".
I'm hesitant to endorse this, however, because:
* we necessarily constrain the use of ts4file by assuming we know how it
will always be used.
* we impose a requirement on the model maker but apparently none on the
EDA
tool to actually do something useful with it
Regards,
Bob
On Mon, Mar 13, 2017 at 12:50 PM, Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx
<mailto:Arpad_Muranyi@xxxxxxxxxx> > wrote:
Walter,
Regarding “If the model maker has on-die interconnect or even package models
included in the Ts4file, then he should make sure that any package/on-die
interconnect that is in the Ts4file not be included in the
package/interconnect model. This is the mechanism…”,
I agree. What I was saying is that we might need two additional
parameters for that if we consider on-die interconnects, as opposed
to one additional parameter if we only considered the package model,
because if we need to tell the tool that the Ts4file includes either
one or both, the tool needs to know which one(s). That “For discussion”
section was only talking about one additional parameter that tells
the tool whether the package is included in the Ts4file or not.
Thanks,
Arpad
========================================================================
From: Walter Katz [mailto:wkatz@xxxxxxxxxx ;<mailto:wkatz@xxxxxxxxxx> ]
Sent: Monday, March 13, 2017 12:35 PM
To: Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx
<mailto:Arpad_Muranyi@xxxxxxxxxx> >; IBIS-ATM <ibis-macro@xxxxxxxxxxxxx
<mailto:ibis-macro@xxxxxxxxxxxxx> >
Subject: RE: [ibis-macro] Comments on BIRD158
Arpad,
I general agree with your changes and comments with one exception:
“#### For discussion: The model maker may want to inform the user and the
EDA tool that the Ts4file includes the package. Then a separate parameter is
needed. Alternatively, the model maker may provide the s4p data of the
package.”
The above discussion will also have to include the on-die interconnect, made
possible
by BIRD189. Possibly through another additional parameter…
Any package interconnect between the pin and the buffer defined in the IBIS
file should be inserted between the pin and
* The “B” element for classical [Model]s
* The External Model for [External Model]s
* The Ts4file when using Ts4file in the .ami file
If the model maker has on-die interconnect or even package models included
in the Ts4file, then he should make sure that any package/on-die
interconnect that is in the Ts4file not be included in the
package/interconnect model. This is the mechanism that he “informs the user
and the EDA tool that the Ts4file includes the package”.
Walter
From: ibis-macro-bounce@xxxxxxxxxxxxx
<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad
Sent: Monday, March 13, 2017 1:08 PM
To: IBIS-ATM <ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx> >
Subject: [ibis-macro] Comments on BIRD158
Hello Everyone,
I decided to read through BIRD158 in great detail because of the
disagreement
Walter and I have on whether this BIRD should appear in the same IBIS
version
with BIRD189 or not. As I am reading BIRD158.6 in “final mode” (markups
hidden),
I am finding quite a few mostly editorial issues that should be corrected:
Correct the spelling of: “makde”
Correct the spelling of: “pachage”
Remove the extra “s”: “Stimulus s calculating”
Correct the spelling of: “cimpulse”
“Reserved Parameters” should either have an underscore between the words or
be spelled lower case. If using it as the name of the subparameter with the
underscore, it should always have the “s” at the end. Otherwise use lower
case first letters…
Change: “This BIRD defines new AMI Reserved Parameters Ts4file, Tx_V, Tx_R,
Rx_R.”
to something like:
“This BIRD defines four new AMI Reserved_Parameters: Ts4file, Tx_V, Tx_R,
Rx_R.”
Change: “The ports 1 and 3 are at the Tx side, and ports 2 and 4 are
connected to the channel. Furthermore, the ports 1 and 2 correspond to the
positive signal path and the ports 3 and 4 to the negative signal path.”
to something like:
“Ports 1 and 3 are at the stimulus source side, and ports 2 and 4 are
connected to the buffer terminals. Furthermore, ports 1 and 2 correspond to
the non-inverting signal path and the ports 3 and 4 to the inverting signal
path.”
Change: “The ports 1 and 3 are connected to the channel, and the ports 2
and 4 serve as the differential input to the Rx algorithmic model.”
to something like:
“Ports 1 and 3 are connected to the buffer terminals, and ports 2 and 4
serve as the differential input to the Rx algorithmic model.”
Regarding the above sentence I wonder, whether the input to the Rx
algorithmic
model is really at ports 2 and 4. Is this box really in series with the
signal,
as if it was a package model? Or is the input to the algorithmic model at
ports
1 and 3? (The BIRD160 approach with [External Model] actually allows for
both).
Keeping the new capabilities described in BIRD189 in mind, the following two
sentences should also mention the possibilities of on-die interconnects:
“This Impulse Response characterizes the differential response of the Tx
analog buffer model, the Tx on-die interconnect model (if exists), the Tx
package model, the interconnect between the Tx component and the Rx
component (the channel), the Rx package model, the Rx on-die interconnect
model (if exists) and the Rx analog buffer model. The Touchstone file
defined in this BIRD is to be used for either the Tx analog buffer excluding
the Tx package and Tx on-die interconnect model and/or the Rx analog buffer
model excluding the Rx package and Rx on-die interconnect model.”
This is the reason I think BIRD158 and 189 should be added to the same IBIS
version.
I agree that this could be added with a later BIRD, but why shouldn’t we do
it now
“proactively” when we are fairly certain that BIRD189 will become part of
the spec?
If we decided to add BIRD158 to IBIS now and BIRD189 in a later version, we
would
have to write a “companion” BIRD to 189 immediately to adjust the
connectivity
around these buffer models so that they would be consistent with the on-die
interconnect
capability described in BIRD189. I think we could save ourselves a bunch of
time
and work if we did this now and added the two BIRDs to the same IBIS
version… The
same argument applies to the following discussion item:
“#### For discussion: The model maker may want to inform the user and the
EDA tool that the Ts4file includes the package. Then a separate parameter is
needed. Alternatively, the model maker may provide the s4p data of the
package.”
The above discussion will also have to include the on-die interconnect, made
possible
by BIRD189. Possibly through another additional parameter…
How many Tx_R parameters do we have? I only see one, so it should be
singular
in the following sentence:
“For Tx models that have the Reserved Parameter Ts4file, the Reserved
Parameter Tx_V is required and the Reserved Parameters Tx_R are optional.”
This sentence should be changed from:
“In other words, for a Tx buffer, the Transmitter Circuit defines the analog
buffer model between the zero impedance stimulus input voltage source and
the channel.”
to:
“In other words, for a Tx buffer, the Transmitter Circuit defines the analog
buffer model between the zero impedance stimulus input voltage source and
the buffer terminals.”
Similarly, the next sentence should be changed from:
“For an Rx buffer, the Receiver Circuit defines the analog buffer model
between the channel and a high impedance probe at the input to the Rx
Algorithmic model.”
to:
“For an Rx buffer, the Receiver Circuit defines the analog buffer model
between the buffer terminals and a high impedance probe at the input to the
Rx Algorithmic model.”
Note that in the paragraph that discusses the drawing of the whole system,
the word “channel” is used in a different way:
“This Impulse Response characterizes the differential response of the Tx
analog buffer model, the Tx package model, the interconnect between the Tx
component and the Rx component (the channel), the Rx package model and the
Rx analog buffer model.”
Here, “channel” does not even include the package (or potentially the on-die
interconnect). This will confuse the reader, not knowing what really the
“channel” consists of.
Please change (we can’t mention “this BIRD” in text that goes into the spec:
“the other component’s contribution to the Channel Step Response (or just
the cimpulse response) follows the .ibs file description, as has been the
case before this BIRD has been introduced.”
to:
“the other component’s contribution to the Channel Step Response (or just
the impulse response) is described by the [Model] keyword in the .ibs file,
as has been the case before the Ts4file, Tx_V, Tx_R, Rx_R parameters have
been introduced.”
In the Definition section of Ts4file, change:
“This parameter contains the name of 4-port Touchstone file to be used in
the Analog Model Circuit. The If the file contains 4-port S-parameter data”
to:
“This parameter contains the name of the 4-port Touchstone file to be used
in the Analog Model Circuit. If the file contains 4-port S-parameter data”
In the Definition of Tx_R, we should probably use plural for the
resistor at the end of the sentence, since there are two of them:
“This parameter is optional and defines the value of the Tx_R series
resistors in ohms.”
While most of these comments are editorial in nature, my main point is
that with a minimal change we could make this BIRD consistent with BIRD189,
and doing that could save us a lot of additional work and time later when
BIRD189 is added to IBIS. I would highly recommend that we make these
changes to this BIRD now and add these two BIRDs together to IBIS.
Thanks,
Arpad
============================================================================
