[ibis-macro] Re: EMD Pole-Zero Models (Why voltage controled sources?)
- From: "C. Kumar" <kumarchi@xxxxxxxxx>
- To: Arpad_Muranyi@xxxxxxxxxx, scott@xxxxxxxxxxxxx
- Date: Mon, 30 Jun 2008 09:08:46 -0700 (PDT)
translating to pole/zero merely hides the inadequacy of the original data.
There is no guarantee that pole/zero produced by two different vendors will
give the same results. They can yield surprising and unacceptably different
answers especially if pole/zero is used non transparently used to extrapolate
the original data. Even the same may produce inconsistent results based on
frequency range. The fundamental problem is the s-parameter data not pole/zero
derived from it.
The only way this problem can be solved is if everybody applies the same
algorithm for pole/zero extraction. (so even if it is incorrect everybody is
equally incorrect!!)
--- On Mon, 6/30/08, Scott McMorrow <scott@xxxxxxxxxxxxx> wrote:
From: Scott McMorrow <scott@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: EMD Pole-Zero Models (Why voltage controled sources?)
To: Arpad_Muranyi@xxxxxxxxxx
Cc: "IBIS Macro" <ibis-macro@xxxxxxxxxxxxx>
Date: Monday, June 30, 2008, 11:41 AM
Arpad
I think the piece you are missing is why controlled sources and
pole-zero are needed. There are several reasons:
1) currently a number of vendors have realized that s-parameter models
do not always simulate well in a number of environments. Sometimes they
are not implemented, and when they are, simulation is often slow,
and/or problematic in numerous ways.
2) having realized this, some companies have developed tools to
translate s-parameters into pole/zero approximations using various
methods. These models are often corrected for passivity and run
extremely fast in classical environments like HSPICE. Ansoft, Sigrity
and IDEMworks, among others, can all provide excellent pole-zero models
from S-parameters or the output of their tools.
3) some manufacturers are already delivering models in multiple forms:
coupled ideal spice, s-parameter, and HSPICE pole-zero format.
4) implementation of the HSPICE pole-zero format requires G-element or
E-element controlled sources, because these are the elements that
implement pole-zero polynomials in that software.
5) There are two ways to implement coupling in classical coupled spice
models, using either K-elements or controlled sources. Controlled
sources are sometimes preferred for simulator performance reasons.
Ansoft coupled Spice models developed by their quasi-static 3D FEM tool
use controlled sources for coupling.
I would restrict the use of controlled sources to only those cases
where they are needed. Otherwise you get into the case where a
controlled source could be arbitrarily placed in signal path, thus
creating an amplified model, or a model with feedback, which then could
break the LTI assumption.
Scott
Scott McMorrow
Teraspeed Consulting Group LLC
121 North River Drive
Narragansett, RI 02882
(401) 284-1827 Business
(401) 284-1840 Fax
http://www.teraspeed.com
Teraspeed® is the registered service mark of
Teraspeed Consulting Group LLC
Muranyi, Arpad wrote:
Walter,
Let's put aside for a moment all detail
questions about LTI,
the format of
pole/zero, and consistency. My question still
stands:
How are we going to guarantee that this
new language will
translate into
anyone's SPICE easily if we put NEW capabilities
into it
(such as pole/zero and impulse response based T-line
data)
which were never supported in any SPICE before? I can
see how
limiting ourselves to LTI may help to simplify this
new
Interconnect-SPICE specification so that this translation
may be
doable (although when it comes to the controlled sources
I am not
that sure), but I don't see how the new T-line features
could
be translated easily. Do you have any thoughts on how that
would be done?
Thanks,
Arpad
===================================================================
From:
ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter
Katz
Sent: Monday, June 30, 2008 8:37 AM
To: kumarchi@xxxxxxxxx; bob@xxxxxxxxxxxxx
Cc: IBIS Macro
Subject: [ibis-macro] Re: EMD Pole-Zero Models (Why voltage
controled sources?)
Kumar,
There is an
interesting analogy between Pole-Zero vs. S element, and RLGC with Ro,
Lo, Co, Go, Rs, Gd data vs. R, L, G, C table data. When ICM was
developed, IBIS chose to exclude Ro, Lo, Co, Go, Rs, Gd because it was
only an approximation to R, L, G, C table data. Should we be consistent and
exclude
Pole-Zero, or should we include both Pole-Zero and Ro, Lo, Co, Go, Rs,
Gd data?
The first
question is “Do we implement Pole-Zero?”. An independent question is
“What format do we choose for Pole-Zero?”. First lets resolve the first
question. You said “I fail to see how
IBIS 'inventing' its own form and convince vendors and users to adopt
the yet another new/even improved form.”. But did we not
take this path already with ICM. There were several forms of s elements
and RLGC element instances but IBIS chose to invent a new one for ICM.
Should we be consistent
with the ICM approach or consistent with one of the various proprietary
simulators.
I believe that
one needs to pick a form that can easily be translated into all EDA
vendors’ simulator tools. Please remember that there are unique node
and element naming rules in each simulator (case sensitive for one).
Should we be consistent
with Hspice, Verilog, VHDL, QCD, …
Some notable
quotes on being “Consistent”
Consistency
requires you to be as ignorant today as you were a year ago.
Bernard
Berenson (1865 - 1959)
Consistency is
the last refuge of the unimaginative.
Oscar Wilde
(1854 - 1900)
A foolish
consistency is the hobgoblin of little minds, adored by little
statesmen and philosophers and divines.
Ralph
Waldo Emerson (1803 - 1882),
Self-Reliance
I would be consistent with “If
IBIS approves a standard, the members of IBIS should the new standard."
What this means to me is that the choice of including Pole-Zero
elements and the format of a Pole-Zero element should be driven by the
needs of the user committee, the ability of vendors to create these
models, and the unanimous consent of EDA vendors to agree that they can
translate these models into their simulation environment.
Walter
-----Original
Message-----
From: C. Kumar
[mailto:kumarchi@xxxxxxxxx]
Sent: Monday, June 30,
2008 7:25 AM
To: bob@xxxxxxxxxxxxx;
wkatz@xxxxxxxxxx
Cc: IBIS Macro
Subject: Re:
[ibis-macro] Re: EMD Pole-Zero Models (Why voltage controled sources?)
before
you go on to pole/zero approximation you should have an 'S' elements
which directly provides the original s-parameter data either in time
domain or freq domain. all the others like pole/zero are
approximations, except in the case of circuit elements with analytic
forms.
You can add to your list controlled sources with data provided multi
dimensional tables; controlled elements with hysterisis; a generic
expression controlled source of the form f(v,i, parameters)=0 and so on
and so forth. It is possible to reduce the controlled sources to a
finite set (order of 10;s)
however I am still with Arpad on this one here, unless I am not seeing
something fundamental . These elements have been already implemented in
equivalent forms in various flavors of spice and yes AMS and -A
flavors. I fail to see how IBIS 'inventing' its own form and convince
vendors and users to adopt the yet another new/even improved form.
--- On Mon, 6/30/08, Walter Katz <wkatz@xxxxxxxxxx>
wrote:
From:
Walter Katz <wkatz@xxxxxxxxxx>
Subject: [ibis-macro] Re: EMD Pole-Zero Models (Why voltage controled
sources?)
To: bob@xxxxxxxxxxxxx
Cc: "IBIS Macro" <ibis-macro@xxxxxxxxxxxxx>
Date: Monday, June 30, 2008, 12:54 AM
Bob,
Thanks for digging this up on the use of controlled voltage source for the
Laplace and Pole-Zero form of transfer functions.
Is the following a correct summary:
* A Touchstone file is a matrix of "Transfer Functions", where the
"Transfer
Function" is represented as a "Vector" of complex coefficients.
o Each element of the "Vector" is the amplitude of the "Transfer
Function"
at a specific
frequency.
* Each of the "Transfer Functions" can be translated to Laplace form
with a
numerator and denominator polynomial.
o Hspice implements the Laplace form using the E and G LAPLACE controlled
voltage source.
* The numerator and denominator Laplace polynomials can be factored, the
numerator polynomial factored into a list of zeros, and the denominator
factored into a list of poles.
o Hspice implements the Pole-Zero form using the E and G POLE controlled
voltage source.
* Alternatively, each of "Transfer Functions" can be translated into
"Impulse Responses"
If this is correct, then Lossy RLGC, Touchstone, Laplace, Pole-Zero
interconnect blocks can simply be represented as an EMD "Block" of
the form:
EMD_Block_xxxx <list of nodes> len=<length> type=<type>
file=<file>
* Where
o EMD_Block_xxxx
* Instance Designator
o <list of nodes>
* List of
nodes
o <length>
* Length of interconnect in meters (applies only to RLGC)
o <type>
* RLGC
* Touchstone
* Laplace
* Pole
* Impulse
o <file>
* RLGC
* Contains RLGC table data
* Touchstone
* sNp
* Laplace
* Contains Laplace polynomial coefficients
* Format needs specification
* Pole
* Contains Pole Zero data
* Format needs specification
* Impulse
* Contains Impulse Response data
* Format needs specification
It is a trivial exercise to convert any one of these "Types" of
EMD_Blocks
to Hspice W, S, E and G elements.
If all of the above is correct, then there is no need for EMD_Blocks that
are specifically voltage controlled sources.
To answer your question: An EMD models for a group of interconnect pins is
essentially an ICM [Nodal Path Description] where each of the N_sections is
essentially an EMD_Block. The [Nodal Path Description]
becomes a subckt with
nodes consisting of external EMD pins and IBIS component pins. In ICM all of
the N_section can either be all RLGC or all Touchstone. In EMD the
EMD_Blocks can be any combination of RLGC, Touchstone, Laplace, Pole,
Impulse, Resistor, Capacitor, Inductor, Conductance or K (coupling)
elements.
Walter
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx]On Behalf Of Bob Ross
Sent: Sunday, June 29, 2008 10:53 PM
To: wkatz@xxxxxxxxxx
Cc: IBIS Macro
Subject: [ibis-macro] Re: EMD Pole-Zero Models (Why voltage controled
sources?)
Walter:
Here are some responses to your questions
1. HSPICE and some other SPICEs implement the Laplace and pole-zero
elements as a network function WITHIN controlled sources including
the VCVS (E) and VCCS (G) elements. The documention is hard to
find, but the HSPICE
syntax is in the HSPICE Applications Manual:
Exxx n+ n- LAPLACE in+ in- k0 k1 ... kn / b0 b1 ... bm
Gxxx n+ n- LAPLACE in+ in- k0 k1 ... kn / b0 b1 ... bm
Exxx n+ n- POLE in+ in- a {cmpl zeros) / b (cmpl poles}
Gxxx n+ n- POLE in+ in- a {cmpl zeros) / b (cmpl poles}
2. I think of a pole-zero block, not as the single Laplace transfer
element, but as an n-port block such as proposed in some private
Touchstone-like formats and possibly implemented internally
and automatically from n-port table data.
My question:
When you say interconnect block modules of Resistor/Inductor/
Capacitor, do you really mean a low-level SPICE or SPICE-like
syntax within "SPICE" subcircuits for interconnect structures?
That is where K and controlled sources are valuable for many
reasons. While we have not really discussed this, I have been
assuming that we need such low-level capability for EMD. We
could formally add a basic SPICE-syntax subcircuit to the list
below as one of the modules with its internal SPICE-like netlist
used for connecting the R/L/C/K/E/F/G/H ... elements.
Bob
Walter Katz wrote:
> All,
>
>
>
> Based on the following assumptions for an EMD:
>
>
>
> * A module as a netlist of IBIS components and external pins
> * Interconnect models between these IBIS component pins consist of a
> netlist of interconnect blocks
> * Interconnect block models are:
> o Resistors
> o Inductors
> o Capacitors
> o Distributed RLGC models
> o S parameter Models
> o Impulse Response Models
> o Pole-Zero Models
>
>
>
> The purpose of this e-mail is to raise the issue of what is a Pole-Zero
> model and why do we need voltage controlled sources.
>
>
>
> I refer to http://www.ece.uci.edu/docs/hspice/hspice_2001_2-217.html
>
>
>
>
> Understanding Pole/Zero Analysis
>
> In pole/zero analysis, a network is described by its network transfer
> function which, for any linear time-invariant network, can be written in
> the general form:
>
>
>
> In the factorized form, the general function is:
>
>
>
> It seems to me that a Pole-Zero model can either be represented as a set
> of numbers like the polynomial coefficients a0, b0, a1, b1, a2, b2, ..
> or the factorized form a0, b0, z1, p1, z2, p2, ?
>
> Where is the controlled voltage source?
>
> I assume that one can model
the pole-zero form into Spice, Verilog, and
> VHDL primitives, and doing so might utilize controlled voltage sources
> and other simulator specific models.
>
> Why it is not sufficient to just have a Pole-Zero model (either with
> polynomial coefficients and/or pole-zero coefficients).
>
> Walter
>
--
Bob Ross
Teraspeed Consulting Group LLC Teraspeed Labs
121 North River Drive 13610 SW Harness Lane
Narragansett, RI 02882 Beaverton, OR 97008
401-284-1827 503-430-1065
http://www.teraspeed.com ; 503-246-8048 Direct
bob@xxxxxxxxxxxxx
Teraspeed is a registered service mark of Teraspeed Consulting Group LLC
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector:
//www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector: //www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
Other related posts:
