Walter,
This is something around the points I made in my IBIS Summit presentation about
4 weeks ago. Placing my comments into your questions below.
* What is the s element N+1 terminal in HSPICE hook up to?
Choosing reference node in the simulator is not the same thing as choosing
reference for ports when doing measurements. Each port has two "terminals", one
of which is considered its reference. When making measurements, we have to
choose the pair of terminals which are electrically close (the distance between
them less than the wavelength of the highest frequency of interest. Termination
on other ports can be provided by connecting a load to their corresponding
terminals. There is NO requirements that all ports should have the same
reference terminals. On the contrary, very often it is physically impossible to
do.
However, when S-parameters are measured (or constructed other way), they
acquire some properties not existing in the original object: (1) is electrical
isolation between the ports inside computational model, and (2) regularity of
each port that requires that the current entering "+" terminal or a port always
equal the current leaving "-" terminal (or reference) of that same port.
From these two things, it follows that in the circuit simulator Spice, ADMS,
many others, the reference terminals of all ports can be connected together and
the equations that relate port voltages/currents or waves, don't change. In
other words, in the simulator we can do things not possible in reality and
still have correct signals at ports.
* How can this be a single reference for the Ports at both ends of the
cable if they are 1 mile away?
It cannot be. Common reference in SPICE, (N+1)-st terminal, is an abstraction,
though useful.
* Does it matter what node in the simulation the N+1 terminal of the
s-element is connected to (in HSPICE, in other SPICE simulators)?
It depends on the task you are solving and the general topology of your design.
If for example you cascade a chain of s4p files, and have differential source
at one end and diff load an another, yes, you can connect all reference
terminals of all ports together, and the resulted port voltages/currents remain
valid. On other cases, such as combining S-parameters describing signal
propagation and PDN effects, this is not right and may produce wrong simulation
results.
* If it is OK to hook up this N+1 terminal to Node 0 in HSPICE, how do
simulators that claim they have no absolute ground (Node 0) hook up this N+1
terminal so they have a legal return path for each Port?
S-parameter models per se don't care about the node to which the ports'
references are connected to. They generate equations embedded into the model.
We need to watch for two things. One is to connect port to port in a way
consistent with their connection in a "real object". The other is making sure
that there exist a path to global ground from any node (otherwise, SPICE will
issue an error). However, when connecting references to Node 0, we should not
create additional contours that obscure the behavior of the S-parameter model
(I gave some simple rules about this in my presentation). On the other hand,
there is NO requirement that reference terminals must be connected to Node 0 or
any other special node, and there is no rule that all ports must have their
references connected to the same node. Sometimes, it's convenient to think this
way, but in other cases imposing such requirements doesn't allow us to
correctly simulate designs with slightly more complicated topology.
* How do suppliers who create s parameter data (measurement or solver)
expect EDA tools to hook up terminal N+1?
They cannot expect anything in particular because they don't know in which
topology this S-parameter model will be used.
* There are people that claim that there is no current flowing through
terminal N+1 because the S-Parameter assumption for interconnect is that the
current flowing from the signal terminal of a port is equal to the current
flowing to the reference terminal of the port. If this is true, then it should
not matter which node should be connected to the N+1 terminal. Is this a valid
argument?
Who are these people? I don't know them. If you connect all ports' terminals
together and make it a N+1-th terminal, then from current conservation law it
follows that the current that is leaving the model through this N+1-th terminal
equals the sum of all currents entering positive terminals of all those ports.
This happens due to rule #2 (regularity of each port) explained above.
Vladimir
From: ibis-interconn-bounce@xxxxxxxxxxxxx
[mailto:ibis-interconn-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz
Sent: Tuesday, February 27, 2018 7:48 AM
To: Scott@xxxxxxxxxxxxx; IBIS-Interconnect <ibis-interconn@xxxxxxxxxxxxx>;
IBIS-ATM <ibis-macro@xxxxxxxxxxxxx>; michael.mirmak@xxxxxxxxx
Subject: [ibis-interconn] Re: [ibis-macro] Re: S-parameter thought experiment
Scott,
Some subtle points here. Are you measuring the s4p or extracting an s4p from a
solver. Is the 1 mile cable coiled up so the test equipment has probes inches
apart or a mile apart. These are questions about the quality (accuracy) of the
s4p.
I will try to re-phrase the question. When making measurements or using a
solver (and a solver does measurement by probes in a simulation or 3-d field
solver, the measurement of a port at one end of the cable are made relative to
a terminal at that end of the terminal. Another way of saying this is that each
port is two terminals at the same end of the cable. What I believe Michael is
asking:
* What is the s element N+1 terminal in HSPICE hook up to?
* How can this be a single reference for the Ports at both ends of the
cable if they are 1 mile away?
* Does it matter what node in the simulation the N+1 terminal of the
s-element is connected to (in HSPICE, in other SPICE simulators)?
* If it is OK to hook up this N+1 terminal to Node 0 in HSPICE, how do
simulators that claim they have no absolute ground (Node 0) hook up this N+1
terminal so they have a legal return path for each Port?
* How do suppliers who create s parameter data (measurement or solver)
expect EDA tools to hook up terminal N+1?
* There are people that claim that there is no current flowing through
terminal N+1 because the S-Parameter assumption for interconnect is that the
current flowing from the signal terminal of a port is equal to the current
flowing to the reference terminal of the port. If this is true, then it should
not matter which node should be connected to the N+1 terminal. Is this a valid
argument?
Walter
Walter Katz
wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>
978.461-0449 x 133
Mobile 303.335-6156
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Scott McMorrow
Sent: Monday, February 26, 2018 8:29 PM
To: IBIS-Interconnect
(ibis-interconn@xxxxxxxxxxxxx<mailto:ibis-interconn@xxxxxxxxxxxxx>)
<ibis-interconn@xxxxxxxxxxxxx<mailto:ibis-interconn@xxxxxxxxxxxxx>>; IBIS-ATM
(ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>)
<ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>>;
michael.mirmak@xxxxxxxxx<mailto:michael.mirmak@xxxxxxxxx>
Subject: [ibis-macro] Re: S-parameter thought experiment
Michael
Define "described". What is the purpose of the black box measurements?
Any 2-wire, 4-terminal black box can be modeled with a measured s-parameter.
The question regarding the purpose is important, as it helps to define the
measurement needed.
Assume, for example, two wires in Free space. We can measure them in 2 ways. We
can use differential and common mode probing to create e black box. Or we can
create a virtual ground reference between the conductors, one in each end and
measure single ended s parameters.
In either case, the measurements are valid and identical. N ports are required.
However, unless the two wires are placed inside a Faraday cage the measurements
will be noisy, since they include the electromagnetic environment of the
universe. Instead, we use shorter sections that we measure, and then
extrapolate then out to one mile long.
I have no comment regarding the bird.
Get Outlook for Android<https://aka.ms/ghei36>
From: Mirmak, Michael
Sent: Monday, February 26, 7:36 PM
Subject: [ibis-macro] S-parameter thought experiment
To: IBIS-Interconnect
(ibis-interconn@xxxxxxxxxxxxx<mailto:ibis-interconn@xxxxxxxxxxxxx>), IBIS-ATM
(ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>)
(apologies if you have heard something like this before - I have discussed it
privately in a couple of contexts)
Imagine that I have a physical interconnect, consisting of a differential pair
of wires not particularly strongly coupled to each other. I want to model the
interconnect using an S-parameter set of at least 4 (four) ports: two inputs P
&N and two outputs P' & N'.
The problem: the physical pair of wires is approximately one mile in length
from P & N to P' & N'.
My questions:
1) Can this structure be correctly and unambiguously described using a
circuit with N terminals and an N+1 reference terminal? If not, how many
additional terminals are required?
2) Does the current BIRD189 draft, as written, enable correct and
unambiguous description(s) of this structure?
3) Does BIRD158.7, as written, enable correct and unambiguous
description(s) of this structure?
Thank you!
- MM
This email and any appended documents are only for the intended person/entity
and may contain information of Samtec, Inc., that is PRIVILEGED, PROPRIETARY,
CONFIDENTIAL, AND/OR PROTECTED BY LAW. If you are not the intended recipient
you are hereby notified that any dissemination, disclosure, use or copying of
this email or its contents is prohibited. If you received this message in
error, please notify Samtec immediately and delete the email, attachments and
all copies. The intended recipient should not disclose the content to third
parties or reproduce the content without Samtec's written consent.