[ibis-macro] Re: Analog Model Isolation Definition
- From: James Zhou <james.zhou@xxxxxxxxxx>
- To: Bob Ross <bob@xxxxxxxxxxxxx>, "fangyi_rao@xxxxxxxxxxx" <fangyi_rao@xxxxxxxxxxx>, "ibis-macro@xxxxxxxxxxxxx" <ibis-macro@xxxxxxxxxxxxx>
- Date: Fri, 23 Mar 2012 11:40:54 -0700
Hi Bob,
You've raised a very important issue of "make files interchangeable between
tools". I agree with you and echo that "As long as the official specification
makes this clear, then I am satisfied". As a user of IBIS models and EDA tools,
I would further argue strongly that models SHOULD be interchangeable between
tools. In other words, proprietary implementation is not the same thing as
proprietary interpretation. The former does not necessarily break
interchangeability if done properly, but the latter often does.
With reference to "My major conclusion is that the stimulus, parameters and
analog model would be specified as an independent group for each type of analog
model or for even IBIS buffers", if indeed this approach is adopted, the
Specification should clarify the differences between "each type of analog
model" so that they can be interchangeable between tools. Within the context of
IBIS AMI, there are two main types of independent analog models:
analog-in-analog-out(AIAO) and, logical-in-analog-out (LIAO), regardless of the
format of the model.
For example, the same T-coil circuit can be represented by RLC SPICE deck, or
S-parameter, or AMS. By hooking up the same source and load, the response
should be the same regardless of its format.
My comments on the four examples in your email (they are labeled Ex1,2,3,4 from
top to bottom):
(a) Ex1,2 involve s4p and Vt curve in IBIS file. s4p is analog-in-analog-out,
IBIS Vt curve is logic-in-analog-out. Some clarification is needed on how to
mix them together. For example, how could a logical-input be driven by an
analog signal? Or if the logical-input is not driven by anything, what is the
concept of an "analog model" that only has output but no input?
(b) Ex1,2,3 involve the "drive" of the analog model. I think the distinction
should be made between (i) the drive when deriving impulse response (ii) the
drive when obtaining Rx waveform in channel simulation. The concept of impulse
response does not apply to a logical-in-analog-out block represented by legacy
IBIS V-T curve. The specification needs to clarify such practices.
(c) Ex4 is a clear example of logical-in-analog-out, which could be a valid
example of "analog" circuit. The question is, how could such a model be
simulated with Tx IBIS AMI model? If such logical-in-analog-out blocks are
allowed by IBIS Specification, clarifications are needed on how to interpret
these data (which is different from how to implement the simulation).
Best Regards,
James Zhou
QLogic Corp.
-----Original Message-----
From: Bob Ross [mailto:bob@xxxxxxxxxxxxx]
Sent: Thursday, March 22, 2012 12:39 PM
To: fangyi_rao@xxxxxxxxxxx; James Zhou; ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Re: Analog Model Isolation Definition
All:
Thanks for your responses and discussion. I agree with
James Zhou below.
My major conclusion is that the stimulus, parameters
and analog model would be specified as an independent
group for each type of analog model or for even IBIS
buffers.
That makes the defined interface discussion moot since
the whole flow might be customized (and even scaled)
for each of the analog choices including s4p files.
As long as the official specification makes this clear,
then I am satisfied and does not overreach by implying
that the Analog blocks are completely interchangeable.
That is the key understanding needed make files interchangeable
between tools and also makes the scaling issue discussion
also moot.
The cookbook could follow later, both to codify existing
practices and to leave the door open to differing practices.
------
As an example of the differences and potential differences
so far:
TX: s4p - practice probably S11 = 0 (per the TI example
and some earlier historical IBM presentations) and s4p
files do not have Cc or Vt parameter (unless the parameter
is used to swap in a different s4p file. The frequency response
is in s4p file and it should be driven by a step. One vendor
could theoretically produce an s5p file with a Vt
port to pass in differential offset values. (Or Vt could
be used to swap in a different s4p file.) This should also work.
Another vendor could make the s4p file as a DC transfer and
move most of the frequency response (V-T) shaping) as a
simplified and controllable a parameterized Ramp input.
TX SPICE (or IBIS-ISS) - simplified fixed model - normally driven
by ramp waveform to approximate the source
frequency response and may have load parameters
for Cc, VT and anything else. So this is not interchangeable
with the s4p model without changing the driver and
parameters and possibly the parameter. Also, any type of
simplified model can be created with arbitrary parameters
that might be impractical with s4p TX models.
TX SPICE (or IBIS-ISS) - general TX driver - could be modeled
with a shaped frequency response and driven
by an ideal step (think Laplace transform pole-zero variables
in IBIS-ISS and isolation Op Amps (E elements)
(or cascaded Op Amps with RLC shaping networks) to configure
an extracted TX model frequency response. This would be
driven by an ideal step.
TX *_AMS models - driven by logical 0 - 1 stimulus
(only case without D_to_A converters), to produce
shaped TX responses - so the frequency shaping is
in the *_AMS model. (Not a real application yet, but
we see a lot of Verilog A embedded in SPICE models,
and this is supported in IBIS.).
Bob
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of fangyi_rao@xxxxxxxxxxx
Sent: Wednesday, March 21, 2012 11:23 AM
To: james.zhou@xxxxxxxxxx; bob@xxxxxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Re: Analog Model Isolation Definition
I agree with James. Spice simulations never need to make any assumption on
the values of S-parameters, and it should be same case here for IBIS analog
model specification. I don't think it's necessary to include examples of how
to treat special S-parameters. Matrix stamping of S-parameters are standard
textbook simulation techniques.
Regards,
Fangyi
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of James Zhou
Sent: Wednesday, March 21, 2012 10:53 AM
To: bob@xxxxxxxxxxxxx; 'IBIS-ATM'
Subject: [ibis-macro] Re: Analog Model Isolation Definition
Hi Bob,
Thanks for providing the link of past references. It is helpful for those of
us who were not on the list a couple of years ago.
With reference to: "the complete interface needs to be known and defined ...
to get the proper gain (S21 term).", this is a proven and well-known
principle in circuit and LTI network analysis. The conclusion at yesterday's
meeting sufficiently satisfies this requirement. The conclusion as I
interpreted was that, IBIS Specification only need to say that the Tx analog
circuit input must be driven by ideal voltage sources (i.e. zero source
impedance), and Rx analog circuit output must drive a load with infinite
impedance. Once this is established, the Specification does not need to say
anything about what goes into, or how to model, the "bigger blockbox" in the
diagram (provided it satisfies other IBIS requirements such as LTI and
IBIS-ISS compliant, etc.).
With reference to: "In both cases, the transfer function S21 term is changed
by a factor of 2...", these two cases represent total reflection by open or
short circuits (S11=1 or -1). In both cases, S21 must be 0. This result is
consistent with the general formula of S-parameter modeling when S11 can
have arbitrary values (including the cases of S11=1 or -1). The treatment
for these specific cases of open and short circuits can be provided in the
Specification as examples instead of requirements of the Specification.
On the issue of setting S12=0, there have been IBIS discussions or
presentations recommending this. However, it is not necessary to impose this
condition on the analog circuits and it should not be treated as a
requirement of the BIS Specification.
Regards,
James Zhou
QLogic Corp.
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Bob Ross
Sent: Tuesday, March 20, 2012 2:57 PM
To: 'IBIS-ATM'
Subject: [ibis-macro] Analog Model Isolation Definition
All:
Per the discussion today on the block interface definition, here is a
derivation based on BIRD122 for just s2p 2-ports (but should apply to s4p
files with more math). It shows the gain sensitivity to the explicit S11
term for TX and S22 term for RX from a reference impedance definition (e.g.,
TX: S11=0 or 50 ohm to infinite impedance S11 = 1 and RX: S22=0 to zero
impedance S11=-1).
In both cases, the transfer function S21 term is changed by a factor of 2
(assuming no S12 term). This is equivalent to specifying the isolation
interfaces with a unity gain ideal voltage amplifier.
//www.freelists.org/post/ibis-macro/Eelement-in-analog-model-BIRD,1
Sorry, but some of the prior material may have involved some lost or private
e-mails and also a presentation or e-mail shown to the ATM committee
according to questions and exchanges in the Dec. 14, 2010 minutes including:
Walter showed an email from Fangyi:
- Fangyi: The equation is at the bottom
- Bob: We have defined S12 as zero
- The real voltage gain is a function of S11
- Todd: Is this because S11 can affect V1 and therefore V2?
- Fangyi: No
- If V1 was fixed V2 would still be affected by S11
- Bob: The defaults should be zero for series impedances and
infinity for impedances to ground.
http://www.vhdl.org/ibis/macromodel_wip/minutes/20101214/mikelabontecisco/Mi
nutes%20for%20the%2012-14-2010%20ibis-atm%20meeting/20101214.txt
The main point is that the complete interface needs to be known and defined
in the Specification (for S-parameters and perhaps different for a SPICE
model) to get the proper gain (S21 term).
Bob
--
Bob Ross
Teraspeed Consulting Group, LCC
http://www.teraspeed.com
bob@xxxxxxxxxxxxxx
Direct : 503-246-8048
Teraspeed Labs: 503-430-1065
Headquarters: 401-284-1827
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
This message and any attached documents contain information from QLogic
Corporation or its wholly-owned subsidiaries that may be confidential. If
you are not the intended recipient, you may not read, copy, distribute, or
use this information. If you have received this transmission in error,
please notify the sender immediately by reply e-mail and then delete this
message.
---------------------------------------------------------------------
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
This message and any attached documents contain information from QLogic
Corporation or its wholly-owned subsidiaries that may be confidential. If you
are not the intended recipient, you may not read, copy, distribute, or use this
information. If you have received this transmission in error, please notify the
sender immediately by reply e-mail and then delete this message.
---------------------------------------------------------------------
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:
