Bob,
Re technical comment,
This was one of the major issues that SiSoft and Keysight had to work out.
We each had customers who took different approaches on how to generate the
Touchstone file. Each of the customers of course thought that they were
doing it the only correct way. This is the solution that we came up with
that made everyone happy.
Walter
From: Bob Miller (APD) [mailto:bob.miller@xxxxxxxxxxxx]
Sent: Tuesday, January 10, 2017 5:44 PM
To: Walter Katz <wkatz@xxxxxxxxxx>
Cc: IBIS-ATM <ibis-macro@xxxxxxxxxxxxx>
Subject: Re: [ibis-macro] Example BIRD 158 .ami file
...and here's a technical comment about 158.4:
I'd like to point out that the use of Tx_R (R1 an R2 in the transmitter
circuit diagram) and Rx_R (R3 and R4 in the receiver circuit) are rendered
irrelevant if their respective termination model touchstone files
incorporate them behind impedance-isolating ideal buffers, also within the
touchstone models. Perhaps requiring the model maker to do so (takes a lot
of head-scratching on the Tx side) is not worth the benefit of eliminating
Tx_R and Rx_R, but I like not caring what these parameters are set to (My
channel simulates the same if they are defaulted to 50 ohms or set to 0.1
ohms.)
But doing so would also eliminate the resistors in the pesky "analog
models". The EDA tool merely drives the Tx algorithmic model out value
(scaled by Tx_V) between port1 and port3 of the tx s4p model and passes the
difference between port2 and port4 in the Rx termination s4p directly to the
Rx model.
Regards,
Bob
On Tue, Jan 10, 2017 at 3:17 PM, Bob Miller (APD) <bob.miller@xxxxxxxxxxxx
<mailto:bob.miller@xxxxxxxxxxxx> > wrote:
Comrades-
[ am somewhat late to this renewed discussion and if there is another active
discussion thread point me to it/include me in it.]
From a model maker/support standpoint, I have long had difficulty
maintaining a robust model manufacturing workflow which keeps the .ibs file
sync'ed with the AMI model. When first developed, Tstonefile (e.g. the
AMI-centric approach) allowed me to produce an analog model as precise as I
wished while also pulling all of the critical modelling details into one
place.
Our .ibs files tend to pair Tx and Rx, which complicates the forward
propagation of changes to either's analog model. (Yes there are ways to
accommodate this but it's Yet Another Thing To do.) Having the analog model
definition concisely wedded into the .ami file (which resides within the ami
model's directory, revision control, etc. was simply too convenient to
abandon, especially after being used to build a small army of models we
continue to support customers on. So I have contributed to this
"bifurcation" in analog model approach.
I personally think its time to just "punt" and accept both methods. I
understand the reasons why the termination models should rightfully be
specified in the .ibs file, but the truth is it's easier to specify them in
the AMI file and the pain of changing is great enough that I'd have to be
pretty much forced to change (say by a customer demand to run their models
on a non-Tstonefile EDA tool) to make the change. And the longer we let
Tstonefile, whose users are probably NOT going away, be "unsupported" we
risk drifting even further apart (If I as a model maker can't use your tool
I will care less deeply whether your tool supports some other cool new
feature).
Anecdote: Avago recently merged with Broadcom and our modeling groups were
delighted to find out we both use Tstonefile.
Completely non-technical argument, I know!! Let's just patch up the status
quo and agree to hang together better in the future.
Regards,
Bob
On Tue, Jan 3, 2017 at 2:11 PM, Walter Katz <wkatz@xxxxxxxxxx
<mailto:wkatz@xxxxxxxxxx> > wrote:
All,
The enclosed zip file contains an IBIS file, .ami files and .mod files that
implements the proposed BIRD 158 with [External Model] syntax that allows an
EDA tool to use the AMI parameters directly to generate Impulse Response in
the frequency domain and allow EDA tools that do not support BIRD 158
parameters to use the [External Model] syntax. BIRD 158.4 is also enclosed.
As written, BIRD 158.4 does not require that the IBIS file has the [External
Model] statements that are shown in the .ibs file example.
Also, the Receiver and the Transmitter Circuit should have the earth ground
symbol replaced with the local ground symbol.
Walter
Walter Katz
wkatz@xxxxxxxxxx <mailto:wkatz@xxxxxxxxxx>
Phone 303.449-2308 <tel:(303)%20449-2308>
Mobile 303.335-6156 <tel:(303)%20335-6156>
