Thanks for the great discussion! To clarify a few points:
1. [Test Load] and [Test Data] do not today specify the targets for or the
method for evaluating the model quality. Today, for analog models, nothing is
provided outside of the names of the models involved, the load structure, and
the output waveforms expected. No criteria are supported for how "close"
matching between the user's simulation and the provided output should be (see
below).
2. I had envisioned that AMI_Parameters_In would be generated by the
simulation tool using the defaults from the .ami file associated with the
[Algorithmic Model] keyword. To go beyond the defaults, we would need to add
support for some sort of configuration file that looks like a .ami parameter
tree, or otherwise supporting parameter-passing as is done elsewhere in IBIS.
3. My original assumption was that EDA tool vendors would enable automated
tests using this new AMI [Test Data] keyword information. I had not assumed
that quality would be automatically evaluated, but that users would manually
perform their own comparisons using the data generated by the tools.
4. As Walter implies, it may be worthwhile to include support for expected
AMI_Parameters_Out from a successful run.
5. If we wanted to include quality metrics, we would have to include a lot
of additional detail, including potentially adding significant new features
(e.g., FSV for waveform matching, or the long-desired eye mask and BER
definition sub-parameters). As Walter suggests, AMI doesn't generally deal in
eyes, because AMI focuses on waveforms, not on eye generation. Adding any
references to eyes would end up enabling tool-to-tool comparisons in addition
to model checks.
6. Lance is absolutely right that, for AMI_GetWave, some number of bits
should be (optionally?) recommended by the model maker for simulation, and that
these should be above the IgnoreBits minimum of the models in the [Test
Load]/[Test Data] setup. I can add that optional number of bits to the list of
sub-parameters for the keyword.
7. We could definitely add support for external files containing the
waveform data, but as Mike stated, we would be setting ourselves up to support
external files for just about any waveforms supported in-line in IBIS files
today. I'm willing to take the risk...
* MM
From: Walter Katz <wkatz@xxxxxxxxxxxxx>
Sent: Wednesday, January 19, 2022 12:04 PM
To: Wang, Lance <Lance.Wang@xxxxxxxxxxxx>; Mirmak, Michael
<michael.mirmak@xxxxxxxxx>
Cc: ibis-macro@xxxxxxxxxxxxx
Subject: RE: Slides, Thoughts on AMI Support for [Test Data]/[Test Load]
Lance,
The output should contain the AMI_Parameters_out. Again, it is the EDA tool
that "ignore bits". For IBISCHK the "input file" needs a single waveform. The
IBSCHK verification code would need to split up the input waveform into blocks
and put output blocks back again. How long, the Model Maker is giving you the
waveform. How to define the stimulus, the Model Maker with give you an input
pattern.
Walter
Walter Katz
Work 508.647-7633
Cell 720.417-3762
[Description: Description: Visit MathWorks.com]
From: Wang, Lance <Lance.Wang@xxxxxxxxxxxx<mailto:Lance.Wang@xxxxxxxxxxxx>>
Sent: Wednesday, January 19, 2022 2:50 PM
To: Walter Katz <wkatz@xxxxxxxxxxxxx<mailto:wkatz@xxxxxxxxxxxxx>>;
michael.mirmak@xxxxxxxxx<mailto:michael.mirmak@xxxxxxxxx>
Cc: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: RE: Slides, Thoughts on AMI Support for [Test Data]/[Test Load]
Hi Walter,
Thanks, that makes sense.
I have another question. Will the Impulse response from AMI_Init and/or the
Waveform output from AMI_GetWave be sufficient? Using the waveform output from
AMI_GetWave as an example, do we need a certain length of bits to represent the
AMI_GetWave functionality fully? How long? How to define?
Thanks,
Lance Wang
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
<ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>> On
Behalf Of Walter Katz
Sent: Wednesday, January 19, 2022 12:15 PM
To: Wang, Lance <Lance.Wang@xxxxxxxxxxxx<mailto:Lance.Wang@xxxxxxxxxxxx>>;
michael.mirmak@xxxxxxxxx<mailto:michael.mirmak@xxxxxxxxx>
Cc: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: Slides, Thoughts on AMI Support for [Test Data]/[Test
Load]
EXTERNAL
Lance,
Re:
Should we also consider adding the parameter settings (if differ than the
defaults in .ami file), Eye Opening/BER for the defined fixture/channel?
"Eye Opening/BER" are not outputs of the channel, they are outputs of the EDA
tool processing the Impulse Response output of AMI_Init or the waveform output
of AMI_GetWave. IBISCHK has no way of verifying this. There is no specification
on how "Eye Opening/BER" is calculated. If the Model Maker sets the "Eye
Opening/BER" generated when running the model with this test data on EDA Tool
A, there is no expectation that the user will get the same answer on EDA Tool
B. Who is to decide which EDA Tool is "Right".
Walter
Walter Katz
Work 508.647-7633
Cell 720.417-3762
[Description: Description: Visit MathWorks.com]
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
<ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>> On
Behalf Of Wang, Lance
Sent: Wednesday, January 19, 2022 12:03 PM
To: michael.mirmak@xxxxxxxxx<mailto:michael.mirmak@xxxxxxxxx>
Cc: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: Slides, Thoughts on AMI Support for [Test Data]/[Test
Load]
Hi Michael,
This is indeed an interesting idea, and it will be a great improvement related
to IBIS quality (for AMI models), I think.
In my experience, AMI users often face the different results given by different
simulators. This new feature would at least give a model vendor's expectation
for the simulations using given test fixture/channel.
Should we also consider adding the parameter settings (if differ than the
defaults in .ami file), Eye Opening/BER for the defined fixture/channel?
Also, there are some "complains" about the IBIS file size (too large due to
[Test data]). We could use this opportunity to redefine waveform attachment for
the [Test data].
My 2 cents,
Best regards,
Lance Wang
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
<ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>> On
Behalf Of Mirmak, Michael
Sent: Tuesday, January 18, 2022 5:27 PM
To: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Slides, Thoughts on AMI Support for [Test Data]/[Test
Load]
EXTERNAL
Please find enclosed my slides from today's IBIS-ATM meeting, regarding
expanding the existing [Test Data] and [Test Load] keywords to support AMI
model testing. I greatly appreciate the thoughtful feedback and suggestions,
and will incorporate them into any eventual BIRD.
Please note that the material is solely my own and does not necessarily
represent the views of my employer or any other person or organization.
- MM
---------------------------------------------------------------------
IBIS Macro website :
http://ibis.org/macromodel_wip/<https://nam04.safelinks.protection.outlook.com/?url=http%3A%2F%2Fibis.org%2Fmacromodel_wip&data=04%7C01%7Clance.wang%40zukenusa.com%7C07ba98263ff247be394c08d9db6f3214%7Cb0eaca87236343d2abd9213a0c7b484d%7C0%7C1%7C637782092910629148%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=DwalFDuHEeK9OYn%2BtVoUZkvx%2BowECIOMaaSa91aNrKE%3D&reserved=0>
IBIS Macro reflector:
//www.freelists.org/list/ibis-macro<https://nam04.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.freelists.org%2Flist%2Fibis-macro&data=04%7C01%7Clance.wang%40zukenusa.com%7C07ba98263ff247be394c08d9db6f3214%7Cb0eaca87236343d2abd9213a0c7b484d%7C0%7C1%7C637782092910629148%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=MM3XwhGI5L%2F4xu2O%2BdJ%2BEBXMOWWSo4vwdzkbCsSQbH0%3D&reserved=0>
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx<mailto:ibis-macro-request@xxxxxxxxxxxxx>
Subject: unsubscribe