Minutes from the 18 August ibis-atm meeting are attached.
IBIS Macromodel Task Group
Meeting date: 18 August 2020
Members (asterisk for those attending):
Achronix Semiconductor: Hansel Dsilva
ANSYS: Curtis Clark
* Wei-hsing Huang
Cadence Design Systems: * Ambrish Varma
Ken Willis
* Jared James
Google: * Zhiping Yang
Intel: * Michael Mirmak
Keysight Technologies: Fangyi Rao
Radek Biernacki
Ming Yan
Todd Bermensolo
Stephen Slater
Marvell: Steve Parker
Mentor, A Siemens Business: * Arpad Muranyi
Micron Technology: * Randy Wolff
* Justin Butterfield
SiSoft (Mathworks): * Walter Katz
Mike LaBonte
Teraspeed Labs: * Bob Ross
Zuken USA: Lance Wang
The meeting was led by Arpad Muranyi. Justin Butterfield took the minutes.
--------------------------------------------------------------------------------
Opens:
- Randy noted there is a tentative agenda for the IBIS summit next Friday.
-------------
Review of ARs:
- Walter to send the updated presentation on DDR5 training protocol.
- Done.
- Ambrish to find out whether his team would have any interest in taking some
of their BCI protocols public if there were a process to do so.
- Ambrish reported this will require more internal discussion on his side.
--------------------------
Call for patent disclosure:
- None.
-------------------------
Review of Meeting Minutes:
Arpad asked for any comments or corrections to the minutes of the August 11
meeting. Walter moved to approve the minutes. Randy seconded the motion.
There were no objections.
-------------
New Discussion:
DDR5 DQ Write BCI Protocol:
Walter commented there was a concern about the performance of the GetWave
simulation, since the protocol uses file IO to save the waveform. Walter did
a simple test to both read and write the waveform to a file, and the
performance is 10x slower in the training simulations. There is potential for
the training in GetWave to be slowed down significantly by this.
Ambrish asked why we need to pass the waveform. Walter replied, for DDR5, the
Tx sets the FFE and Rx DFE taps. This is how it is done in the hardware,
where the Tx counts the errors at the Rx. Ambrish noted, in SerDes, the
waveforms are not passed back and forth. Walter replied, in SerDes, the Rx
does the DFE adaptation to set the taps. In the case of DDR5 writes, the Rx
does not do the adaptation. The memory can only send back the sequence of
bits it received. The easiest thing for the Tx to process in GetWave is to
accumulate and create a persistent eye. It is easy to pass the waveform from
the Rx to the Tx. The alternative is to have the Rx return some metrics. In
the hardware, the DQS will be skewed against the DQ and the VREF level swept.
The Tx will then look for errors. The question is what metric is sufficient
for the Rx to return, and we have to specify in the protocol how to make the
measurement for this metric.
Arpad asked about the performance and what this is compared to. Walter
replied he compared to a normal GetWave simulation. Arpad asked if we need to
define the format if it is ASCII or binary. Walter replied he has used a
string similar to the AMI parameter file format. Arpad asked if binary could
improve the performance. Walter replied this could possibly improve
performance by a factor of 2.
Walter commented if you are training in GetWave the time to train will
dependent on the number of GetWave calls. The alternative would be to create
an inner eye contour.
Arpad asked if there is any plan to write the protocol in a word document.
Walter stated he plans to do this, but he would like to get more feedback from
other vendors.
Supporting PI modeling/simulation in IBIS:
Zhiping did not have any specific updates on this. He encouraged everyone to
attend the IBIS Summit presentations on PI modeling.
SPI IBIS Summit Agenda:
Randy noted the SIPI IBIS Summit is on 8/28. He reviewed the tentative agenda.
Randy commented it is a full technical program over the scheduled 4 hours. If
you are planning to attend, please contact Lance Wang. Wei-hsing asked if it
is free. Randy replied the EMC meeting is not free, but the IBIS Summit is
free. You can also sign up with IEEE to attend the IBIS Summit. Bob noted
the agenda will go out Friday 8/21. Walter asked what time it starts. Randy
replied it starts at 1pm CST. Walter asked about the IBIS Open Forum
meeting on this day. Randy replied it has been canceled.
- Michael: Motion to adjourn.
- Walter: Second.
- Arpad: Thank you all for joining.
-------------
Next meeting: 25 August 2020 12:00pm PT
-------------
IBIS Interconnect SPICE Wish List:
1) Simulator directives
