[ibis-macro] Re: How would you model this On Die Termination? (continuation from Question on dividing up the Tx behavior thread)
- From: James Zhou <james.zhou@xxxxxxxxxx>
- To: Scott McMorrow <scott@xxxxxxxxxxxxx>
- Date: Fri, 9 Mar 2012 15:27:53 -0800
Hi Scott,
Thanks for your response, please see my comments below:
(1) "Your fundamental assumption may be incorrect": which "fundamental
assumption" are you referring to? Please be specific. I could not see any
assumptions in my email.
(2) "Tx driver is a high impedance current source": This could be true for
this example. However, the question needs to be answered is: does or should
IBIS AMI flow also work for other impedance levels?
(3) "that is isolated from the channel by a terminated T-coil.": This is
incorrect. The T-coil is passive RLC as shown in the diagram and is completely
reciprocal forward and backward. If it has isolation backward, it must also
have the same isolation forward, which will prevent it from working at all. The
fact that Tx driver is high impedance has little to do with whether or not "it
is isolated from the channel".
(4) "The node that the driver is attached should be "by-design" a high
impedance node.": Similar to (2) above, the question is whether IBIS AMI is
"by-design" to only work for high-impedance Tx drivers? Or should the flow work
well regardless of the impedance levels of Tx driver?
Regards,
James Zhou
From: Scott McMorrow [mailto:scott@xxxxxxxxxxxxx]
Sent: Friday, March 09, 2012 2:28 PM
To: James Zhou
Cc: Terry.Chen@xxxxxxxxxx; IBIS-ATM
Subject: Re: [ibis-macro] Re: How would you model this On Die Termination?
(continuation from Question on dividing up the Tx behavior thread)
James
Your fundamental assumption may be incorrect. The Tx driver is a high
impedance current source that is isolated from the channel by a terminated
T-coil. The node that the driver is attached should be "by-design" a high
impedance node. What Terry does not show in his circuit drawing is that the
two inductors on either side of the termination are coupled spiral inductors.
Regards,
Scott
On Fri, Mar 9, 2012 at 3:18 PM, James Zhou
<james.zhou@xxxxxxxxxx<mailto:james.zhou@xxxxxxxxxx>> wrote:
Hi Terry,
I am not the expert but would like to offer my "0.02" on this topic.
If we name the differential Tx driver current summing nodes as <TDO+,TDO->, it
is obvious that the voltages at TDO are dependent on the loading at Tx PKG
output (i.e. input impedance of the channel). This is a very basic concept in
circuit design and there is nothing new about it.
Now if we can agree that the waveforms at TDO are dependent on the loading at
Tx PKG output, then there are several subsequent questions one must answer when
creating TDO waveforms:
(1) what loading can (or should) be put at Tx pkg output when generating TDO
waveforms?
(2) is it necessary and if so, how to inform the users and EDA tools about the
loading impedance?
(3) under what conditions can we assume that the impact to TDO waveform caused
by channel loading at Tx PKG output is negligible and, is it actually the case
in real silicon/package design? What are the errors caused by this assumption?
Existing IBIS ATM BIRDs propose to enforce isolation between Tx AMI output
(TDO) and channel loading. Regardless of the implementation details, the
"isolation" approach forcefully make TDO waveform "independent" of the channel
loading, which is equivalent to say that the impact of channel loading to TDO
waveform is negligible.
There are established methods to model this circuit rigorously in the most
general case (when loading is not negligible). That would require the knowledge
of the output impedance of Tx Driver. This is the impedance looking from the
location of "current summing" arrow towards the left. I have not seen such
proposals at IBIS ATM.
Based on IBIS 5.0 and various existing BIRDs for 5.1 and 5.2, I think this
circuit can be modeled by the following approach:
(1) choose TDO nodes as Tx AMI output and, Tx analog input
(2) obtain Tx AMI waveform at TDO nodes by loading the Tx PKG with 50ohms or
any other impedance of choice. The loading impedance used in obtaining TDO
waveform should be recorded in ibis file.
(3) provide Tx Term circuit in IBIS [External Circuit] or [External Model}
keywords using one of the supported languages, such as Spice, IBIS-ISS or
Touchstone.
This flow allows the model creator to provide accurate waveform at Tx AMI
output and, true-to-silicon Tx analog circuitry without any unnecessary
restrictions and assumptions. It would require minor changes to the IBIS Spec
by forcing the disclosure of load impedance used in TDO waveform simulations.
Alternatively, you may put the T-Coil circuit and the caps inside the Tx Driver
block and have two short wires (i.e. all pass filter) linking Tx driver to Tx
pkg. This is the approach David and you have mentioned in earlier emails.
Strictly speaking there is nothing wrong to use such structures in circuit or
channel simulations. However this may not be the best way to fit into the
existing IBIS AMI flow for reasons stated above.
Best Regards,
James Zhou
QLogic Corp.
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>]
On Behalf Of Chen, Terry
Sent: Friday, March 09, 2012 6:28 AM
To: IBIS-ATM
Subject: [ibis-macro] How would you model this On Die Termination?
(continuation from Question on dividing up the Tx behavior thread)
Hi IBIS experts,
I have been getting a ton of great feedbacks on why it may be a bad idea to
model the TX driver with as an ideal output step function (b/c it will not
model the impedance mismatch and reflections properly). To clarify, I have
roughly sketched my TX output driver and the T-Coil "like" ODT structure, and
attached it as a picture in order to serve as a point for further discussion. I
have also sketched how I am "attempting" to model it.
So the question I have is:
1. How would you model this?
2. Can you do this using original IBIS 5.0 directives and still get the
frequency dependent effect of my ODT?
I had (possibly mistakenly) thought that this is not possible with IBIS 5.0, so
I proceeded down the path of factoring out the term and the parasitics as S4P.
Regards,
Terry Chen
________________________________
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.
--
Scott McMorrow
Teraspeed Consulting Group LLC
121 North River Drive
Narragansett, RI 02882
(401) 284-1827 Business
(401) 284-1840 Fax
http://www.teraspeed.com<http://www.teraspeed.com/>
Teraspeed(r) is the registered service mark of
Teraspeed Consulting Group LLC
________________________________
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.
Other related posts:
