[ibis-macro] Re: AMI Touchstone Analog Model BIRD

  • From: Taranjit Kukal <kukal@xxxxxxxxxxx>
  • To: "wkatz@xxxxxxxxxx" <wkatz@xxxxxxxxxx>, IBIS-ATM <ibis-macro@xxxxxxxxxxxxx>
  • Date: Sat, 16 Feb 2013 11:34:41 +0530

Hi Walter,
Why are we putting s-parameter file (describing analog IO-buffer) into AMI 
file. Why can't we handle it in EXTERNAL_CIRCUIT/MODEL? This way, we keep it 
clean where IBIS file has the IO info (analog portion) and AMI deals with just 
the algorithmic portion.

Hope I did not misunderstand the BIRD intent...


From: ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] 
On Behalf Of Walter Katz
Sent: Saturday, February 16, 2013 2:40 AM
Subject: [ibis-macro] AMI Touchstone Analog Model BIRD


I am enclosing the BIRD that defines AMI Parameters to support Touchstone 
Analog Buffer Models.

I reviewed last week's meeting minutes. There was a request that I indicate 
whether the specification should define how the impulse response should be 
generated for a channel. The generation of the impulse response of a channel is 
well defined when the Tx and Rx buffers specify a Touchstone file in accordance 
with the enclosed BIRDs.  Since the channel is LTI and these analog buffers are 
LTI there are many methods that can be used to generate the impulse response, 
and one can prove these methods should give identical results within numerical 
accuracy. The first method is to do a step response SPICE simulation using the 
differential voltage range specigied by Tx_R. The impulse response is the first 
derivative of the differential signal at the Rx Touchstone nodes specified on 
this BIRD. A second way is to convert the channel into an s4p Touchstone field. 
Using the S Parameter arithmetic James presented recently, all one needs to do 
is cascade the Tx s4p, channel s4p and Rx s4p into a single s4p, inverse 
Fourier transform this s4p, and voila, one has the impulse response of the 

On can similarly describe various methods to generate the impulse response of a 
channel when  the Tx and Rx are legacy models with linear IV curves. This is 
true because one can generate an s4p analog model for these Tx and Rx legacy 

Defining a method to generate the impulse response of a channel when the Tx 
and/or Rx have non-linear IV curves is problematic since the premise of AMI 
modeling implies Linear Time Invariant (LTI). Which is another way of saying a 
"Small Signal Interface". Since most implementations of the IBIS B element 
require a digital stimulus, and according to the common understanding that IBIS 
B elements (and external model) have digital stimulus, then an impulse response 
generate with a SPICE simulator using the IBIS B element can only generate a 
step response across the full swing of the Tx.


Walter Katz
Phone 303.449-2308
Mobile 303.335-6156

Other related posts: