[ibis-macro] Re: another silly IBIS-AMI question about deriving the impulse response of a channel

  • From: "Walter Katz" <wkatz@xxxxxxxxxx>
  • To: <scott@xxxxxxxxxxxxx>, "IBIS-ATM" <ibis-macro@xxxxxxxxxxxxx>
  • Date: Thu, 27 May 2010 18:14:07 -0400


Your picture is incorrect:

1 --- Network --- 3

It is really

Tx_Pad --- Network --- Rx_Pad --- Input_to_Equalization ---

There is a high impedance connection between ports Rx_Pad and

This should answer your question.


Walter Katz
Mobile 720.333-1107

-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx]On Behalf Of Scott McMorrow
Sent: Thursday, May 27, 2010 6:04 PM
Subject: [ibis-macro] another silly IBIS-AMI question about deriving the
impulse response of a channel

Since the analog channel is required to be LTI from driver to receiver, I
believe that in order to create a correct impulse response it is necessary
to know what the termination impedance should be.  In the case where a
driver or a receiver is terminated at the end of the line, the termination
impedance is obvious.  So, for a 100 ohm differential system the
differential impedance used for impulse response extraction is 100 ohms.
Essentially the high-impedance receiver amplifier, or driver current sources
are considered to be lumped at the termination network.

How are people modeling, extracting, and simulating drivers and receivers
where the high-impedance ports are physically isolated from the termination,
where the waveform in the algorithmic domain (driver/receiver) is
substantially different than the waveform in the analog domain at the
termination?  I'm confused, since IBIS-AMI does not include the on-die
termination network, except in the rudimentary and unsatisfying form of a
*.ibs model, and that model is a lumped approximation.  In my mind, in order
to correctly extract an impulse response of a channel that will give a
correct representation of the waveform at the receiver, one has to either
perform a transient simulation of the full circuit from the driver to the
receiver, or perform frequency domain modeling to impulse response
transformations with known driver/receiver input impedance.  Otherwise, the
resulting impulse response is wrong.

For a hypothetical 50 ohm single-ended transmitter or receiver, the simplest
circuit that can correctly model many transmitters and receivers is the
following network:

1 --- Network --- 3

            Port 1 = 50 ohm
            Port 3 = 50 ohm
            Port 2 = some high impedance.
            and Port 1 to Port 2 is the path used for computation of the
impulse response

If the Network is lumped into:

1 --- Network --- 2

The resulting impulse response and waveform is different.  How are people
solving this problem?



Scott McMorrow
Teraspeed Consulting Group LLC
121 North River Drive
Narragansett, RI 02882
(401) 284-1827 Business
(401) 284-1840 Fax


Teraspeed® is the registered service mark of
Teraspeed Consulting Group LLC

Other related posts: