RE: LVDS models and s2ibis2

Hi,

As The presentation shows, we classify LVDS buffer into two categories:

        - Without internal termination: (If we are lucky)
                - This one I consider it just a normal differential buffer.
No special setting for it, just the normal differential buffers settings.

        - With internal termination: (The usual received sub circuit)
                - If the internal termination (the resistance shunted
between the +ve & -ve PADS) has a discrete value (and a linear one) then you
can use the approach of decomposing the resistance behavior as shown in
Munich summit meeting presentation.

                - If the termination is not linear or you can't specify its
value then you can use the variable resistance approach in the presentation
you already have.

Regarding your questions:

>>>> 1. If differential inputs or outputs
        switch between 1.1 and 1.3 volts, what
        range of voltages do you use for the I/V
        curves?

The IBIS standard recommends the I-V curves extraction from -VDD to 2*VDD
But if you have difficulties in that you can sweep in the active region only
(from 0 to VDD) in this case you should expect bad validation out of the DC
range (if you have some high reflection above/under VDD/zero respectively).

>>>> 2. What load do you use on differential
outputs to generate the V/T curves? (What
are Rfix, Rref, Vref, ... and how are these
plus the terminating resistor connected?)
By the way, this has to be done in a way
that HPSICE converges.

The loads are normally 50 Ohm to the reference voltage. Apparently the
reference voltage in your case is 1.2 volt as the wave form swings between
1.1 & 1.3 (i.e. around 1.2) .
Rfix is the load you but while extracting the V-T tables, Rref&Vref are the
value of the timing test loads stated in the data sheets of that buffer.

BR,

Hazem Hegazy
Development Engineer

Tel: (20 2) 4135458
Fax: (20 2) 4186945
Mobile: (+20) 105129341
Email: hazem_hegazy@mentor.com

-----Original Message-----
From: Greg Haynes [mailto:haynes@utmc.aeroflex.com]
Sent: Saturday, May 26, 2001 1:23 AM
To: ibis-users@eda.org
Subject: LVDS models and s2ibis2

I'm trying to create IBIS models for some
parts that we make using s2ibis2, and for
single-output buffers it works pretty well.
LVDS buffers are another story. In some
cases I've had a little success, but I
don't know if s2ibis2 is even supposed to
be capable of creating the required SPICE
decks for simulating LVDS buffers.

From a previous post I was referred to
a presentation on LVDS modeling by
Hegazy and Korany, but it's just foils
with no detailed explanations.

Among my questions:

1. If differential inputs or outputs
switch between 1.1 and 1.3 volts, what
range of voltages do you use for the I/V
curves?
2. What load do you use on differential
outputs to generate the V/T curves? (What
are Rfix, Rref, Vref, ... and how are these
plus the terminating resistor connected?)
By the way, this has to be done in a way
that HPSICE converges.

-- 
Greg Haynes		  haynes@utmc.aeroflex.com
Aeroflex UTMC Microelectronic Systems
4350 Centennial Blvd           phone: 719 594-8197
Colorado Springs, CO 80907       fax: 719 594-5541