Re: BIRD18 Column Headers for IBIS Ver. 2.1

Arpad

Here are my views on the questions you raise:

1) BIRD18 applies only to [Diff Pin] and is totally unrelated to BIRD17
While cutting and pasting, I accidentally tranmitted a BIRD18 which
had some BIRD17 verbage, and have subsequently retranmitted BIRD18.
If you need the correct copy, I can send it to you.

2) Regarding the range of any VI table in IBIS, I interpret the mandated
ranges as the Minimum ranges for IBIS compliant models. I believe there
is nothing to prevent the [Power_clamp] table to be extended on either
end. So IBIS does not prevent you from putting in a "weak", non-linear
always on pullup resistor using the [Power_clamp] keyword for any range
including to 0 or -Vcc. The [Gnd_clamp] data from 0 to Vcc (and beyond)
would be set to 0 since the clamping curve currents are added.

3) For normal IBIS buffer applications, you would not need to extend
the clamp ranges beyond what is specified (beyond Vcc for [Gnd_clamp] and
below Vcc for [Power_clamp] because the clamping currents in these regions
are typically negligible. So even if simulators do not do extroplation,
the error would be neglibible. Your pullup resistor is an example where
you need to extend the IBIS range because the currents are not negligible,
and I believe IBIS already supports this application.

Bob Ross,
Interconnectix, Inc.

> So is BIRD 17 or 18 the same thing?

> One more comment I forgot in my last EMAIL:

> If anything, I would like to see both clamps have the same range, -Vcc to Vcc,
> or even -Vcc to (2 x Vcc). Currently, the POWER clamp is only -Vcc to 0V, and
> the GND clamp is -Vcc to Vcc.

> We all agreed on earlier open forums that the clamp curves can also be used for
> internal pullup or pulldown structures ("resistors") for devices that have it.
> These "resistors" are usually weak transistors which are always ON. Since they
> are transistors, they have an I-V curve, not a straight resistive line. The
> best place to put these in an IBIS model is the clamp curve.

> Hovever, currently there is only one place to put such curves, the GND_clamp
> curve. This creates a few problems when I have a pullup "resistor" (which I
> have to make VCC relative) and I have to put it into the GND_clamp curve. If we
> extend the Vcc_clamp range to the same coverage as the GND_clamp, I could put
> the pullup "resistors" into the POWER_clamp, and the pulldown "resistors" into
> the GND_clamp curve and my problems would be gone.

> Arpad Muranyi
> Intel Corporation
Received on Mon Jul 25 14:23:28 1994