Bird2.1, Useful timing data

To IBIS Committee

I basically agree with Jon Powells comments regarding having the Input
Threshholds, and feel that Vinh and Vinl should be provided in all Input
and I/O type models.

I lean toward requiring this data if there is another "Model_type" such as
"Analogue" or "Terminator" or whatever that would allow a device to be on
the net for which no timing information is needed or desired. One application
that is within IBIS Version 1.1 functionality is to add a clamping diode
terminator package at the end of a net which has no need for threshhold
timing calculations.

I am assuming that the Vinl and Vinh values are in practice, Data Book
specification limit values. However, they also could represent an
attempt to represent actual values which may be within a smaller range.
Furthermore, they could also represent values for a Hysterisis Input
where the Vinl value is greater than the Vinh value.

One underlying assumption is that any simulator that needs threshhold
data can enter that data for each model and/or override any data that is
presented in the model. Because of this, the issue is not critical. The
workaround is to enter the value externally and/or use some default values.

However, a consideration is that the Model provider will probably know this
information and therefore should provide it. Some devices (larger ones)
may have several types of input and outputs and may have several different
threshhold levels specified. TTL/ECL translator chips will have different
levels. It is easier for values to be provided than to have the user tediously
go through literally pin-by-pin, especially when there are significant
differences.

Concerning "VM", I presume both a voltage and a correction time would be
provided. I think that at least a VM voltage would be a good addition since
that is also provided in Data Books and is relevant to timing analysis
measurement points. The whole timing issue is a "sticky wicket", so I can
understand the safer position of doing nothing at this time. However, I
do believe that these timing "enhancements" provide measurement parameters
closely related to "SI analysis" and in many cases timing is the PASS/FAIL
crieria of SI on nets. So I strongly recommend all model providers include
at least the Vinh and Vinl information.

Bob Ross,
Interconnectix, Inc.
Received on Wed Mar 16 20:27:13 1994