Re: Ground/Power Noise modeling of sim switch outputs

From: Bob Ross <bob@icx.com>
Date: Tue Oct 24 1995 - 15:36:00 PDT

Michael:

I tend to agree with your assessment that knowing I_pwr_n(t) and I_gnd_n(t)
obtained under the test load condition will give a more accurate
Simutaneous Switching Output analysis. There may be some feedback
interactions when switching several outputs which may cause the total
power and ground currents to be different than the sum of the components
found individually. However, this approach should still be reasonably
accurate under the test load conditions.

Presently, IBIS can be used to estimate the currents based on assuming
the Pullup table provides currents from the Power Bus, and the
Pulldown table provides currents from the Ground Bus. The currents
that are provided by each BUS into the test loads during the
low-to-high and high-to-low transitions depend on the simulator
switching or blending algorithm with respect to the Pullup and Pulldown
tables. So this approach can only be approximate. The IBIS format does
not give any more details about internal dynanic impedances which would
effect the current distributions.

If providing I_pwr_n(t) and I_gnd_n(t) tables gives sufficient information,
the IBIS format could be easily extended using syntax similar to [Rising
Waveform] and [Falling Waveform] to provide this data. However, I am
wondering how this information can be accurately applied in real loading
situations (not just the test load case) to simulate actual power and ground
bounce noise. Also, what are the best test loads?

Bob Ross,
Interconnectix, Inc.

> Hello, IBISians !

> (Sorry, but I've forgotten to name a subject for posting my
> mail. Therefore I sent this message once again. The content is
> the same as in my first mail.)

> I want to use IBIS for modeling of Ground/Power Noise of simultaneously
> switching drivers.
> Spec version 2.1 can handle static V/I curves and Vout(t) curves
> for a given load. If I've got this information then I am able to
> determine the rising/falling waveform of the output current Iout(t)
> of a switching driver.

> What I need for example to calculate the voltage drop over the
> ground inductance L_gnd_pin (see below) is the contribution of every
> (simultaneously) switching driver to the current I_gnd(t) flowing
> through their common ground pin inductance (if I assume for
> simplification that the inductance of the ground path is caused only
> by that ground pin).

> o PWR
> |
> |
> C
> C L_pwr_pin
> C
> |
> |
> V I_pwr(t)
> I_pwr_1(t) | I_pwr_n(t)
> |
> -----<--------------*------------- - - ----->---
> | | |
> | V I_pwr_2(t) |
> | | |
> ---------- ---------- ----------
> | | Iout_1(t)| | Iout_2(t) | | Iout_n(t)
> |Driver_1| |Driver_2| |Driver_n|
> | |--->-| | |--->--| | |--->--|
> | | | | | | | | |
> ---------- --- ---------- --- ---------- ---
> | --- CL_1 | --- CL_2 | CL_n ---
> | | | | | |
> | | | | | |
> | --- | --- | ---
> | | |
> V I_gnd_1(t) V I_gnd_2(t) V I_gnd_n(t)
> | | |
> | | |
> --------------------*-------------- - - --------
> |
> |
> V I_gnd(t)
> |
> C
> C L_gnd_pin
> C
> |
> |
> ---

> It would be useful for Ground/Power Noise calculations to have the
> currents of the internal ground/power path of every driver
> I_gnd_1(t)/I_pwr_1(t) ... I_gnd_n(t)/I_pwr_n(t) with respect to
> the corresponding Vout(t) curves for specified loads.

> Unfortunately it is not possible to describe these ground/power path
> currents with current IBIS format in my opinion. I'm quite aware
> of the fact that this request perhaps contradicts the basic
> concept of IBIS because these internal ground/power path currents are
> not measurable from outside of the package.

> But at my point of view I don't see any other way for an
> efficient ground (power) noise analysis.

> However, what do you think about the opportunities to model
> power/ground noise and the resulting switching noise at the
> outputs of active/quiescent output buffers by using present
> or future IBIS format ?

> Any comments to that problem will be appreciated.

> Thank you very much in advance

> Michael Gutzmann

> Cadlab
> Group Analog System Engineering
> Joint R&D Institute
> University of Paderborn
> Siemens Nixdorf Informationssysteme AG
> Fuerstenallee 7
> 33102 Paderborn
> Germany

> phone: +49-5251-60-6164
> fax : +49-5251-60-6155
> e-mail: micha@cadlab.de
Received on Tue Oct 24 16:02:30 1995

This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:52:29 PDT