Fwd: Re[4]: Non-Monotonic Drivers

From: MR RONALD J CHRISTOPHER <EGJJ77A@prodigy.com>
Date: Thu Aug 31 1995 - 08:08:16 PDT

-- [ From: Ron Christopher * EMC.Ver #2.10P ] --

Forwarding to the reflector.

Ron Christopher
------- FORWARD, Original message follows -------

> Date: Wednesday, 23-Aug-95 07:26 PM
>
> From: Arpad Muranyi \ Internet:
> (arpad_muranyi@ccm.fm.intel.com)
> To: Ron Christopher \ PRODIGY: (EGJJ77A)
>
> Subject: Re[4]: Non-Monotonic Drivers
>
> Text item:
>
> John,
>
> I see your point in suggesting a spec on how to handle the I-V
curves.
> However, I am afraid that this will never happen. We had similar, or
even
> more important issues surface up in previous Open Forum meetings, yet
> these discussions ended with the conclusion that the software tool
vendor
> should not be told how to do things. That is their specialty, and
that
> is how they differentiate on the market. Each one of them knows
> everything better than the other, so they wouldn't listen anyway. :-)
>
> You, the user must decide which tool suits your needs the best.
>
> Arpad
>
=======================================================================
==
> =====
> Arpad,
>
> I agree 100% with your point #2. In fact non-monoticity was a non-
issue
> with our ASX simulator because we knew ahead of time that these
waveforms
> were non-standard. Our ASX analysis did not match up that well
because the
> Vout vs. Time waveforms were not symmetric about the center voltage,
the
> shape of the waveform assumed by the vendor tool was different from
> actual. IBIS has addressed this in ver 2.1.
>
> Thus we analyzed the VI cureves and did what you said in #2. The
problem
> is ofcourse that the standard does not tell one writing a simulation
> program how to handle these IV curves. Current programs that I looked
at
> changed the data to be monotonic, I beleive so they could calculate
> impedances. I would imagine different programs would do different
things.
> I suspect you have seen other non-monotonic drivers that are non-
monotonic
> for other reasons that may need to be treated different yet. A
challenge
> would be to write the spec to handle all of the known cases.
>
> We are not getting sufficient demand for CMOS 4 IBIS models to
justify any
> lengthy studies on how to model these drivers.
>
> John
>
>
***********************************************************************
> John Brennan Internal: jbrennan@btv
> IBM Microelectronics Internet:
jbrennan@vnet.ibm.com
> G77V/862-1 Tel: 802-769-6982 (Tie-Line:
446)
> Burlington, VT Fax: 802-769-5882 (Tie-Line:
446)
>
***********************************************************************
>
> ------------------ Referenced E-Mail ----------------------------
>
> Date: Wed, 23 Aug 1995 10:46:08 PST
> cc: jbrennan@VNET.IBM.COM, EGJJ77A@prodigy.com, jonp@qdt.com,
> Will_Hobbs@ccm2.jf.intel.com
> From: Arpad Muranyi <Arpad_Muranyi@ccm.fm.intel.com>
> Subject: Re[2]: Non-Monotonic Drivers
> Message-Id: <950823104608_4@ccm.fm.intel.com>
>
> Text item:
>
> John,
>
> Thanks again for your reply. I have mixed feelings about what to say
now.
>
> 1) This kind of a behavior is most likely caused by a feedback
circuit in
> the
> buffer. It is true, IBIS currently does not have provisions to
describe
> buffer
> s
> with feedback. However, we had discussions on this subject in the
open
> forum
> meetings before, and we could pick it up again if there is a real
need for
> modeling such buffers in IBIS.
>
> 2) On the other hand, I am not sure if IBIS is to blame for the
situation
> that
> you described, that simulators alter the curve this and that way. I
> imagined
> how I would use this data in my HSPICE behavioral (=IBIS) models. I
am
> using
> piecewise linear (PWL) sources for the I-V curves. If I make the
> voltage value
> s
> the independent variable, and current the dependent variable, it will
work
> no
> problem, because each voltage value is assigned only one current
value.
> (It
> doesn't matter whether the current values are repeated). If I turned
> it around
> ,
> however, to have current values as the independent variable and
voltage
> the dependent one, it would not work, because the simulator would
have to
> pick between two or more voltages for one current value. What I am
trying
> to say here is that this kind of a curve should work, as far as IBIS
is
> concerned (as it does in my HSPICE behavioral models). It depends on
what
> the simulator tool does with it.
>
> Arpad
>
=======================================================================
==
> ====
> Arpad,
>
> Each point on the curve is what you would get if you put a battery DC
> voltage at the output of the driver and measured the current through
the
> pad with all other operating conditions constant. The curves were
> generated with multiple DC analysis using ASX(the IBM SPICE program).
I do
> not think IBIS can handle this today. Specifically, simulation
programs
> will try to alter the DC curves, and there is nothing that tells them
how
> to do this(I believe the correct thing to do would be to not alter
the
> curve).
>
> Thanks,
> John
>
>
***********************************************************************
> John Brennan Internal: jbrennan@btv
> IBM Microelectronics Internet:
jbrennan@vnet.ibm.com
> G77V/862-1 Tel: 802-769-6982 (Tie-Line:
446)
> Burlington, VT Fax: 802-769-5882 (Tie-Line:
446)
>
***********************************************************************
>
> Text item:
>
> John,
>
> Thanks for the graphical response to my question. This is the kind
> of response
> I really wanted to get. I have a question, though. Is this I-V
curve a
> function of the voltage on the output pin, or time. I mean, is
> "translating
> between two Vgs curves" done with a feedback circuit by monitoring
the
> output
> voltage, or is it a function of time as slew rate controlled devices
> usually do
> it?
>
> If the answer is time, IBIS can definitely do it, but if it is
voltage (as
> in feedback), IBIS would not be able to help you at this time.
However,
> the open forum committee did consider this issue in meetings before,
and I
> don't think that it would be impossible to incorporate features that
can
> model it.
>
> Arpad
>
=======================================================================
==
> =====
>
>
> Arpad,
>
> The drivers Ron refers to have the following IV curve shape...
>
> Ids * *
> * *
> * *
> * *
> * ****************
> *
> *
> *
> * Vds
>
>
> Instead of the normal transistor curve shape like...
>
> Ids * ***********************
> *
> *
> *
> *
> *
> *
> *
> * Vds
>
>
> The top curve comes about because we are manipulating the gate
voltage to
> control di/dt noise in the maximum Vds region of the large output
device.
> Essentially we are translating between two Vgs curves for the output
> transistor. Hence the non-monotonicity is real and not noise related.
> Clamp diodes would add another bend in the curve and create 3 current
> values for 1 voltage value.
>
> The simulation tools we looked at would "correct" the data by making
the
> top curve look like the bottom. We saw some extreme differences in
> benchmark nets after this manipulation from the full transistor
> simulation. Since there is no standard on how to "correct" the above
> curve, we would not know how various simulation vendors would treat
the
> model.
>
> The above and the time schedule of IBIS 2.1 and CMOS 4 makes
supporting
> the IBIS modeling strategy unattractive for our CMOS 4 products.
>
> Thanks,
> John Brennan
>
>
***********************************************************************
> John Brennan Internal: jbrennan@btv
> IBM Microelectronics Internet:
jbrennan@vnet.ibm.com
> G77V/862-1 Tel: 802-769-6982 (Tie-Line:
446)
> Burlington, VT Fax: 802-769-5882 (Tie-Line:
446)
>
***********************************************************************
>
> Text item: External Message Header
>
> The following mail header is for administrative use
> and may be ignored unless there are problems.
>
> ***IF THERE ARE PROBLEMS SAVE THESE HEADERS***.
>
> From: "John Brennan" <jbrennan@VNET.IBM.COM>
> Date: Tue, 22 Aug 1995 13:34:03 -0400
> Subject: Non-Monotonic Drivers
> To: EGJJ77A@prodigy.com, arpad_muranyi@ccm.fm.intel.com, jonp@qdt.com
> Message-Id: <9508221734.AA33207@btv.ibm.com>
> Received: by btv.ibm.com (AIX 3.2/UCB 5.64/fs4.03)
> id AA33207; Tue, 22 Aug 1995 13:34:05 -0400 Received: from
> eeyore.btv.ibm.com by btv.ibm.com (AIX 3.2/UCB 5.64/1.9)
> id <AA38196@btv.ibm.com>; Tue, 22 Aug 1995 13:34:10 -0400
> Received: by BTV (XAGENTA 4.0) id 3245; Tue, 22 Aug 1995 13:35:23 -
0400
> Received: from BTV by VNET.IBM.COM (IBM VM SMTP V2R3) with BSMTP id
7677;
> Tue, 22 Aug 95 13:38:48 EDT
> Received: from vnet.ibm.com by aurora.intel.com (5.65/10.0i); Tue, 22
Aug
> 95 10
> :
> 38:52 -0700
> Received: from aurora.intel.com by fmmail.fm.intel.com with smtp
> (Smail3.1.28.1 #2) id m0skxGs-000aoAC; Tue, 22 Aug 95 10:37 PDT
>
> Text item: External Message Header
>
> The following mail header is for administrative use
> and may be ignored unless there are problems.
>
> ***IF THERE ARE PROBLEMS SAVE THESE HEADERS***.
>
> From: "John Brennan" <jbrennan@VNET.IBM.COM>
> Date: Wed, 23 Aug 1995 09:01:18 -0400
> In-Reply-To: Your message of Tue, 22 Aug 1995 15:47:55 PST.
> <950822154755_9@ccm.fm.intel.com>
> Subject: Re: Non-Monotonic Drivers
> Cc: jbrennan@VNET.IBM.COM, EGJJ77A@prodigy.com, jonp@qdt.com,
> Will_Hobbs@ccm2.jf.intel.com
> To: Arpad Muranyi <Arpad_Muranyi@ccm.fm.intel.com>
> Message-Id: <9508231301.AA12722@btv.ibm.com>
> Received: by btv.ibm.com (AIX 3.2/UCB 5.64/fs4.03)
> id AA12722; Wed, 23 Aug 1995 09:01:19 -0400 Received: from
> eeyore.btv.ibm.com by btv.ibm.com (AIX 3.2/UCB 5.64/1.9)
> id <AA12775@btv.ibm.com>; Wed, 23 Aug 1995 09:01:20 -0400
> Received: by BTV (XAGENTA 4.0) id 3602; Wed, 23 Aug 1995 09:36:15 -
0400
> Received: from BTV by VNET.IBM.COM (IBM VM SMTP V2R3) with BSMTP id
2183;
> Wed, 23 Aug 95 09:36:24 EDT
> Received: from vnet.ibm.com by hermes.intel.com (5.65/10.0i); Wed, 23
Aug
> 95 06
> :
> 40:31 -0700
> Received: from hermes.intel.com by fmmail.fm.intel.com with smtp
> (Smail3.1.28.1 #2) id m0slG1j-000algC; Wed, 23 Aug 95 06:39 PDT
>
> Text item: External Message Header
>
> The following mail header is for administrative use
> and may be ignored unless there are problems.
>
> ***IF THERE ARE PROBLEMS SAVE THESE HEADERS***.
>
> From: "John Brennan" <jbrennan@VNET.IBM.COM>
> Date: Wed, 23 Aug 1995 16:22:17 -0400
> In-Reply-To: Your message of Wed, 23 Aug 1995 10:46:08 PST.
> <950823104608_4@ccm.fm.intel.com>
> Subject: Re: Re[2]: Non-Monotonic Drivers
> Cc: jbrennan@VNET.IBM.COM, EGJJ77A@prodigy.com, jonp@qdt.com,
> Will_Hobbs@ccm2.jf.intel.com
> To: Arpad Muranyi <Arpad_Muranyi@ccm.fm.intel.com>
> Message-Id: <9508232022.AA27032@btv.ibm.com>
> Received: by btv.ibm.com (AIX 3.2/UCB 5.64/fs4.03)
> id AA27032; Wed, 23 Aug 1995 16:22:19 -0400 Received: from
> eeyore.btv.ibm.com by btv.ibm.com (AIX 3.2/UCB 5.64/1.9)
> id <AA60513@btv.ibm.com>; Wed, 23 Aug 1995 16:23:16 -0400
> Received: by BTV (XAGENTA 4.0) id 3950; Wed, 23 Aug 1995 16:29:18 -
0400
> Received: from BTV by VNET.IBM.COM (IBM VM SMTP V2R3) with BSMTP id
1866;
> Wed, 23 Aug 95 16:29:25 EDT
> Received: from vnet.ibm.com by hermes.intel.com (5.65/10.0i); Wed, 23
Aug
> 95 13:
> 29:31 -0700
> Received: from hermes.intel.com by fmmail.fm.intel.com with smtp
> (Smail3.1.28.1 #2) id m0slMPX-000altC; Wed, 23 Aug 95 13:28 PDT
>
>
>

------- FORWARD, End of original message -------
Received on Thu Aug 31 08:58:17 1995

This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:52:28 PDT