RE: different rising and falling Ramp dV ( AGTL+)

Eckhard,
I have seen buffers modeled in IBIS ver 3.2 as you describe with the strong
pull down in the top level model, the "on die" termination in the power
clamp, and the weak pullup in a sub model. But the case I mentioned below
was a ver 2.1 level model and the IC has no "on die" termination. My intent
was to encourage further discussion about (alternate?) ways to model
buffers.

Aubrey Sparkman
Signal Integrity
Aubrey_Sparkman@Dell.com
(512) 723-3592

-----Original Message-----
From: Lenski Eckhard [mailto:Eckhard.Lenski@icn.siemens.de]
Sent: Monday, March 26, 2001 12:01 AM
To: 'Aubrey_Sparkman@Dell.com'; arpad.muranyi@intel.com;
ibis-users@vhdl.org
Subject: AW: different rising and falling Ramp dV ( AGTL+)

May it be, that this should be modelled ( or is modelled ) with a driver
schedule?

This is my experience with AGTL+, as the weak pullup-transistor is not
always on.
It is modelled in a driverschedule, with the pulldown as the top-level-model
and the pullup as a 'sub-model'
Furthermore there appear now models, which includes the termination on chip,
( and the termination is always 'on' )
which is then modeled in the powerclamp of the top-level-model.

Eckhard Lenski
Siemens AG ICN M TC 6
Hofmannstr.51 81359 München
Tel : 0049 89 722 27776
Fax: 0049 89 722 44692
Email: eckhard.lenski@icn.siemens.de

> -----Ursprüngliche Nachricht-----
> Von: Aubrey_Sparkman@Dell.com [SMTP:Aubrey_Sparkman@Dell.com]
> Gesendet am: Freitag, 23. März 2001 23:42
> An: arpad.muranyi@intel.com; ibis-users@vhdl.org
> Betreff: RE: different rising and falling Ramp dV
>
> The case I had in mind has the dV rising about 1/3 the dV falling. The dV
> rising seems to be taken from the Rising V/T curve with 50 ohms to ground
> and the dV falling seems to be taken from the Falling V/T curve with 50
ohms
> to Vtt. There is no pullup in the power clamp for this buffer. I didn't
> ask my supplier whether the pullup was always on but I don't think it is.
> This buffer is used on a bus that has at least one pullup. This seems to
be
> a little different from what you discussed; perhaps an alternate method?
> Perhaps I am missing something?
>
> BTW, my simulator seems to be OK with this as my simulations match what we
> measure in the lab.
>
> Aubrey Sparkman
> Signal Integrity
> Aubrey_Sparkman@Dell.com
> (512) 723-3592
>
>
> > -----Original Message-----
> > From: Muranyi, Arpad [mailto:arpad.muranyi@intel.com]
> > Sent: Friday, March 23, 2001 2:58 PM
> > To: ibis-users@vhdl.org
> > Subject: RE: different rising and falling Ramp dV
> >
> >
> > Sorry, I got dyslexic a bit... For the open-drain GTL buffer
> > the resistor is not connected to GND, but to Vtt, and the
> > "ON" endpoints of the Vt curve will definitely not go to GND,
> > but the "OFF" endpoints should be at Vtt.
> >
> > Arpad
> > ==============================================================
> >
> > -----Original Message-----
> > From: Muranyi, Arpad [mailto:arpad.muranyi@intel.com]
> > Sent: Friday, March 23, 2001 12:52 PM
> > To: ibis-users@vhdl.org
> > Subject: RE: different rising and falling Ramp dV
> >
> >
> > Aubrey,
> >
> > In light of the original question from Mike and my answer to it,
> > I don't expect anything different in this case. I assume that
> > the weak pullup you are talking about is always on. So the
> > two Vt curves you will make will be generated with a resistor
> > to ground, and the rising and falling edge will characterize
> > how the switched pulldown is turning on and off. The amplitude
> > of the rising and falling edge will be the same, but may not
> > start or end on the rail (GND) exactly due to the weak pullup.
> >
> > I am not sure if this answered your question, or whether I
> > understood your question fully. Did this help?
> >
> > Arpad
> > ===============================================================
> >
> > -----Original Message-----
> > From: Aubrey_Sparkman@Dell.com [mailto:Aubrey_Sparkman@Dell.com]
> > Sent: Friday, March 23, 2001 10:49 AM
> > To: arpad.muranyi@intel.com; ibis-users@vhdl.org
> > Subject: RE: different rising and falling Ramp dV
> >
> >
> > Arpad,
> > Would you comment on what you expect for something like AGTL+ which is
> > almost an open drain, but has a weak pullup? The weak pullup
> > alone can't
> > pull a 50 ohm resistor up very far.
> >
> > Aubrey Sparkman
> > Signal Integrity
> > Aubrey_Sparkman@Dell.com
> > (512) 723-3592
> >
> >
> > > -----Original Message-----
> > > From: Muranyi, Arpad [mailto:arpad.muranyi@intel.com]
> > > Sent: Friday, March 23, 2001 10:28 AM
> > > To: ibis-users@vhdl.org
> > > Subject: RE: different rising and falling Ramp dV>
> > >
> > >
> > > Mike,
> > >
> > > The reason dV can be different for rising and falling edges
> > is because
> > > the voltage swing does not have to be the same (in contrary of your
> > > statement).
> > >
> > > A falling edge ramp requires that the R_load resistor is
> > connected to
> > > the supply rail. This means that the falling edge characterizes the
> > > pulldown turning on. A rising edge ramps requires that the R_load
> > > resistor is connected to GND. This means that the rising edge ramp
> > > characterizes how the pullup turns on. Since the pullup
> > and pulldown
> > > IV curves do not have to be the same (strength), you can
> > get different
> > > swings for these two edges, hence the dV numbers will also be
> > > different.
> > >
> > > Now, if you consider an open drain type buffer, where the rising and
> > > falling edges are characterized with R_load connected to the
> > > same rail,
> > > your statement is correct, the swing will be the same, therefore dV
> > > should also be the same. However, notice that in this case one edge
> > > describes how the device turns on, the other edge describes how the
> > > same device turns off.
> > >
> > > Arpad
> > > ==============================================================
> > > =========
> > >
> > > -----Original Message-----
> > > From: Mike LaBonte [mailto:mike@labonte.com]
> > > Sent: Thursday, March 22, 2001 3:19 PM
> > > To: ibis-users@vhdl.org
> > > Subject: different rising and falling Ramp dV
> > >
> > >
> > > Many IBIS models have different values for dV in the dV/dt_r
> > > and dV/dt_f.
> > > The clipping below from IBIS spec shows an example. But the
> > > usage rules
> > > imply that dV must be 60% of the voltage swing. If the
> > > voltage swing is
> > > measured between the 2 steady-state voltages, then how
> > could you have
> > > different values for rise and fall dV? If there is variation in how
> > > simulators handle the Ramp dV values, then this may matter.
> > >
> > > Mike LaBonte
> > >
> > > |=============================================================
> > > ==============
> > > ==
> > > | Keyword: [Ramp]
> > > | Required: Yes, except for inputs, terminators, Series and
> > > Series_switch
> > > | model types.
> > > | Description: Defines the rise and fall times of a buffer.
> > > The ramp rate
> > > | does not include packaging but does include
> > > the effects of
> > > the
> > > | C_comp parameter.
> > > | Sub-Params: dV/dt_r, dV/dt_f, R_load
> > > | Usage Rules: The rise and fall time is defined as the time
> > > it takes the
> > > | output to go from 20% to 80% of its final
> > > value. The ramp
> > > | rate is defined as:
> > > |
> > > | dV 20% to 80% voltage swing
> > > | -- = ----------------------------------------
> > > | dt Time it takes to swing the above voltage
> > > |
> > > | The ramp rate must be specified as an
> > > explicit fraction and
> > > | must not be reduced. The [Ramp] values can
> > > use "NA" for the
> > > | min and max values only. The R_load subparameter is
> > > optional
> > > | if the default 50 ohm load is used. The
> > > R_load subparameter
> > > | is required if a non-standard load is used.
> > > |-------------------------------------------------------------
> > > --------------
> > > --
> > > [Ramp]
> > > | variable typ min max
> > > dV/dt_r 2.20/1.06n 1.92/1.28n 2.49/650p
> > > dV/dt_f 2.46/1.21n 2.21/1.54n 2.70/770p
> > > R_load = 300ohms
> > >
> > >
> >
> >
> >
> >
> << Datei: Sparkman, Aubrey.vcf >>
 
Received on Tue Mar 27 07:36:47 2001