3 issues (was Re[3]: Interpretation of I/O data )

Hello all!

I'd like to thank Arpad Muruanyi taking the time to answer my questions.

These questions arose from internal discussions here at Alcatel as to
whether an IBIS model can replace the electrical characteristics
specification for a component.

It would be nice if a combination of IBIS and VHDL models could provide, as a
minimum, the same information as a paper datasheet. (This would mean
extending the definition of the IBIS model to more than behavioral models.)

Below are some ideas on

   1) INTERPRETING INPUT V/I CURVES
   2) ABSOLUTE MAXIMUM RATINGS
   3) NEW INPUT SUB-PARAMETERS

I hope you find them of interest. Depending on feedback, I may write 2 & 3
up as a BIRD. (Does Alcatel need to be an IBIS member if I want to submit
a BIRD? Should I have said EGG?)

Regards,
John

1) INTERPRETING INPUT V/I CURVES
 
Arpad asked:
> What are you referring to when you say FAST-type?

Significant input current.
(I was a bit confused by the receiver model, which contains only clamping diode
curves.)
 
Has anybody modelled an input with significant input current?
 

Arpad asked:
> Shouldn't the Bus-hold circuit be modeled as an output rather than an input?

Hmm. I guess so - if that's the buffer's only function.
But "bus hold" is usually a feature associated with an input buffer,
so I think it would be better simply to include the current in or out
as part of the input buffer's V/I curve.

Maybe someone from TI could post an example of model of an LVT input?

2) ABSOLUTE MAXIMUM RATINGS

I asked:
> >Can a buffer which doesn't clamp when powered, clamp
> >when unpowered (for applied voltages in the range 0 to Vcc)?
Arpad replied:
> If the power on the input disables the clamping somehow, I can see that it could
> clamp when it is not powered. But it would be helpful if you would state what
> kind of input you are referring to. Regular CMOS, or something unusual, which I
> don't know about? CMOS DOES clamp ewen when it is powered (0.6 volts above
> Vdd).

It does seem natural to make the assumption that if the buffer
doesn't clamp from Vcc to 2Vcc, over the specified Vcc range, then it won't
clamp inputs from 0 to ~|Vcc| when Vcc = 0.
Similarly, a 3.3V buffer, which clamps at Vcc+2V normally, will clamp at
2V when its Vcc = 0.
But this is only my interpretation of the IBIS data.

On re-reading my question, I realise that what I am looking for is an
Absolute Maximum Ratings section in the IBIS model.

With IBIS 2.1, it's not possible to answer questions like:

   - What is the maximum over/undershoot voltage or current tolerated by
     a component?
   - Over what time-scale can an over/undershoot safely be tolerated?
   - If the power supply fails, is it sufficient protection to limit the
     input current? To what value? Over what time-scale?
   - etc
   
A good example of the "modern" absolute maximum ratings is that
specified by Xilinx:

  "Maximum DC overshoot above Vcc or below GND must be limited to either
   0.5 V or 10 mA, whichever is easier to achieve. During transitions, the
   device pins may undershoot to -2.0 V or overshoot to Vcc + 2.0 V, provided
   this overshoot or undershoot lasts less than 20 ns.

Can the IBIS model evolve to take this type of specification into account?

My suggestion is to supply a table of voltage v.time for each
[Model] entry:

-----------------------start of example-----------------------------

[Abs Max Applied Voltage Pos Ref]
POWER | Simple example, only one power supply
                            | If not present, assume GND

[Abs Max Applied Voltage Neg Ref]
GND | If not present, assume GND

[Abs Max Applied Voltage]
| Time Positive_Voltage Negative_Voltage
    0 2.0V -2.0V
   20n 0.5V -0.5V

-----------------------end of example-------------------------------

Other absolute maximum rating may also be necessary (supply voltage, temperature, etc)

The maximum current can be derived from the V/I curves *IF* these can
be extrapolated for use over absolute maximum supply voltage range.

Any comments?

3) NEW INPUT SUB-PARAMETERS

Arpad said:
> If it
> is important to you, you might want to raise this (and the hysteresis) issue in
> the Open Forum Meeting. I brought it up last time, and people are open to these
> kinds of comments, requests

My suggestion is that two new sub-parameters might be
added to the [Model] keyword.

Vh = 0.4V | minimum hysteresis value
dt/dV_in = 10ns/1V | maximum input transition rise or fall time

These values could be used to issue warning during a simulation.

Any comments?

--
John Fitzpatrick   <John.Fitzpatrick@alcatel.fr>
Alcatel CIT, 4 rue de Broglie, 22304 Lannion, France
Tel: (+33)96.04.79.33  Fax: (+33)96.04.85.09