Re: Modelling bus switches and other series elements

Thanks to Chris and Stephen for your comments,

Hopefully we'll be able to discuss this topic on Monday.

Below is a clarification of some points:

1) Path info

I agree with Chris when he says that extra information is needed
within IBIS to reduce the number of potential simulations. But this
means giving more than an electrical description of the component;
a functional description would be necessary, something IBIS doesn't
do (yet!).

The proposal I sent would require all possible series paths to be
listed. This, in my opinion, is a minimum requirement. (With this,
a simulator could ask the user to select the paths to simulate).
A further optimisation might be to group series paths together to
indicate to the simulator that if one path is selected, then all
paths are selected.

What do simulator companies do today to model series devices
such as bus-switches?

2) Keywords

Regarding the choice of keywords, my original idea was to use
only the keywords [GND Clamp] and [Pulldown]. [GND Clamp] would
be for switch that is OFF, [Pulldown] for a switch that is ON.
To measure [Pulldown], one side of the series device would be
connected to GND.

Optionally, a [PWR Clamp] keyword could be used, but I don't like
this keyword !! In my opinion, it is redundant, but that's
another debate.

Instead, the proposal I make is more general, based somewhat
on network-analysis approach:
      - one input impedance measurement per operating state
      - one through (thru) impedance measurement per op. state

So for a switch, which can be ON or OFF, there are two
operating states => 4 measurement curves.

I think this is enough to describe fully the device. What is the
opinion of simultor companies?

Aside:
      For a standard buffer (one-port), only input impedances
      are measured. There are, at most, three operating states:
      HI-Z, LO and HI => 3 measurement curves.
      IBIS defines 4 curves, hence my view that one curve is
      redundant.

Regards,
John

-- 
John Fitzpatrick   <John.Fitzpatrick@ln.cit.alcatel.fr>    
Alcatel Telecom, 4 rue de Broglie, 22304 Lannion, France
Tel: +33(0)2.96.04.79.33  Fax: +33(0)2.96.04.85.09
Original proposal:
IBIS cannot correctly model series devices. This isn't
much of a problem for simple passive devices (R,L,C) or diodes.
However, it's a major handicap for bus-switches (e.g. 74CBT range
from Texas Inst). 
I've seen seen this feature on the wish-list, but am not aware
of any proposals to handle it.
So here's my suggestion for extending IBIS to model series switches. 
It requires 3 new keywords and one new modeltype.
1) Define a new keyword [Series Pin]
   This keyword is at the same level as, say, [Diff Pin]
   [Series Pin]
   |Pin1      Pin2    Buffer   
       3         4    switchA     |switchA is buffer from 3 to 4
       4         3    switchAneg  |Buffer from 3 to 4 is non-symmetric
      12        13    switchA
      12        14    switchB     |12 can be routed to 13 or 14
         
   The two pins of the series element are identified, along with
   the the associated series buffer. This buffer corresponds
   to measurements made at Pin1.
   If the buffer is not symmetric, a different buffer may be
   specified with the pin order swapped.
   In the case of a crossbar switch, all possible series
   connections must be listed. IBIS contains no information
   concerning how or when one of the connections is selected.  
   The buffer name given here supersedes that given under the
   [Pin] keyword. 
2) Define new model type SERIES-SWITCH
   Define new keywords [Series] and [Series OFF]
   Four curves are obligatory: [Pulldown]
                               [GND Clamp]
                               [Series]
                               [Series OFF]
   [GND Clamp] and [Series OFF] are measured for switch OFF.
   [Pulldown] and [Series] are measured for switch ON.
   [GND Clamp] + [Pulldown] are standard measurements, made at one pin.
   The other pin, identified under the [Series Pin] keyword, 
   is assumed connected to ground.        
 
   [Series] and [Series OFF] are through measurements ie between
   the two pins.
   The [Ramp] keyword is also required. The test load must
   include a voltage source.
   This gives enough information to model a switchable non-linear
   resistor.
   We could also define a model type SERIES which would not
   be switchable: only [Pulldown] and [Series] needed in this case.
   To allow for series frequency-dependent devices, additional
   R,L,C paths might be defined.
Looking forward to your feedback,
Regards,
John
-- 
John Fitzpatrick   <John.Fitzpatrick@ln.cit.alcatel.fr>    
Alcatel Telecom, 4 rue de Broglie, 22304 Lannion, France
Tel: +33(0)2.96.04.79.33  Fax: +33(0)2.96.04.85.09