Jean,
The "worst case" depends entirely on what you are trying to measure
and the topology of your nets. There is no easy way to define
"worst case." However, some of your questions can be answered.
Most likely your simulator ignores the temperature variations.
The temperature range indicates a reasonable operating range
for the part, and is supposed to have some connection to min
and max corners, but there is no direct tie between the temperature
you choose and the model that gets simulated unless your simulator
specifically defines one, in which case you'd better ask the
simulator vendor what is intended.
If you want a weak driver then use the "min" of most everything,
but use the "max" C_comp.
If you want a high impedance receiver, you can figure out what
to use as well as anyone.
If you want the "worst case" for a particular net, all bets are
off. If by worst case you mean the most ringing, or the longest
time to "final settle", then that depends on the net topology
as well as the driver and receiver characteristics. For example,
if you are using a balanced topology, say a star, then the worst
case is probably unbalanced receivers, some high impedance, some
low impedance.
There is no easy answer, although there is some theory behind this
which can be used to get an idea of what might be "worst case."
The only way to be sure is to run a Monte-Carlo type variation
of all the drivers and receivers, not to mention process variations
on you board, actual lengths of lines, etc.
Christopher Reid
Interconnectix
http://www.icx.com
JEAN-CHRISTOPHE_PAUTRAT@HP-France-om5.om.hp.com wrote:
>
> Hello,
>
> maybe someone already asked this question in the past, I don't know.
>
> I'm trying to figure out how to tweak all the parameters to exercise
> worst case simulations. Let's take the example of a strong and fast
> driver. The C_comp and parasitics will be the only parameters of the
> driver to be weak. All receiver parameters will be weak (parasitics
> and clamping diodes only). Let's say I tune the parameters of the PCB
> to be weak also. So far everything is logical.
>
> I've got a problem with the voltage and temperature parameters. Let's
> say I have ibis models for 5V cmos device, where min and max
> parameters were extracted in the following conditions :
>
> min = minimum voltage, max temp deg C
> max = maximum voltage, min temp deg C
>
> then, I would have to supply the driver with 5.5V and the receiver
> with 4.5V which is absolutely not realistic. This case never happens
> on a motherboard. Also what voltage should be used to connect the
> components tied to Vcc in between them ?
>
> Same thing for the temperature. There will never be let's say @ -50C
> on one side and +70C on the other side. Moreover, what temperature
> value shall I give to the simulation software for the components in
> between the driver and receiver ?...
>
> In summary : what are the Vcc and temp values I shall use for the
> corner case simulations ?
>
> thanks and best regards,
>
> Jean-Christophe PAUTRAT,
> R&D Engineer, Design Engineering
> Commercial Desktop Computer Division,
> H E W L E T T - P A C K A R D
Received on Wed Jul 16 09:16:58 1997
This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:53:46 PDT