Signal Integrity and IBIS Model Question

[ku637]

Hello,

Im new to Signal Integrity analysis , but i have a doubt regarding the suppply voltage ranges used inside an IBIS model

An IBIS Model as i understand includes slow weak,fast and typical corner behaviour of the chip


I see in a typical IBIS Model for an IC

| Variables typ min max
[Pullup Reference] 3.3 2.97 3.6
[Temperature Range] 25 125 -40
[Voltage Range] 3.3 2.97 3.6

So while running the simulation in different corners im seeing the output voltage steady levels at these voltages.

So i see for a different IBIS model this voltage range will be different

So incase of a driver and reciever , it looks like there will be difference in the voltage steady states

But my question is in a single PCB we have a single 3.3V which may supply to driver and as well as to receiver, the tolerance for that voltage is say 3%. In that case does this model defintion is reasonable , to increase accuracy do we need to edit the driver and receiver voltage ranges to 3% of 3.3V . Is that acceptable.

 

[dick_freebird]

"Acceptable" means what?

If you are trying to squeeze a "right answer" result out
of an IBIS based simulation, and do it by narrowing the
range of supply voltage, are you simply ignoring that
supply voltage can easily have excursions outside 3%?
Is the "right answer" during a design exercise more
important than having design margin in pocket, for
the things you didn't know about or anticipate?

This is always a tension in product development.

 

[ku637]

What exactly i was thinking is that.. Im having a power supply designed with say sufficient confidence with worst case analysis all done and also test measurements so that the onboard 3.3V is within say 5% tolerance limit.

That will be 3.135V to 3.465V across all load and temperature variations.

Im using this supply to power two ICs one is a driver (say U1) and other one is a receiver (say U2).

Im doing SI analysis now to validate the overshoot,undershoot and timing aspects of the digital signals

But the IBIS models from the manufacturer defines the supply voltage range of U1 and U2 differently inside their IBIS models

say for U1

| Variables typ min max
[Pullup Reference] 3.3 2.97 3.6
[Temperature Range] 25 125 -40
[Voltage Range] 3.3 2.97 3.6



say for U2

| Variables typ min max
[Pullup Reference] 3.3 2.85 3.65
[Temperature Range] 25 125 -40
[Voltage Range] 3.3 2.85 3.65


Now when i run simulations and analyzing the combinations of different corners the driver can be sending out data in say steady state level 3.3V and say receiver at that time is receiving signals with reference to 2.85 and the overshoot measurements at receiver may be taking from 2.85V steady level, this may cause failure.

But this scenarios will not exist actual system as my 3.3V to both U1 and U2 is within 5%.

So in this case what to do with IBIs model supply definitions?

 

[dick_freebird]

It's been a long time and probably many revisions to IBIS but are those supply "corner" values not indices for PWL table interpolation?

If so then you only need them to encompass the range, and timing would be interpolated if the index (Vdd) is not set to the min, Max or type value exactly?

Might see what the IBIS docs say about the functioning of those args.