Publications

IBIS Model Data Generation Methods

IBIS Data Generation Methods

By Timothy Coyle

The Different Ways to Get IBIS Data

You can really only generate IBIS data from SPICE model simulations or actual lab measurements from the physical device itself. Here are a couple caveats about each approach.

SPICE versus Lab Data

There has been a long standing debate on which source is better for IBIS model data: SPICE models that can cover the full process range of the device transistors or lab measurements that will give the actual performance of the device based upon process corner.

From my experience I think the best method is to use SPICE models as the source for IBIS model data as it provides the most robust and straight-forward way to generate the IBIS data. The process corners and simulation setup can easily be changed and the affects of physical measurement equipment do not have to be de-embedded or considered. The downside to this approach is that it relies on the SPICE model to be accurate and representative of the device (which it always should be anyways) but sometimes is not. The bigger issue for some users is that the actual device process corners will probably not be as bad as the simulated device process corners but that debate really turns into a simulation methodology discussion we will save for a later time. I do agree that the IBIS data should be measured in the lab and used as a correlation vehicle to ensure that the SPICE model is within the device specification.

AC versus DC IV Curve Data Generation

When generating the IBIS IV data from a SPICE model simulation the easiest method is to perform a DC sweep with a voltage source. Sometimes though this won’t work due to circuit convergence issues or perhaps because the IO buffer SPICE model includes a clock so you need to exercise that in order to generate the IV data. You can use a transient simulation by applying a very slow ramp and measuring the current-voltage so it is essentially a DC sweep to get the same results.

The above image shows a partial listing of an example generic SPICE file that was used to generate a pulldown IV curve of an IO buffer. A DC voltage source was placed on the output node and a DC sweep was applied to this node from –Vcc to 2*Vcc. The current was printed in an output file so the current-voltage table results can be formatted into the IBIS model. Notice that depending on how the flow of current is defined you may need to artificially change the polarity sign of the current.

The above image shows a partial listing of an example generic SPICE file that was used to generate a pulldown IV curve of an IO buffer. A very slow voltage ramp in the form of a PWL source is applied to the output node from –Vcc to 2*Vcc with a transient analysis. The voltage and current was printed in an output file so the current-voltage table results can be formatted into the IBIS model. Notice that in the voltage ramp sweep the polarity of the voltage (-Vcc to 2*Vcc) is reversed in order to get the correct sweep range polarity.

References and Links

“A First Approach to IBIS Models: What They Are and How They Are Generated”, Analog Devices, Application Note # AN-715, 2004.




Learn How IBIS DRC Can Automatically Validate IBIS Models for Signal Integrity Simulation

IBIS DRC uses over 75+ built-in quality checks based off of the IBIS Specification and the IC vendor datasheet to provide a comprehensive quality report.