Publications

Modeling Ccomp in IBIS Model

Modeling Ccomp in IBIS Model

By Timothy Coyle

The Many Faces of C_comp in IBIS

The C_comp parameter in the IBIS Specification represents the buffer die capacitance without anything else. So this does not include the package capacitance. The C_comp parameter is one of the more interesting elements of an IBIS buffer model and you’ll see why as we discuss it more below.

Why the C_comp Parameter is Important

Since the C_comp parameter is the buffer die capacitance and IBIS buffer models are used for system level simulations involving transmission lines it should be readily apparent why C_comp is important. It’s really all about the reflections in a system. Remember that a signal will travel down a transmission line to the receiver, bounce back from the receiver to the driver, and keep on doing this until the signal settles out. The magnitude of the ‘bounces’ is based upon the reflection coefficient but the capacitive loading plays a part in the reflections as well.

The bottom line is that the C_comp parameter value has to be representative of the actual device or the simulation result will not be accurate. For example, let’s say we have a high speed differential interface and the receiver is supposed to have a 0.5 pF input capacitance represented in the IBIS model with the C_comp parameter. What if that value was changed to be 10 pF? You would certainly see a worse looking eye diagram. And the same holds true for the C_comp parameter for an IBIS output buffer as well. If the reflected signal from a receiver bounces back to the driver and expects to see 0.5 pF of capacitance but instead sees 10 pF the reflection will be a lot worse.

Different Methods for C_comp Extraction and Accuracy

One of the issues with the C_comp parameter in IBIS is that you are only allowed to use one static value. This may not seem like a big deal since most datasheets specify an input or output capacitance with one value but when you start to think about how to extract the buffer capacitance you will see some of the issues. For example, what affects capacitance? Is it voltage dependent, frequency dependent, or both?

In the image above an example C_comp parameter extraction methodology is shown using a SPICE model of a buffer. The idea here is pretty simple: apply a voltage ramp to the input with the same slew rate for the rising and falling edge and measure the current on the output. The capacitance (C_comp in our case) is calculated by the equation: I = C * dV/dt This method provides voltage dependent C_comp values.

The above image shows the C_comp capacitance versus voltage and its dependency. Remember that you are only allowed to choose one value so you have to decide if you want to use the min, max, or perhaps the average as the final value. As long as the delta between the min and max are not that large I typically use the average as this tends to give the best correlation results.

But we also know that capacitance is frequency dependent as well based upon the following equation:

C = Im (I) / (2 * pi * f * Vac)

So another method to extract the C_comp capacitance would be to apply a current source on the output of the buffer at different frequencies and voltages. This way you can get a frequency and voltage dependent C_comp capacitance like the graph shown below.

Choosing a Final C_comp and Some Accuracy Issues

At the end of the day you have to put in one value for C_comp in the IBIS buffer model. It can be difficult to choose from the different extraction methods one C_comp value that will work with all simulation scenarios. But the reality is that in most cases as long as the C_comp value is close enough the simulation accuracy will not be impacted. The best thing to do is to simulate the IBIS buffer model in a realistic system level environment and correlate it to a known source measurement (SPICE model, lab data, etc) to ensure that the final C_comp value is performing adequate enough.

There are provisions in the IBIS Specification that allow you to split the C_comp parameter value into different corresponding values for the various pullup and pulldown structures. Some model vendors also provide multiple buffer models based upon different C_comp values for frequency and voltage. Some IO buffers have also shown a dependency whether it is in drive or receive mode and some model vendors provide multiple models for this as well. So while there are some alternatives to using one C_comp value they are not directly supported by the IBIS Specification or the value of doing so has not been sufficiently examined to warrant a recommendation.




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.