|
|
|
Modeling & Simulation
Our Methods
Services We Provide
Tools, Resources &
|
Error: the Double Counting of Clamp Curves in an IBIS ModelThe IBIS Spec has always made it clear that model I-V current Vs voltage data is to be entered in the file under four [keywords]: [Pulldown], [Pullup], [GND Clamp] and [POWER Clamp. All simulators assume this and will add the currents back together, when it is appropriate, to arrive at a rising edge and falling edge composite curve that represents the behavior of the total I/O cell. Appropriate is determined by the structures that are actually present in the I/O cell, the voltage across it and the state the device is being held in. Thus, this problem will not occur if the cell is an input, if it does not contain one/both of the clamps, etc. Consider what happens if a simulation program has been properly instructed to add the power clamp current to the pulldown curve (and/or the ground clamp current to the pullup curve). What will happen if originally when the data was entered into the file if the power clamp has not been subtracted from the pulldown curve (and/or the ground clamp has not been subtracted from the pullup curve)? The composite curve will then double count the clamping action resulting in a much more robust (and therefore erroneous) clamping action on overshoot (high and low) swings. On a graphical plot this error is readily seen. The composite I-V curve will have double the slope of the applicable clamp in the overshoot region. An example from the example bug file follows. It is possible that the power clamp and the pulldown curve (and/or the ground clamp the pullup curve) are of the same order of magnitude. Then, adding them together for the composite curve will result in a composite curve slope that is double the clamping curve or pullup/pulldown curve But, such devices are unusual and need extra validation that they are accurate.
Figure 1: V-I curves with incorrect double counting of the clamp currents (courtesy of Cadence Design Systems through a loan of SpecctraQuest) By contrast, look at the curves of a correctly created IBIS file that follow. While there may be offsets in the composite curve due to the magnitude of the pullup or pulldown curves, the slope of the composite curve is close to that of the clamp curve(s) alone. This is because the clamp currents are normally significantly larger and increase much faster than that of a pullup or pulldown. After all, those structures are there to protect the output from overshoot.
Figure 2: V-I curves modeled correctly (courtesy of Cadence Design Systems through a loan of SpecctraQuest) |
|
|
|
