Last weekend, I told one of colleagues how to tweak datasheet and do a guesswork for the current model. He learned much fast and did a trial analysis to a real design.
Here is the trial reportI(I translated it in English). From his report, I think it's so critical and most important that teaching HL user how to establish the model if user can get
real engineering meaning from the tool.
SSNSimulation.doc 721.0 KB
You make a very good point. The AC current model is essential to getting accurate results from the plane noise analysis in HyperLynx PI. Because of that, I think it's important to get some accurate information from the IC vendor. In your original post, you mentioned delta I/delta t and it's obvious that this is the critical parameter for plane noise analysis. Since the majority of the transmission plane model is reactive elements, the changes in current are what create the noise on the planes.
It seems to me that an IC vendor should know both the maximum dI/dt of the IC silicon, and the band pass frequency of the IC package. The lower of these frequencies establishes the dI/dt of the AC current source model that we should use in HyperLynx PI.
I also agree with your comments Yanfeng. One thing that has been brought up a few times is the use of CPM (Chip power model) files to profile the transient current behavior. From my understanding based on a few conversations, these are somewhat of a defacto standard for modeling the die power requirements. Do you know if you can one of these models from your IC vendors? The look like simple piecewise linear current sources and should accurately reflect the dI/dt behavior.
In your report, I noticed that you did not change the rise time on the Broadcom chip between simulation runs but did change the current amplitude. You mentioned that the results were off by half in the simulation, but I wonder if this is because the rise time of the current pulse wasn't accurate to the devices behavior. You may want to try and have it switch faster and see why type of results you get.
Well, in general we need an industry standard for IC power models. Perhaps a simple extension of IBIS spec.