Frequenty Asked Questions


Why is PhysPack called a True 3D electromagnetic solver?
PhysPack solves 3D Maxwell’s equations for 3D structures. There are no assumptions of infinite dielectric layers or TEM-wave behavior. While there are indeed a few 3D commercial solvers around, some commercial solvers which should be termed 2.5D because of inherent restrictions, and which therefore are not applicable for many beyond-die applications, are still marketed as 3D solvers. Hence the assertion that PhysPack is a True 3D EM solver.


What is Maxwell Accuracy?
PhysPack is an accurate EM solver with unprecedented speed and capacity. In general, some speed and capacity can be obtained with any EM solution method by reducing the accuracy. However, PhysPack’s scalability, speed, and capacity is obtained through deep algorithm IP which ensures that exceptionally high accuracy is maintained. Hence all coupling and electromagnetic effects are captured as expected by a true Maxwell equation solver. We call this Maxwell Accuracy.


What is True Parallelization?
Distributed computing has been used in several competing products. This allows multiple CPUs to run different simulations at different frequencies, or for different designs, simultaneously. True parallelization is significantly more powerful, and entails the parallel solution of the underlying electromagnetic system, even for a single frequency and single design. The power of true parallelization is especially evident for large-scale problems. Distributed computing simply tries to solve multiple copies of the problem on different CPUs. However, if the problem does not fit on the memory associated with a single CPU or core of a CPU, such an approach either fails or has to resort to time-expensive disk storage. This is especially true for the increasingly prevalent shared-memory multi-core architectures where a severe memory bottleneck is created as the multiple distributed processes feed from the shared memory pool. True parallelization, on the other hand, ensures the same turnaround by parallelizing the underlying electromagnetic solution such that the memory of multiple CPUs and cores can be effectively utilized. This kind of parallelization requires inherent parallel architecture and algorithms, understanding of the constraints of Amdahl’s law, and thread-safe methods. PhysPack’s true parallelization enables large-scale solutions on multiple cores, and also enables distributed computing across multiple CPUs.


What is linear scaling?
Most electromagnetic solvers need to either solve or timestep large matrix systems of size N. In the case of finite element solvers, these are large sparse systems, while for boundary element methods, these are smaller but dense systems. While the cost of classical direct methods scale cubically with N, even the best commercial techniques tend to scale as the square of N, and with similar memory scaling. The breakthrough fast solver technology at Physware enables memory and time that scales only lineearly in N, enabling dramatic speedups and memory savings for large problems. As problems become larger, the relative advantage of linear scaling methods grows even more significantly.


What is Broadband Solution? Many high-frequency techniques develop issues at low frequencies. These methods tend to have a floor below which the methods are unstable and inaccurate. Often separate products are sold as band-aids for these frequency regimes. Physpack is a true broadband solver. It produces maxwell-accurate high frequency results but can also produce equally accurate results at as low a frequency as desired, which is particularly useful for spice models. This is made possible by special transformations and preconditioners deep in the electromagnetic solution process, all of which are transparent and seamless to the user. In addition, PhysPack provides, in the same gui and flow, a quasi-static mode for direct model generation that bypasses S parameters.


What is Design at the Speed of Analysis?
The fast linear scaling solver IP at Physware enables large scale design for integrity solution while preserving Maxwell-Accuracy. When used in conjunction with True Parallelization, dramatic speedups of one to two orders of magnitude in solution can be obtained. While speed and scale are one metric, a more disruptive paradigm also arises. Due to the speed, scalability, accuracy, and due to the inclusion of several design-friendly features and related automation, designers can now complete full designs with PhysPack in times where one or a few simulation passes are possible with competing products. We call this paradigm Design at the Speed of Analysis.


What does think beyond the die™ mean?

The classical approach of simulation and design tends to consider chips, packages, and boards as separate. However, with increasing speed, functionality, and flexibility, such demarcations are getting blurred. RDL layers and I/O placement can be considered to be part of both chips and packages. SoCs, SiP’s and SoP’s have characteristics of all three entities. think beyond the die™ is a vision to move beyond traditional boundaries of simulation, analysis, and design, which requires flexible and efficient design tools which understand and move beyond classical rigid interfaces.


What is Design for Integrity™?
System-on-package designs are becoming extremely complex, with simultaneous requirements of power integrity, signal integrity, and electromagnetic interference mitigation. All of these have different requirements, and often require different simulation techniques. PhysPack’s scale, speed, accuracy, and holistic approach enable these seemingly disparate SI-PI-EMI needs to be addressed in a unified manner. Furthermore, PhysPack’s approach is not simply about faster analysis but also in helping in design through several design-specific features, We consider this as enabling Design for Integrity™.