Yuta Inoue

Fast Circuit Simulation Based on Parallel-Distributed LIM using Cloud Computing System

This paper describes a fast circuit simulation technique using the latency insertion method (LIM) with a parallel and distributed leapfrog algorithm. The numerical simulation results on the PC cluster system that uses the cloud computing system are shown. As a result, it is confirmed that our method is very useful and practical.

GPGPU-FDTD method for 2-dimensional electromagnetic field simulation and its estimation

For signal/power integrity analysis of the high density packages and printed circuit boards, the FDTD (Finite-Difference Time-Domain) method has been widely used. In order to apply to large-scale problems, a variety of acceleration techniques are required. This paper describes a GPGPU-FDTD (General Purpose computing on GPU (Graphic Processing Unit)-Finite-Difference Time-Domain) method for massively parallel electromagnetic field simulation. Finally, it is confirmed that GPGPU-FDTD method shows the high-performance when the computational algorithm is programmed suitably for the architecture of GPU.

Parallel-Distributed Block-LIM for Transient Simulation of Tightly Coupled Transmission Lines

In this paper, the parallel-distributed block-latency insertion method (block-LIM) is proposed for the fast transient analysis of a large-scale circuit that includes lots of coupling elements, such as mutual inductance and mutual capacitance. A conventional SPICE-like simulator requires an enormous cost for transient analysis of large-scale equivalent network that includes tightly coupled transmission lines, which is derived by using well-established commercially based extractors. The proposed method is based on the leapfrog algorithm, and can efficiently analyze tightly coupled transmission lines. First, the original LIM and the block-LIM are reviewed briefly. Next, the parallel-distributed block-LIM is proposed for the fast transient simulation and is implemented on the personal-computer-cluster system. Finally, some numerical results are shown, and it is confirmed that the proposed technique is useful and efficient for the simulations of the tightly coupled transmission lines.

A Hybrid Implicit–Explicit and Conformal (HIE/C) FDTD Method for Efficient Electromagnetic Simulation of Nonorthogonally Aligned Thin Structures

In a conventional finite-difference time-domain (FDTD) method, the computational cost increases as a grid spacing decreases because the number of unknowns increases and a time step size decreases to fulfill a numerical stability condition. Especially, it is inherently time-consuming to analyze thin structures such as a printed circuit board (PCB) and nonorthogonally aligned objects such as interconnection patterns on the PCB. In this paper, we propose an efficient electromagnetic simulation technique which is constructed by combining a hybrid implicit-explicit (HIE)-FDTD method and a conformal FDTD (CFDTD) method. The proposed technique can avoid the aforementioned problems by exploiting advantages of the HIE-FDTD and CFDTD methods. An example model of the interconnection pattern on the PCB is analyzed by the proposed and conventional methods, and we compare their accuracy and efficiency.

Parallel-distributed block-LIM-based fast transient simulation of tightly coupled transmission lines

In this paper, parallel-distributed block latency insertion method (block-LIM) is proposed for the fast transient simulation of the large scale circuit which includes lots of coupling elements such as mutual inductance and mutual capacitance. First, the block-LIM formulation for the network with mutual inductance and mutual capacitance is shown. Next, the parallel-distributed block-LIM is described. Finally, some numerical results are shown and it is confirme that the proposed technique is useful and efiicien for the simulations of the tightly coupled transmission lines.

GPGPU-based Latency Insertion Method: Application to PDN simulations

With the progress of high-density integration technology of the circuits, a variety of signal and power integrity problems have become serious and important for the electronic design. This paper describes the fast circuit simulation by GPGPU-LIM (GPGPU-based Latency Insertion Method). First, LIM is reviewed, which is a fast algorithm. Next, implementation of LIM on the general purpose computing on graphic processing unit (GPGPU) is shown. Furthermore, this method is applied to the simulation of power distribution networks (PDNs). Finally, it is confirmed that GPGPU-based LIM is very practical and efficient for the large-scale PDN simulations.

Multi-GPU HIE-FDTD method for the solution of large scale electromagnetic problems

In this paper, Multi-GPU hybrid implicit-explicit finite-difference time-domain (HIE-FDTD) method is proposed for the solution of large-scale electromagnetic problems. The HIE-FDTD method is one of the weakly conditionally stable algorithms which can use a larger time step size than that for the conventional FDTD method. Furthermore, the HIE-FDTD method is suitable for parallel computing because this method requires only the local matrix operations. First, the HIE-FDTD method is reviewed briefly. Second, the proposed method is described for the efficient electromagnetic field simulation. Finally, the efficiency of the proposed method is evaluated by some numerical results.

Fast fullwave simulation based on parallel-distributed HIE-FDTD method

In this paper, parallel-distributed hybrid implicit-explicit finite-difference time-domain (HIE-FDTD) method is proposed for an efficient fullwave simulation of large-scale problems. The HIE-FDTD method is one of the weakly conditionally stable algorithms which can set a larger time step size than that for the conventional FDTD method. In addition, the HIE-FDTD method is suitable for parallel computing. First, the HIE-FDTD method is reviewed briefly. Second, the proposed method is described for the fast fullwave simulation. Finally, the efficiency of the proposed method is evaluated by performing computer simulations.

Multi-GPU based fast electromagnetic simulation method for analysing PCB

This paper describes the efficiency of the fast electromagnetic simulation method for analysing the PCBs. This method is constructed by combination of the multi graphics processing unit (GPU) systems and the weakly conditionally stable algorithms. The weakly conditionally method is faster than the conventional ones in the case of fine cell including the computational domain. In addition, this method is suitable for parallel implementation with multi-GPU computing. First, the weakly conditionally stable algorithm is described briefly. Second, this method is shown. Finally, the efficiency of the multi-GPU based method is verified by analysing an example of the practical PCB.

A framework for fast SI/PI simulation with acceleration techniques

For the signal/power integrity (SI/PI) and EMI design, the simulation of large-scale network, for example, printed circuit board (PCB) and package system with PDN (power distribution network), is strongly demanded. In this paper, a framework for a fast SI/PI simulation with several characteristic techniques is described. First, the concept of locally implicit block leapfrog scheme (based on the block-LIM) for the efficient PDN simulation, which can be more efficient than the original leapfrog method even for the network with tightly coupled subnetworks, is introduced briefly. Furthermore, the derivative of the block-LIM is also described in the cases 1) with conformal meshes and 2) of implementation on the multi-GPUs 3) with multi-rate technique.