微波EDA网,见证研发工程师的成长!
首页 > 研发问答 > 微波和射频技术 > 电磁仿真讨论 > Simulation time in HFSS and best way of system resources utilization

Simulation time in HFSS and best way of system resources utilization

时间:03-29 整理:3721RD 点击:
Hi
when I run a HFSS simulation on my laptop, convergence time is reasonable, but in case of a frequency sweep each step (frequency) takes a very long time!
I have Lap Top with this Specifications:
CPU: Core i7 1.73GHz (it shows 8 threads in CPU Usage history)
6 GB of RAM
64-bit Win7

HFSS Simulation Machin Options:
Number of processors: 6
Number of processors distrbuted: 1
Desired RAM limit: not set
Maximum RAM limit: not set

Also when an analyzing is in progress my laptop is very very slow. CPU usage is not high, but RAM usage of the HFSS is 3 or 4 GB!
How can it be resolved? thank you

Hi rizoeda,

We can't really tell how much RAM the simulation should take, since you haven't said anything about your setup, but for a domain size of roughly 50,000 tetrahedra, that sounds about right.

Thank you PlanarMetamaterials.
In fact my problem is that the main analysis doesn't take long time (in case of no frequency sweep, simulation ends in
appropriate time and with no errors) but when it starts sweeping frequencies each step takes too much time!

Each frequency sweep shouldn't take much longer than the initial convergence -- you're right, something sounds incorrect there. I would check your settings and ensure that there's not something limiting your resources during frequency sweeps.

Hello,
The issue is with your ram: you should,upgrade to more ram so you like be able to run multiple frequencies in parallel.
Currently you are forcing the frequencies to run in parallel by setting distributed manually . You force hfss to run out memory.If you set auto setting, hfss will check if enough ram and will run as much freq in parallel it can on your PC. The auto setting is available since version 16 (2015).

I agree that parallel operation surely force your system out of RAM so a good option could be to use auto settings or upgrade RAM. Another issue could be that may be at some certain frequencies which you sweep solution takes more time to converge which adds to the time.

What type of sweep are you using? Discrete sweep is very slow, but required if your materials are very frequency dependent. Interpolating is the fastest, fast sweep is still faster than discrete and allows you to save the fields at each frequency point.

Copyright © 2017-2020 微波EDA网 版权所有

网站地图

Top