Learning Forum

[fachienchong]

Hello,

I need to use lumopt for inverse design, and during the simulation process, I found that my model remained unchanged for a long time. In the log file, I noticed the following issues. Should I adjust the auto shutoff time.And i noticed that my simulation process has lasted for several days,is it works normally?

thanks 

best

[Guilin Sun]

It does not diverge and autoshutoff is still 1 for so long simulation time might mean that you are using metal or periodic boundaries conditions without PML boundaries and with lossless materials.  While for periodic structure with plane wave excitation you will for sure use the periodic boundaries in the periodic directions, and PML along propagation, non-periodic structures usually use PML surrounding the structures. When the substrate is a metal, to aovid long simulation time, one may use perfect electronic metal to replace the real metal material but the other side of boundary should be PML.

Other causes of long simulation time might be due to many other factors, please refer to 

Ansys Insight: Slow convergence in FDTD

Please check your settings. If you still have the same issue, please post some screen copies here for the structure and settings. Otherwise please write a new post for any other questions.

[fachienchong]

Hello,
Thank you for your response.
Does this mean that my model is not converging? I have confirmed my PML condition settings and reduced the simulation bandwidth, but the same situation still occurs. My simulation is based on the terahertz band, and the device size has reached the millimeter level. Is the grid setting related to it?
best

[Guilin Sun]

If you use the default automesh with mesh accuracy 2 or higher, it should be ok. However, from what you said, it might be because the simulation time is too short. The default 1000fs is for optic frequency. at the THz range, the simulation time is much longer. Please check the source signal and see if it is a complete pulse without cutting off.

In addition, if you can read those posts it would be helpful:

为什么说仿真时间要足够长才能得到正确的频域结果

 

 One suggestion is to signficantly increase the simulation time, and let the autoshutoff min to terminate it.