| Author |
Message |
krazydav
Joined: 31 May 2003
Posts: 4
Location: San Diego
|
 31 May 2003 4:27 How to minimize EM simulate time in MWO? |
|
|
|
I am trying to run the EM simulator for a very sensitive edge coupled filter. So I set the cell size to 1 mil---- not only will it take 42hrs.(I can live with  ) but I run out of memory on my computer. I get an error message that I need to free up some memory. I have 1G ram and 40G HD... What do I need to do to get this thing to run? Any help would be apprecieated.
|
|
| Back to top |
|
 |
goxy
Joined: 25 Jul 2002
Posts: 268
Helped: 1
|
| 31 May 2003 14:35 |
|
|
|
In MWO installaton dir. you have Users Guide and chapter:
MWO User Guide > Electromagnetic Analysis > 7.8 Using the EM Simulator Effectively > 7.8.4 Minimizing Solution Time
Maybe this can help you, but no help with big structures and very small cells.
If you have not enought RAM mem. this is killing proces for simulator because OS use HD for virtual mem and simulator work with all matrix so that will to slow simulation and 10 times. One solution is more memory and more processors. Memory is the most important thing for EM simulation.
Also, you can upload your MWO project here, this is the best way that somebody help you.
|
|
| Back to top |
|
|
loucy
Joined: 26 Aug 2001
Posts: 424
Helped: 10
|
| 01 Jun 2003 0:30 |
|
|
|
1. how are the coupling edges oriented? if they are not parallel to the sidewall, go with other mom tools. if mwo is your only choice, change your design so that the coupling edges are pararell to the sidewall, e.g., by adding some bends.
2. reduce the size of enclosure box and position the metals so that the cell size could be bigger.
|
|
| Back to top |
|
|
toonafishy
Joined: 18 Jun 2002
Posts: 425
Helped: 19
|
| 01 Jun 2003 6:26 |
|
|
|
| I have found that a 1 mil grid will always take much too long. One thing you can try is to make the thick center of your line a coarse mesh. Then put a strip along the edge and make this a fine mesh. In most cases, you only need a coarse mesh in the middle but a fine mesh along the edge. I have also seen that a grid that is bigger, say 2 or 5 mil, will give a result that is close enough but that may not be the case in your circuit. M/o/m/e/n/t/u/m seems to be better in this respect since it can handle trapezoids.
|
|
| Back to top |
|
|
krazydav
Joined: 31 May 2003
Posts: 4
Location: San Diego
|
| 02 Jun 2003 2:27 |
|
|
|
| Thank you all for the input. I have used the help menu and guides quite a bit. For the box I am simulating the actual box that I will use. I am still learning about the meshing densities, perhaps I should focus on that possible solution. Again thanks for all the input.
|
|
| Back to top |
|
|
loucy
Joined: 26 Aug 2001
Posts: 424
Helped: 10
|
| 02 Jun 2003 3:05 |
|
|
|
| the matrix filling time in MWO is mostly determined by the size of the 2D FFT, in this case, =box size/cell size. Making the mesh coarse would reduce the size of the matrix, but it has a secondary effect on the fill time. Suppose the structure is a patch which is 0.2X0.2mm, and it is meshed into only 4 rooftops--say it is a coarse mesh. The cell size is 0.1mmX0.1mm. Now if the enclosure box is 1000mmX1000mm, the program would still take a lot of time to calculate the 10000X10000 FFT in order to get the 16 matrix elements. This FFT also requires a lot of memory.
|
|
| Back to top |
|
|
