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 > ie3d edge meshing

ie3d edge meshing

时间:03-22 整理:3721RD 点击:
Hi,

When I use IE3D, I found when I consider all the metal thickness (3D metal) and mesh it. There is not much difference in the total cell number with or without edge cell. So can I say for 3D metal meshing we do not need to worry about the edge cell size?

For metal growing, there are "1st Degree Compensaton" and "2nd Degree Compensatoin with Pre-Processing"? which one is more accurate?

what is the difference between Contemporary and Classic Meshing scheme? Which one is better?

Sometime when I mesh the structure, I got some warning message "between ... and ... forced meshing cell size is smaller than minimum cell size". Can I ignore it?

Thank you.

danda

Hi, Danda821:

On IE3D, when you build the metal thickness, we basiclaly put some cells on the vertical sides, as well as the top and bottom sides. The vertical cells will allow current to flow vertically and horizontally. The horizontal edge current there already takes care of the edge condition very well. Normally, the simulation results are already very good and we do not suggest users to use edge cells. By default, the "classical" meshing scheme will not put edge cells there anyway.

For thick metal modeling, the "2nd degree compensation with pre-procesing" will yield more accurate results on modeling the loss. For modeing the structure effect of thickness, both schemes are the same.

The "classical" scheme is a legacy scheme. It has been used for 14 years and it is very robust. However, when your polygons are more arbitrarily shaped, the "classical" schemes may try to use all triangles without using rectangles on the irregular polygons. As it is explained in the manual or other places before, one rectangle is at least equivalent to two triangles. Using more rectangles and fewer triangles will reduce the unknowns and it may improve the efficiency (not accuracy) significantly. For this reason, we have introduced the "Contemporary" scheme in last year. The "Contemporary" scheme will normally create fewer cells for the same structure with the same density because it tries to mesh a structure into more rectangles and fewer triangles. However, the "Contemporary" scheme is still new. It is less stable than the "Classical" scheme. When it becomes mature, it will replace the "Classical" scheme eventually. Also, you can use "contemporary" scheme to create multiple non-uniform edge cells for extreme accuracy. You can also create edge cells on thickness trace for ultra high accuracy simulations.

IE3D allows you to draw a structure as a set of polygons. Each polygon is describes as a set of vertices. When you do the simulation, IE3D will try to mesh your structure into small rectangular and triangular cells. Whenever possible, it will use rectangles in order to reduce the number of cells with the same degree of density. In case it can not fit rectangular cells into the irregular regions, it will try to use triangular cells. Internally, we always use the drawn vertices as the starting point for the meshing. For example, if you draw 2 vertices to be very close or if your edge cell size is very small, the edge between these 2 vertices will be an edge of a cell (normally triangular cell). When the edge is very small, this edge will become very small compared to other edges with normal cell size. When this edge is approaching 0 compared to other edges, it is creating a "singlar" cell or a cell with the area approaching 0 while the circumference is not approaching 0. Eventually, this "singular" cell may cause numerical difficulty or instability.

Internally, before a simulation, we will find the minimum distance between any 2 vertices. If this distance is smaller than some artificially determined small value, we will give user a warning. The artificial limit is in fact to be some very small value (may be 1% of the regular cell size). If you see some message "... size is A (<B) ...", where A and B are some values, the A is normally the minimum distance between any 2 vertices and the B is normally the artificially determined limit. As I indicated before, normallly, if A is slightly smaller than B, you do not need to worry about it even it is a warning. Only when A is about 3 orders smaller than B, you may want to check the current distribution of the structure, if you see some "hot" spot or some place where the current is unreasonably dense compared to other locations, it may cause accuracy problem in the radiation pattern while the accuracy of the s-parameters may be slightly affected. However, in case A is about 4 to 5 orders smaller than B, it may affect the s-parameters significantly. For the worst case, it may cause singular matrix and the simulation can fail. Such situation seldom happens. In fact, in the last few years, I have not seen it happened once. However, I would say we can not avoid it completely because users can always use inserted vertices to control the meshing and we allow it. That is the reason why we don't ignore any vertices in polygons in the meshing because users may manually insert vertices to control the meshing for higher accuracy and efficiency. Normally, experienced users can achieve better combined performance in accuracy and efficiency by manually controlling the meshing.

I hope my comments will help you in understanding those messages. Regards.

Hi Jian :

Does "merge selected polygons" in IE3DZeland have any influence in simulated results or is only a "visual" tool?? The same geometry done with different polygons put together yields always to the same results?
Thanks in advance

I've almost finished my graduate, all of you are invited to the party ;)

HI, Jian

There are "Strong 3D" "Weak 3D" "Strong 3D Enhanced" "Weak 3D Enhanced" options in the simulator. Which one is sutable for on-chip 3D structure? Thank you.

danda

Hi, Javirulo:

On IE3D, we allow users to describe a structure using a set of polygons. Then, we will perform the meshing based upon the shape of each polygon and the interconnection between them. In some sense, we would suggest you to make your polygons as "rectangular" as possible. What I mean "rectangular" is that your strucutre have kinds of rectangular corners. Our meshing program will based upon the shapes and try to mesh your strucures into more rectangular cells (and fewer rectangular cells) for higher efficiency in the simulation. Normally, if possible, we suggest you to divide your polygons into more "rectangular" polygons and it will be help the meshing program much in optimizing meshing because normaly human beings are smarter than programs.

However, in some extreme cases, you or the program may have created a structure some strange polygons. In such a case, we will suggest you to merge those strange polygons together so that the simulator may be able to mesh better. For the contemporary meshing scheme, it is certainly much smarter in meshing a structure into more rectangular cells. Normally, simulation results are related to the meshing. For most cases, IE3D's results are normally insensitive to the meshing when some kinds of rules are followed (using AEC for coupled structures, using aligned meshing for closely coupled structures such as MIM capacitors, etc). Merged polygons will change the meshing of the structure in simulation. In some sense, there is a potential to change the results slightly while the results should normally be good if you follow the meshing rules.

You may ask when you will need to merge some polygons. Let me give you a simple case. You created a probe-fed patch antenna on IE3D. The patch will be automatically broken down into a set of polygons to guarantee the connection between the pach Later on, you want to change the feed. If you want to change the feed location, you can select all the related vertices (not only the vertices for the probe but also the related vertices of the broken down patch for better meshing) and move them to a new location. This is smple. However, if you want to change the radius of the probe, at this time, it is not simple unless you are using the IE3D MM08X with the AGIF capability. If you want to change the radius, you need to remove the probe and re-create it. How to remove the probe? Normally, a probe-fed antenna on IE3D will involve 3 layers: (1) negative level (normally at z = 0) called Zn. (2) positive level called Zp (at about 1% of the substrate thickness) (3) and the patch level called Za. What you can do is to select the vertices at Zp and delete them. Then, the probe and the port will be gone. Then, you can build your new probe with port using the ENTITY->PROBE-FEED TO PATCH command. The problem is that the divisions on patch antenna are still there and they may make the meshing of the antenna very messy. In such a sitution, you should select all the broken down polygons on the patch and merge them.

I hope my comments help you. Thanks!

Best regards,

Added after 28 seconds:

Hi, Danda:

On IE3D, we are able to solve true 3D metallic and dielectric structures as well as planar 3D structures. We are able to solve the current distribution on arbitrarily shaped 3D structures with variation in all the directions of the 3D structure. Normally, it takes more effort to solve full 3D problems.

In the IE3D 11, we improved the schemes both in accuracy and efficiency in modeling full 3D structures. We changed some default settings in it for higher accuracy. When we released it, we immediately received complaints from old customers. They found that the IE3D 11 was much slower than the IE3D 10 for some simple probe-fed antenna. It tooks 1-2 seconds per frequency point on IE3D 10 while it took 10 seconds per frequency point on IE3D 11 while the results were the same. They got used to the high speed of IE3D 10. For this reason, we revert back the settings on IE3D 11. We introduced the "Strong 3D" and "Weak 3D" options. Basiclally, the default "Weak 3D" option will make IE3D 11 the same speed of IE3D 10 because it is the same setting. The "Strong 3D" option will try to create fewer cells for strongly 3D structures while it will sacrifice speed in calculating the Green's functions. The meshing is changed. The "Strong 3D" option will normally create fewer cells so that the matrix size is smaller. However, the simulation time for the Green;s function is increaseed. For most cases, you will see little difference in accuracy (because of change in meshing) while you may see some speed differences between the two settings. Normally, the "Strong 3D" option will make it faster for large structures while the "Enhanced Weak 3D" option will make it faster for small structures. For mainly planar with 3D structures (vias, wire bonds, solder balls, PCBs, on-chip inductors, etc), the "enhanced weak 3D" option is perfect. Only when you have strong 3D structures with big span in the z-direction, the "Strong 3D" option is suggested.

Regards.

上一篇:a question about HFSS
下一篇:最后一页

Copyright © 2017-2020 微波EDA网 版权所有

网站地图

Top