simulation time problem in hfss

Hi everybody,

I am simulating a microstrip antenna, results are somehow good BUT the problem is the simulation time is long and I can't do parametric study with this. I use hfss13 and core i5 laptop 4G RAM. also the laptop screams at each simulation."high cpu usage " "high disk usage' "high memory usage".

what should I do?

Tnx

Don't use HFSS.

HFSS is an excellent piece of simulation software, but at the cost of consuming large amounts of system resources. If you are just simulating a basic microstrip patch antenna, I would recommend a different software solution, such as Agilent's Advanced Design System for planar structures.

Thank you PlanarMetamaterials 4 ur reply.

BUT I have to use hfss -no choice- and I'm working on a paper and its not that simple.
any suggestions to speed up my simulation?

REGARDS
israa

israa.k said:

Thank you PlanarMetamaterials 4 ur reply.

BUT I have to use hfss -no choice- and I'm working on a paper and its not that simple.
any suggestions to speed up my simulation?

REGARDS
israa

Click to expand...

To state the obvious - get a better computer. You are using a laptop, which by their very nature are more expensive for a given amount of computing resources than a desktop. If you are running out of RAM, that will kill you dead. I don't think CPU time will be such an issue for you. Hence you need to get something with more than 4 GB RAM. A cheap desktop will do you, but even a modern laptop with an i3 with 8 GB RAM would probably be better than your i5 with 4 GB. But personally I'd buy a desktop machine.

Laptops are not designed to run heavy CPU intensive programs for long periods. The heat will shorten the life of the battery. The high fan speed will shorten the life of the fan. If you can remove the battery whilst it is powered from the mains, I would do so.

You could also see if it's possible to disable any ofter software which is running. Things like anti-virus software, Skype, all use RAM, and you really want to disable as much of that sort of stuff as you can. Many programs are installed which start automatically when you boot the machine.

I have an i7 laptop with 8 GB RAM, but for serious use of HFSS, I use a Xeon based server. Servers have two big problems though.

* RAM for them is expensive. You can't use the cheap RAM you use in common PCs
* They are noisy.

If you are using the Optimetrics, and have a DSO license, then you are basically going to be analyzing multiple instances of the same structure at the same time. If that exhausts your RAM, then it would be better to just run single-threaded, and not use the multiple CPUs.


Deborah

Besides using a better computer, the best way to reduce simulation time and increase accuracy is to understand meshing. From my experience in EM solver support (not HFSS), new users tend to mesh the problem inefficiently. By understanding the mesh settings and using a good, approriate mesh, an experienced user can often reduce the memory requirement and simulation time by a large amount (10x or more).

DeboraHarry,
I can't thank u enough for the detailed information, it's just exactly all I wished to know.

volker_muehlhaus
Thank u, ur advice hit me.


israa :razz:

As volker_muehlhaus said, meshing is important.

I posted some material from an Advanced HFSS course here some time back.

https://www.edaboard.com/threads/258213/

There is quite a bit in there about meshing.

That said, I would not say a patch antenna is exactly stretching HFSS. If you post your .hfss file, I'll simulate it and tell you the memory usage. That might give you some idea if you are close to your 4 GB laptop's limit, or miles away.

FWIW, here is some environmental data, including fan speeds, of a IBM x3650 server running flat out doing CPU intensive things. This has a pair of quad core Xeons and 48 GB RAM. Note the fan running the fastest is only running at 58% of full capacity. I expect the fan in your poor laptop is running flat out. Also, the fans in this server are hot-swappable, so If one fails, one can replace it whilst the computer keeps running. I think IBM say up to two fans can fail at any one time with no ill effect on the computer. In fact, given it ships with only 5 fans, and another 5 are added if a second PSU is added, I suspect far more than two fans could fail unless they were all in the same area of the machine.

In contrast, if a fan fails in your laptop, your laptop will stop working and you will probably have to send your laptop back to get the fan replaced. Really, although I have done it myself, and will continue to do it, a laptop is not an ideal computer for serious HFSS.

Hi DeboraHarry,

I'd love to, if you give ur e-mail I'll send it to u

Tnx

israa.k said:

Hi DeboraHarry,

I'd love to, if you give ur e-mail I'll send it to u

Tnx

Click to expand...

You can attach the file on the forum - then anyone can analyze it and perhaps comment on how you could make it use less RAM. But if you want to send it to me personally, send me a private message and I'll give you my email address.

Deborah

I'll send u a private msg.

I want to be more specific here.
usually the first run of the simulation or the run after a geometry modification doesn't bother me usually acceptable. however, when I change the analysis setup or sweep and then run the simulation or do successive simulations of the same geometry the problem becomes rather complex and I don't know why and how to avoid it.

could anyone help?

israa.k said:

I'll send u a private msg.

I want to be more specific here.
usually the first run of the simulation or the run after a geometry modification doesn't bother me usually acceptable. however, when I change the analysis setup or sweep and then run the simulation or do successive simulations of the same geometry the problem becomes rather complex and I don't know why and how to avoid it.

could anyone help?

Click to expand...

If you have an HPC or DSO license, and are using parallel processing in an attempt to speed up your simulations, you might actually be slowing them down! Parallel processing will use extra memory. If the memory usage increases to the point where you are swapping to disk, then doing things in parallel will probably grind the system to a halt.

There is also the option of the itterative solver. That uses less RAM, at the expensive of CPU usage. I forgot about that one before, but it might help. Send me the file you have simulated, before making any changes to it, and I'll take a look