1. Reduce number of running process while using HFSS. It will give you more RAM for computation.

2. Design the structure optimally, it will reduce the number of tetrahedron.

3. Try to draw the HFSS structure as per practical scenario, besides, consider the theoretical aspect. For example, you you draw coaxial probe having the rad of inner conductor r = 0.63mm and relative permitivity = 2.2 (teflon), it will create large no of tetrahedron. Try to draw the coaxial connector using r = 0.63mm and er = 1, it will create less number of tetrahedron.

Theory is: design 50 ohm coaxial cable. Similar facts have to be implemented while designing any structure in HFSS. Main objective: less number of tetrahedron.

1-Simplify model for reducing mesh size

2-Add more RAM

3- Use mixed order + curvilinear meshing options instead of 2nd order linear(no curvilinear).

> Reduce number of running process while using HFSS

- it is ineffective. Most windows processes are swapped to disk while an active process is running. HFSS reserves specifically physical RAM, so WIndows/Linux automatically frees the requested amount.

Replacing 3D microtrip models with simple planes, layered impedance applied, helps greatly reducing RAM requirements, and computation time.

Take advantage of symmetry in your structure and use perfect E plane and perfect H plane boundries to divide the model solution time by the number of symmetries found. In the chapter about Assigning Symmetry Boundaries this begins to explain the method and work flow.

Try to simplify your model using PE-PM boundaries along symmetries as suggested. Using very coarse meshes is also an option but your results may not be very accurate. Alternatively, you may want to get a computer with bigger memory.

Yes, the geometry can be simplified using symmetry by applying PEC and PMC. In that case, to obtain general solution, we have to simulate he structure for all possible combinations by changing the position of PMC and PEC.

For example, plz, see the theoretical development on Ridge Waveguide.

-I think you just extend your RAM memory as I have also faced the same situation and the only solution is to get RAM extended.

-you can also try to decrease the number of mesh cells. that also lowers the memory requirement

RAM can help to solve the problem, but it is our duty to know the actual error. In future, you may get same issue and you can further extend the RAM.

This is not the solution. 6GB RAM is huge for academic work. Try to draw the HFSS structure efficiently. I prefer to use HFSS manual v10. All basic things are there. No advance process, hence less complex. As a result of that, less complexity in generating mesh. This will take less time.

Thanks all Respected researchers,

Overall I found that only memory requirements are more to simulate.

Please you contribute ur views how much memory ( RAM and HDD is sufficient) for proper simulation.

One easy way is to use the iterative solver or domain-decomposition method instead of the direct solver in the setup. This will cost more time or more storage space, but it requires less memory. This method does not impair the accuracy considerably, according to my models. But reducing meshes may cause problems.

If the error occurs while sweeping, try fewer sampling points.

In my humble opinion, it is better to follow the steps suggested by the experienced researches. However, you might face the same problem in the future; if it is unavoidable, you might upgrade your RAM to 12GB (or might be 16GB depending upon budget and PC hardware requirements). 1 TB HDD is enough, also HDD doesn't have much impact on the simulation performance (might be SSD have higher simulation performance).

Selecting the optimal Mesh size also reduces unnecessary memory usage. If server PC available then better to use that.

Raja Usman sir, I understand the thing that increasing memory is not all time solution,

Sir I don't know how to reduce mesh size, kindly suggest this stepwise

In HFSS...

1) Click Setup

2) Click on Driven Solution Setup tab

3) Click on Initial Mesh Options.

You can try to change the Mesh, I hope it will work.

Also kayode oyesina sir please explain simple way to how to simplify PE PM boundaries all ng symmetries. It would be more helpful then.

Thanks respected researchers for your response.

6 GB RAM is not very much. I use 32 GB.

For some methods RAM requirements scale as the sixth power of size in wavelengths (the square of the number of elements). Double frequency = 64 times more memory. Even the most economical methods scale as fourth power of size. Do everything you can to use the lowest number of elements that will give a valid answer.

16GB for a typical lab research budget. I managed with 8GB. If the solver demands, that's because your model and solver generate a large mesh. Example of situation may be a model with a wire antenna on a warship. One should therefore consider employing a different numerical solver like Feko. It is too restrictive for EM engineers to just stick to one particular solver like FEM. I would have a time domain solver as well. You will find choosing the right solver for your problem usually avoid memory issues. For me the frequency of the memory is of more importance than the size of the memory.

Best is to reduce the mesh size or to use domain decomposition method so that u can utilize efficiently your memory

1 u can increase RAM

2 u can reduce mesh sample frequency passes

3 one more u should go to the folder where ur projects are saved u find there lots of files with project files which are created when u simulate ur project and u can delete them because when u simulate again these files will be created again so u can delete these files and can increase memory space

Ok thanks Malcolm white sir,Simon Tse sir,Ajit Kumar sir and Ashwii Kumar sir for ur suggestions.

I think you should minimize your boundary box.

Greeting of the day !!

You can try these probable solution,

1. Optimize your design.

2. Try to design in a practical way keeping all theoretical aspects. You can modify 3D microstrip structure with simple plane. This will help to reduce the memory.

3. Reduce sample frequency passes.

4. Reduce frequency sweep intervals and range.

5. Optimize radiation box.

6. Reduce background running process while using HFSS.

7. Increase RAM to minimum 16GB. ( 16GB RAM and 1TB HDD is sufficient).

Best Regards

Anik Ghosh

Ok thanks Sajjad sir n Anik sir for your suggestions

I am also getting the same problem. What to do??

Solutions are already provided above. In summary,

(1) Increase RAM (if possible)

(2) Reduce the number of tetrahedron in your structure. this step is extremely important to get the results quickly from a complicated geometry.

Enable (if you did not enabled it before) and try to increase "Swap fle" in Windows. Don't be shy, set the swaf file, for exaple, to 15-20 GB, or even more, if needed. It should help.

Hfss generates a simulation folder. Try to delete that folder.

After you have done all the things mentioned above and you still get "out of memory", you can get an answer by reducing the number of iterations that you try and/or increase the Delta S you set. Your resolution may be a bit less but often it is still good enough to be useful. Often these settings are more precise than need to be to get a useful answer.