Discussion Forum

Posted: 1 decade ago 18 feb 2010, 02:36 GMT-5

Hi As I understand you are hitting one of the difficulties of representing "small" structures in detail, that end up in complex systems. COMSOL (as any other FEM programme) is not the limit, but its the RAM and processing speed to get a prepresentative model running. Clear, if you are out of RAM with 6 items, there is little chances that with 1000 it will be possible, like that, but: My only advice is to reconsider your model, what can be simplified ? using shell instead of bulk 3D, or ways with symmetry, can some periodic conditions help ? ... The limit is only our imagination, or rather the time availbale to let our imagination run. Unfortatey, I do not have ay better advice Good luck Ivar

Posted: 1 decade ago 19 feb 2010, 16:27 GMT-5

Good Day Prasad, If I might ask, what solver are you using? My understanding is that the direct solvers like Spooles or Paridso perform the solution entirely in memory. Using a linear iterative solver, such as conjugate gradient may extend your capability. Perhaps a slightly coarser mesh, at least at the outset, may be useful to demonstrate the model functionality. Best of luck, Marty

Posted: 1 decade ago 20 feb 2010, 01:56 GMT-5

Thanks for ur kind reply sir, i use linear solver only.also i use the mesh as extremely coarser,but then also it shows the error as out of memory.what can i do. sir i attach my project paper along with this reply.kindly give me a solution, thanking you sir

Attachments:

• rfmems.pdf

Posted: 1 decade ago 20 feb 2010, 11:14 GMT-5

Hi Are you simulating in full 3D or in 2D with a fixed thickness ? (the latter is easier to model and less computationally heavy), and have you considered dividing up your model, as if you have those tousends of small structures it is clear that your meshed device will require dozens of GB RAM, while you might get around by considering the overall shape and replace the different ranges of fingers by some simplified equivalent model. In anycase, you have a challenging and interesting project there Good luck Ivar

Posted: 1 decade ago 20 feb 2010, 19:56 GMT-5

I agree with the suggestions of everyone else. Simplifying the model (if possible) or trying something of the iterative solvers might be helpful. However, the iterative solvers can be a little trickier to ensure convergence. One further option that hasn't been mentioned is the type of OS you are using. If you are not currently running a 64 bit system, I would highly recommend switching to one if at all possible and/or adding more RAM to your computer. This will allow you to solve much more complicated models without running into the dreaded "Out of Memory" error. Good luck, Mike

Posted: 1 decade ago 9 mar 2010, 00:21 GMT-5

sir, thanks for your reply. but i use 2Dfor simulating the model.then also it shows the same error.

Posted: 1 decade ago 4 lug 2013, 03:01 GMT-4

[QUOTE] Hi sir, i am doing my project using comsol software.title was A NON-CONTACT TYPE RF MEMS SWITCH FOR 24GHz RADAR APPLICATIONS.here my switch is a capacitive shunt type switch which is actuated by combdrive actuators.i did a modal design but in that paper the switch requires 1000 comb fingers.my modal design having only 6 combfingers,even for this design also after giving the solve option it shows an error as out of memory during lu factorization.then how can i design a switch having 1000comb fingers.is it possible with comsole? also i need a graphs between isolation and frequency.how to draw using comsol software. thanking you sir [/QUOTE] Hi, I have been trying to find out the pull in voltage of a very basic cantilever based MEMS switch. I referred to the tutorial given in MEMS module for resonator pull in voltage calculation and tried to simulate a switch. However, I keep getting errors indication lack of convergence and NaN/Inf not found and so on. Were you successful in finding out the pull in voltage? I would appreciate it if you help me by giving some tips on this issue.