Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Coupling Electrostatic & Electrical Circuit into Solid mechanics

Please login with a confirmed email address before reporting spam

Please,
I am trying so hard to get the frequency response from my resonator model with electrostatic simulation in Comsol4.2 but still no luck. I need to apply sinusoidal oscillations from AC source and DC voltage to my model in order to generate electrostatic actuator . Electric field causes an electrostatic force which induces the changes in gap . The displacement of the resonator dependent in time induces the vibrations with different frequencies. I tried so many ways but it didn’t work. I can’t figure out how to couple electrostatic(es) & electrical circuit (cir) into solid mechanics(solid) . I would be very grateful if anyone can help me solve this problem. I attached my model here.

Many thanks.


9 Replies Last Post 23 juin 2012, 11:47 UTC−4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 19 juin 2012, 03:08 UTC−4
Hi

I have a few suggestions:
you have a nice symmetric design, so why not start with 1/8 and then mirror out, this will give you some internal boundaries, but these are good as it will improve the symmetry of your mesh and your design is symmetric, no reason to reduce the response by an asymmetric mesh.

Then make your attachment pads by a union and remove the "keep internal boundaries" tick, you can also make the volumes less interfere by doing some boolean operations like differences. You should also split your small beams as the stress build-up in these regions is critical for our survivability and as it is now you will get naturally a large mesh in this region

When you fix the side pads be sure you fix only the true fixed squares and not everything around, including the beam going to your "flag"

When solving, make a first run with ONLY the eigenfrequency, then get the first mode and make a second study for the frequency domain around the first resonances, I see one around 23kHz but you are studying your device around 21[MHz] why this 1k difference ? if that is correct you need to check the wavelength of your modes and the relative meshing density too

I do not see why you use a CIR physics, you can as well imply a sinus source by an equation directly into ES. And do not forget to add the domain Force node to extract forces you can apply as boundary or domain loads on you structural analysis.

Finally your "spring foundation" is not clear for me what it represents bu it xould come from other features of your model

--
Good luck
Ivar
Hi I have a few suggestions: you have a nice symmetric design, so why not start with 1/8 and then mirror out, this will give you some internal boundaries, but these are good as it will improve the symmetry of your mesh and your design is symmetric, no reason to reduce the response by an asymmetric mesh. Then make your attachment pads by a union and remove the "keep internal boundaries" tick, you can also make the volumes less interfere by doing some boolean operations like differences. You should also split your small beams as the stress build-up in these regions is critical for our survivability and as it is now you will get naturally a large mesh in this region When you fix the side pads be sure you fix only the true fixed squares and not everything around, including the beam going to your "flag" When solving, make a first run with ONLY the eigenfrequency, then get the first mode and make a second study for the frequency domain around the first resonances, I see one around 23kHz but you are studying your device around 21[MHz] why this 1k difference ? if that is correct you need to check the wavelength of your modes and the relative meshing density too I do not see why you use a CIR physics, you can as well imply a sinus source by an equation directly into ES. And do not forget to add the domain Force node to extract forces you can apply as boundary or domain loads on you structural analysis. Finally your "spring foundation" is not clear for me what it represents bu it xould come from other features of your model -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 20 juin 2012, 07:47 UTC−4
Dear Ivar,
Thank you very much for your quick reply . I’m sorry that I read very carefully but I’m a newbie to Comsol so some paragraph you have written are still not clear for me.


you have a nice symmetric design, so why not start with 1/8 and then mirror out, this will give you some internal boundaries, but these are good as it will improve the symmetry of your mesh and your design is symmetric, no reason to reduce the response by an asymmetric mesh.

Then make your attachment pads by a union and remove the "keep internal boundaries" tick, you can also make the volumes less interfere by doing some boolean operations like differences. You should also split your small beams as the stress build-up in these regions is critical for our survivability and as it is now you will get naturally a large mesh in this region


If I’m not mistaken, you were talking about ALE module, weren’t you? Is it need to put ALE into this model?Ale is another hitch to me. I have no idea about ALE although I’ve heard that ale maybe needed in electrostatic simulation. I have not found any documentation which has a close state about ale.

23kHz in eigenfrequency makes no sense to my study. About frequencies domain analysis , I am studying my device around 21Mhz because at that frequency , I saw the desired resonant shape is matched with the result in theory . But unfortunately , the resonant frequency in theory is 2.1Mhz which doesn’t correspond with the desired resonant shape. This plate resonator works in square-extensional mode because of its side length .

“When you fix the side pads be sure you fix only the true fixed squares and not everything around, including the beam going to your "flag"”
--
Finally your "spring foundation" is not clear for me what it represents bu it xould come from other features of your model

Excuse me, I guess you mean I fixed the T-stems and that’s wrong. I removed the spring foundation, I thought I should add the effective spring constant. Thank you for your correction. But once again, it’s strange that if I do the same simulation with the plate resonator works in wine-glass mode(in attached file) , the first resonant frequency in eigen is matched resonant shape in theory so everything’s okay with that model. I even fix 4 corners without drawing the structure of anchors and T-stems. I also check the relative mesh density , there’s not much different meshing between 2 models.

Return to the electrostatic simulation, the most important thing to me now and I need your help. With your instructions, I guess what I need to do is:
- First, I imply a sinus source by an equation directly into ES. Simultaneously, add the boundary loads on my structural analysis . The matter makes me confused here is Comsol will calculate the electrostatic force or I have to do it by myself? If Comsol do it, how to indicate to boundary loads the total force which is generated when I applied the sinus oscillation in ES ? If I make it right, will I get the frequency response plots such as frequency response versus V applied, Order-of-magnitude by frequency or Electrostatic force versus V applied ?
- I need to put all the model in the air as dielectric environment .
- Finally, if I don’t success with electrostatic simulation, I have to directly excite the force to resonator by using single module solid mechanics. To do this, I need to turn on damping so would you please tell me the structural loss factor of Silicon or Silicon single crystal. Some documentary I have read shows so many kinds of damping but I still can’t find the exact number of this property of Silicon.

Many thanks once again.
Dear Ivar, Thank you very much for your quick reply . I’m sorry that I read very carefully but I’m a newbie to Comsol so some paragraph you have written are still not clear for me. [QUOTE] you have a nice symmetric design, so why not start with 1/8 and then mirror out, this will give you some internal boundaries, but these are good as it will improve the symmetry of your mesh and your design is symmetric, no reason to reduce the response by an asymmetric mesh. Then make your attachment pads by a union and remove the "keep internal boundaries" tick, you can also make the volumes less interfere by doing some boolean operations like differences. You should also split your small beams as the stress build-up in these regions is critical for our survivability and as it is now you will get naturally a large mesh in this region [/QUOTE] If I’m not mistaken, you were talking about ALE module, weren’t you? Is it need to put ALE into this model?Ale is another hitch to me. I have no idea about ALE although I’ve heard that ale maybe needed in electrostatic simulation. I have not found any documentation which has a close state about ale. 23kHz in eigenfrequency makes no sense to my study. About frequencies domain analysis , I am studying my device around 21Mhz because at that frequency , I saw the desired resonant shape is matched with the result in theory . But unfortunately , the resonant frequency in theory is 2.1Mhz which doesn’t correspond with the desired resonant shape. This plate resonator works in square-extensional mode because of its side length . [QUOTE] “When you fix the side pads be sure you fix only the true fixed squares and not everything around, including the beam going to your "flag"” -- Finally your "spring foundation" is not clear for me what it represents bu it xould come from other features of your model [/QUOTE] Excuse me, I guess you mean I fixed the T-stems and that’s wrong. I removed the spring foundation, I thought I should add the effective spring constant. Thank you for your correction. But once again, it’s strange that if I do the same simulation with the plate resonator works in wine-glass mode(in attached file) , the first resonant frequency in eigen is matched resonant shape in theory so everything’s okay with that model. I even fix 4 corners without drawing the structure of anchors and T-stems. I also check the relative mesh density , there’s not much different meshing between 2 models. Return to the electrostatic simulation, the most important thing to me now and I need your help. With your instructions, I guess what I need to do is: - First, I imply a sinus source by an equation directly into ES. Simultaneously, add the boundary loads on my structural analysis . The matter makes me confused here is Comsol will calculate the electrostatic force or I have to do it by myself? If Comsol do it, how to indicate to boundary loads the total force which is generated when I applied the sinus oscillation in ES ? If I make it right, will I get the frequency response plots such as frequency response versus V applied, Order-of-magnitude by frequency or Electrostatic force versus V applied ? - I need to put all the model in the air as dielectric environment . - Finally, if I don’t success with electrostatic simulation, I have to directly excite the force to resonator by using single module solid mechanics. To do this, I need to turn on damping so would you please tell me the structural loss factor of Silicon or Silicon single crystal. Some documentary I have read shows so many kinds of damping but I still can’t find the exact number of this property of Silicon. Many thanks once again.


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 20 juin 2012, 16:51 UTC−4
Hi

No first of all I'm talking about the geometry, ALE might be an issue when linking to the ES, later. The thing I noticed is that your geometry is very symmetric, you could draw only 1/8 in a workplane, extrude it and then mirror it 3 times. This could give you a few advantages: your meshing will be slightly more symmetric, your thin beams are split so you have naturally more elements in these critical high stress regions and you can start to model only 1/8 which solves must faster, at least until your are happy with all model set-ups ,then yuo can move to the full model, to avoid anaysing all combinations of symmetric and antisymmetric BCs.

The first models I see are at 23kHz, but clearly you have also modes much higher up, but these are surely in plane pressure modes and not bending modes (or a combination of). Then at 21 MHz, what is the wavelength expected ? and do you have at least 10 mesh elements per wavelength period in your model, for thsi frequency ? If not how can you expect to resolve the modes then ?

And if you wnat to do an eigenfrequency analysis around 21 MHz and skip the first mods, you need to tweak your solver accordingly with a "search around 21 MHz frequency range setting

Spring constant, which one I cannot understand where you "spring fundation" comes from, but you might have some "hidden features" in your model I cannot know about".
For the fixed parts, indeed I talk about the T's in the 4 vertexes, I assume these are only fixed on some faces and not on the full domains ?

You have a Force node that make ES to calculate the force applied to the domain selected, this force can then be used in the Structural physics as boundary loads.

You can define a port and apply a sinus voltage thereon, and do a time series solver, but you will have to do many time steps and wait for long, I suppose you could use a frequency domain sweep (if ES is set up for that ? if not have you checked EC ?)
From my understanding, at these frequencies your ES induced displacements are very small, so I assume you can solve first for the ES to get forces and then use the forces as a harmonic load onto the structure (assuming the deformations of the structure are small and would not couple back and change the force values again. (But here I might be wrong depends on your model and what you are really looking for, which is not yet 100% clear for me :)

Indeed for ES or EC you need the air domain all around to allow the E field to be defined.

Damping is a delicate issue, whcih value to use ? material damping is little known and difficult to find in the litterature, and small objects like this at high frequency have damping effects from several sources, including air film damping. You can add a damping sub node to the structural elastic material physics node, then choosing an isotropic damping factor, but if thsi is crystalline Silicon, I would use < 0.001, but the other air ... effects might be signficantly higher (and will be very mode shape dependent

--
Good luck
Ivar
Hi No first of all I'm talking about the geometry, ALE might be an issue when linking to the ES, later. The thing I noticed is that your geometry is very symmetric, you could draw only 1/8 in a workplane, extrude it and then mirror it 3 times. This could give you a few advantages: your meshing will be slightly more symmetric, your thin beams are split so you have naturally more elements in these critical high stress regions and you can start to model only 1/8 which solves must faster, at least until your are happy with all model set-ups ,then yuo can move to the full model, to avoid anaysing all combinations of symmetric and antisymmetric BCs. The first models I see are at 23kHz, but clearly you have also modes much higher up, but these are surely in plane pressure modes and not bending modes (or a combination of). Then at 21 MHz, what is the wavelength expected ? and do you have at least 10 mesh elements per wavelength period in your model, for thsi frequency ? If not how can you expect to resolve the modes then ? And if you wnat to do an eigenfrequency analysis around 21 MHz and skip the first mods, you need to tweak your solver accordingly with a "search around 21 MHz frequency range setting Spring constant, which one I cannot understand where you "spring fundation" comes from, but you might have some "hidden features" in your model I cannot know about". For the fixed parts, indeed I talk about the T's in the 4 vertexes, I assume these are only fixed on some faces and not on the full domains ? You have a Force node that make ES to calculate the force applied to the domain selected, this force can then be used in the Structural physics as boundary loads. You can define a port and apply a sinus voltage thereon, and do a time series solver, but you will have to do many time steps and wait for long, I suppose you could use a frequency domain sweep (if ES is set up for that ? if not have you checked EC ?) From my understanding, at these frequencies your ES induced displacements are very small, so I assume you can solve first for the ES to get forces and then use the forces as a harmonic load onto the structure (assuming the deformations of the structure are small and would not couple back and change the force values again. (But here I might be wrong depends on your model and what you are really looking for, which is not yet 100% clear for me :) Indeed for ES or EC you need the air domain all around to allow the E field to be defined. Damping is a delicate issue, whcih value to use ? material damping is little known and difficult to find in the litterature, and small objects like this at high frequency have damping effects from several sources, including air film damping. You can add a damping sub node to the structural elastic material physics node, then choosing an isotropic damping factor, but if thsi is crystalline Silicon, I would use < 0.001, but the other air ... effects might be signficantly higher (and will be very mode shape dependent -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 juin 2012, 12:59 UTC−4
Dear Mr.Ivar,

I really appreciate your kindly help. Your detailed explanation helps me so much. I wish Comsol could be friendly like you. I’ve accomplished simulation with each module one by one. When I ran Time dependent analysis for only electrostatic simulation with Vbias parametric sweep, I got the plot of Vbias versus Electromagnetic force. And when I ran frequency domain analysis for only tructural mechanics, I got the plot of frequency versus displacement. With single simulation, I can’t get the plots which represent the relationship between frequency response and Vapplied, or order-of-magitude versus frequency with different Vbias applied as some kind of plots I attached.
I’ve also tried to simulate by coupling es into structural with your instructions. I gave electrostatic force a name in Force calculation node and then used it in the Structural as boundary load but when I computed Time dependent analysis , this error occurred:
Failed to evaluate variable Jacobian.
- Variable: Fel
- Geometry: 1
- Boundary: 22
then I ran Frequency domain analysis, Comsol says
Failed to evaluate variable.
- Variable: Fel
- Geometry: 1
- Boundary: 22 27Failed to evaluate expression.
- Expression: d((Fel*test(mod1.u)*mod1.solid.surfScale/mod1.solid.vol_bndl1)*(dvol),{test@6})

In eigenfrequency analysis, I had done as your suggestion before. I tweak solver accordingly with a "search around 21 MHz frequency range” setting but the visual mode shape I observe is not the resonant mode shape as expectation so then I use frequency domain analysis. That’s the only model which has no compability between results from 2 kinds of study. Maybe my PC’s configuration can’t handle with a large amount of finite elements due to the structure of model. I remember that model compute too long.
P/s: I have no idea what BCs represents :D

Many thanks to you once again, Ivar. I will try to solve it and write to you if I succeed.
Sincerely,

Dear Mr.Ivar, I really appreciate your kindly help. Your detailed explanation helps me so much. I wish Comsol could be friendly like you. I’ve accomplished simulation with each module one by one. When I ran Time dependent analysis for only electrostatic simulation with Vbias parametric sweep, I got the plot of Vbias versus Electromagnetic force. And when I ran frequency domain analysis for only tructural mechanics, I got the plot of frequency versus displacement. With single simulation, I can’t get the plots which represent the relationship between frequency response and Vapplied, or order-of-magitude versus frequency with different Vbias applied as some kind of plots I attached. I’ve also tried to simulate by coupling es into structural with your instructions. I gave electrostatic force a name in Force calculation node and then used it in the Structural as boundary load but when I computed Time dependent analysis , this error occurred: Failed to evaluate variable Jacobian. - Variable: Fel - Geometry: 1 - Boundary: 22 then I ran Frequency domain analysis, Comsol says Failed to evaluate variable. - Variable: Fel - Geometry: 1 - Boundary: 22 27Failed to evaluate expression. - Expression: d((Fel*test(mod1.u)*mod1.solid.surfScale/mod1.solid.vol_bndl1)*(dvol),{test@6}) In eigenfrequency analysis, I had done as your suggestion before. I tweak solver accordingly with a "search around 21 MHz frequency range” setting but the visual mode shape I observe is not the resonant mode shape as expectation so then I use frequency domain analysis. That’s the only model which has no compability between results from 2 kinds of study. Maybe my PC’s configuration can’t handle with a large amount of finite elements due to the structure of model. I remember that model compute too long. P/s: I have no idea what BCs represents :D Many thanks to you once again, Ivar. I will try to solve it and write to you if I succeed. Sincerely,


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22 juin 2012, 08:37 UTC−4
Hi

you are going too quick, you haven't got all the subtleties of COMSOL yet, it takes some time I agree there is much to learn.

your ES seems OK, the SOLID almost, normally one would rather fix the boundaries of the block and not the full block, furthermore, the transition from the fixed block to the bean is "without fillets" and here you will get stress concentration points/edges, and you need several mesh elements in the thickness to correctly resolver your beam

The your boundary load the variable is not only "Fel" but (es.Forcex_Fel, es.Forcey_Fel, es.Forcez_Fel) and its not really correct to apply it only on one side, I would rather split it on each boundary opposed to the electrodes

Then what really need some improvements are your study nodes, here you need to read the doc a few times and try out simpler first. you cannot just dump in all parametric time series frequency domain into one node, they interact interfere and mix the outcome like that. Make 2-3 studies with fist a time series, then a frequency domain but separate, then the parametric on the adequate solver sequence etc.

if you chose time series you need to define the time varying voltage, but then you need also to use a time stepping in your MHz and ensure that all inertial transient effects settles (and that your mesh resolves the models you will excite) ...
I prefer to use harmonic developments but if you check the ES physics you will see it does not accept harmonic developments (which makes sense as it's a ElectroSTATIC, and not a DYNAMIC model. You need to consider EC instead. In fact the EMI physics from the MEMS module I suspect is better suited, with a pre stressed Frequency domain, which allows to use a VDc for the pre stress and then a Vac load for the frequency harmonic behaviour

Finally, are you sure you have 7% loss in your silicon beam ? I would have expected 0.1 to 0.01% but that depends how its "attached" too, anyhow your first mode is around 1.6 MHz here

--
Good luck
Ivar
Hi you are going too quick, you haven't got all the subtleties of COMSOL yet, it takes some time I agree there is much to learn. your ES seems OK, the SOLID almost, normally one would rather fix the boundaries of the block and not the full block, furthermore, the transition from the fixed block to the bean is "without fillets" and here you will get stress concentration points/edges, and you need several mesh elements in the thickness to correctly resolver your beam The your boundary load the variable is not only "Fel" but (es.Forcex_Fel, es.Forcey_Fel, es.Forcez_Fel) and its not really correct to apply it only on one side, I would rather split it on each boundary opposed to the electrodes Then what really need some improvements are your study nodes, here you need to read the doc a few times and try out simpler first. you cannot just dump in all parametric time series frequency domain into one node, they interact interfere and mix the outcome like that. Make 2-3 studies with fist a time series, then a frequency domain but separate, then the parametric on the adequate solver sequence etc. if you chose time series you need to define the time varying voltage, but then you need also to use a time stepping in your MHz and ensure that all inertial transient effects settles (and that your mesh resolves the models you will excite) ... I prefer to use harmonic developments but if you check the ES physics you will see it does not accept harmonic developments (which makes sense as it's a ElectroSTATIC, and not a DYNAMIC model. You need to consider EC instead. In fact the EMI physics from the MEMS module I suspect is better suited, with a pre stressed Frequency domain, which allows to use a VDc for the pre stress and then a Vac load for the frequency harmonic behaviour Finally, are you sure you have 7% loss in your silicon beam ? I would have expected 0.1 to 0.01% but that depends how its "attached" too, anyhow your first mode is around 1.6 MHz here -- Good luck Ivar

Bjorn Sjodin COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22 juin 2012, 15:55 UTC−4
The MEMS Module in the version 4.3 release comes with a built-in user interface called Electromechanics. Using the Electromechanics user interface for modeling this type of device will be much easier than if you try to combine the physics yourself. Please see:
www.comsol.com/products/4.3/#mems
for more information.
The MEMS Module in the version 4.3 release comes with a built-in user interface called Electromechanics. Using the Electromechanics user interface for modeling this type of device will be much easier than if you try to combine the physics yourself. Please see: http://www.comsol.com/products/4.3/#mems for more information.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23 juin 2012, 11:07 UTC−4
Dear Mr.Ivar,

In according with what you say, there are so many problems to my simulation. The only thing I had figured out and fixed before your reply is reference name & boundary selections of electrostatics force on boundary loads. Then it’s been solved but something appears in the plots electrostatics force vs time (with Vdc sweep) makes me wonder. ES physics does not accept harmonic developments bcuz it’s a static model as you said . Force doesn’t vary vs time(that means it doesn’t include AC sinus) but if I just plot a graph Fel vs time without Vdc, Fel is a sinus oscillation…? At least, I can plot the graph which demonstrates the relation between Displacement vs Freq (with Vdc sweep) that only available for coupling 2 module simulation. I’m sorry that I’m not clear where EMI physics comes from.. My version at present does not have a separate MEMS module.

“Then what really need some improvements are your study nodes, here you need to read the doc a few times and try out simpler first. you cannot just dump in all parametric time series frequency domain into one node, they interact interfere and mix the outcome like that. Make 2-3 studies with fist a time series, then a frequency domain but separate, then the parametric on the adequate solver sequence etc.”

I’m not sure that I really get what you mean here. If I made study1 with time series first then a freq domain in study2, all the results in study2 are zero. But when I tried a time series first, then I add freq domain as step 2 of this study then solve it without clear all the job and configuration before (like I usually do), I vaguely guess what you mean with the word ‘separate’.. I saw the results are different from the results if I solve 2 steps of study at the same time. There I got an add-on graph displacement vs time. When I add another step, Comsol will solve the first step with update information from the second due to the stored information before, won’t it? Comsol only solve the second analysis when I clear all the job and configuration before?


If you chose time series you need to define the time varying voltage, but then you need also to use a time stepping in your MHz and ensure that all inertial transient effects settles (and that your mesh resolves the models you will excite) ...

I’m confused with “time stepping in your MHz” . In time-dependent, I set a range(0,1e-7,1e-6) Is it fitting? Which doc I have to read? There’s a lot documentation in doc folder but they are separate or mixed together or mutual reference so sometimes I lost in that maze. I’ve just read Introduction to SME and Introduction to Mems module users guide. They show how to plot frequency response vs displacement but don’t specific how to deal with dynamic model . I see so many difficulties in time-dependent study here so I should stop using this way. Unfortunately, since yesterday my PC often crash the program when computing & restart after 2 day run contineous simulations . I know there’re so many mistakes still exist in my model and simulation but I think if I kept using ver4.2 to simulate , I would consider EC as your suggestion or accept the results so far.
As you said, Damping is a delicate issue so I’m not sure about 7% loss. It’s just a guess.

So many thanks to your nice reply . You are my life-saver.

Sincerely,


Dear Mr.Ivar, In according with what you say, there are so many problems to my simulation. The only thing I had figured out and fixed before your reply is reference name & boundary selections of electrostatics force on boundary loads. Then it’s been solved but something appears in the plots electrostatics force vs time (with Vdc sweep) makes me wonder. ES physics does not accept harmonic developments bcuz it’s a static model as you said . Force doesn’t vary vs time(that means it doesn’t include AC sinus) but if I just plot a graph Fel vs time without Vdc, Fel is a sinus oscillation…? At least, I can plot the graph which demonstrates the relation between Displacement vs Freq (with Vdc sweep) that only available for coupling 2 module simulation. I’m sorry that I’m not clear where EMI physics comes from.. My version at present does not have a separate MEMS module. [QUOTE] “Then what really need some improvements are your study nodes, here you need to read the doc a few times and try out simpler first. you cannot just dump in all parametric time series frequency domain into one node, they interact interfere and mix the outcome like that. Make 2-3 studies with fist a time series, then a frequency domain but [B]separate[/B], then the parametric on the adequate solver sequence etc.” [/QUOTE] I’m not sure that I really get what you mean here. If I made study1 with time series first then a freq domain in study2, all the results in study2 are zero. But when I tried a time series first, then I add freq domain as step 2 of this study then solve it without clear all the job and configuration before (like I usually do), I vaguely guess what you mean with the word ‘separate’.. I saw the results are different from the results if I solve 2 steps of study at the same time. There I got an add-on graph displacement vs time. When I add another step, Comsol will solve the first step with update information from the second due to the stored information before, won’t it? Comsol only solve the second analysis when I clear all the job and configuration before? [QUOTE] If you chose time series you need to define the time varying voltage, but then you need also to use a time stepping in your MHz and ensure that all inertial transient effects settles (and that your mesh resolves the models you will excite) ... [/QUOTE] I’m confused with “time stepping in your MHz” . In time-dependent, I set a range(0,1e-7,1e-6) Is it fitting? Which doc I have to read? There’s a lot documentation in doc folder but they are separate or mixed together or mutual reference so sometimes I lost in that maze. I’ve just read Introduction to SME and Introduction to Mems module users guide. They show how to plot frequency response vs displacement but don’t specific how to deal with dynamic model . I see so many difficulties in time-dependent study here so I should stop using this way. Unfortunately, since yesterday my PC often crash the program when computing & restart after 2 day run contineous simulations . I know there’re so many mistakes still exist in my model and simulation but I think if I kept using ver4.2 to simulate , I would consider EC as your suggestion or accept the results so far. As you said, Damping is a delicate issue so I’m not sure about 7% loss. It’s just a guess. So many thanks to your nice reply . You are my life-saver. Sincerely,

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23 juin 2012, 11:10 UTC−4
Dear Bjorn Sjodin,

Thank you for your suggestion. Actually, Comsol is not prevalent in my country although it’s an intensive FEM software. It’s not easy for me to approach or possess new version quickly.. Do you know where I can get or buy a trial version?

Sincere,
Dear Bjorn Sjodin, Thank you for your suggestion. Actually, Comsol is not prevalent in my country although it’s an intensive FEM software. It’s not easy for me to approach or possess new version quickly.. Do you know where I can get or buy a trial version? Sincere,

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 23 juin 2012, 11:47 UTC−4
Hi

I would not say its that many "problems" with your model, it's just that COMSOL is allowing to simulate almost all physics, and that is in fact quite complex. And you are jumping in directly into multi-physics with coupling

ES does not have frequency domain, but what you can do is to make a model based ONLY on ES, then use a Parameter "Param" and ramp your voltage on the electrode as V0*Param, and set the advanced continuation tab of your STATIONARY solver to Param and range(0,0.1,1)

Then you can plot the Static force versus voltage.

In the next model you can simulate SOLID ONLY, and use a boundary force on each side of the half of the ES calculated force, and here you can use a frequency domain sweep.

This is not as a fully coupled model, as the force is probably not linear with the voltage, and the force will deform the beam, hence the voltage induced force would also change in a fully coupled mode.

You would calculate the effects when working with paper and a pencil like this, so why not also with COMSOL ;)

The issue you had with the solvers is that you must make a study per case with one solver, and that generate ONE specific Data-set that you can plot and analyse.
mixing several solver cases in THE SAME study is rather complex and you must know what is going on


--
Good luck
Ivar
Hi I would not say its that many "problems" with your model, it's just that COMSOL is allowing to simulate almost all physics, and that is in fact quite complex. And you are jumping in directly into multi-physics with coupling ES does not have frequency domain, but what you can do is to make a model based ONLY on ES, then use a Parameter "Param" and ramp your voltage on the electrode as V0*Param, and set the advanced continuation tab of your STATIONARY solver to Param and range(0,0.1,1) Then you can plot the Static force versus voltage. In the next model you can simulate SOLID ONLY, and use a boundary force on each side of the half of the ES calculated force, and here you can use a frequency domain sweep. This is not as a fully coupled model, as the force is probably not linear with the voltage, and the force will deform the beam, hence the voltage induced force would also change in a fully coupled mode. You would calculate the effects when working with paper and a pencil like this, so why not also with COMSOL ;) The issue you had with the solvers is that you must make a study per case with one solver, and that generate ONE specific Data-set that you can plot and analyse. mixing several solver cases in THE SAME study is rather complex and you must know what is going on -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.