| Author | Message | |||||
     Alexej Kushunin (kaa_mmf) New member Username: kaa_mmf Post Number: 1 Registered: 06-2006 |
I've created script with FlexPDE v 5.0.7 for timedependent problem, stress with diffusion. See atteched file. On third step I've became error Time Step has fallen...and so on. See attached screenshot. Boundary conditions are very simple. I've read posts about the same problem but not found solution. Best Regards Alexej.
| |||||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 698 Registered: 06-2003 |
Your attached script does not seem to be downloadable. Please repost by inserting the text rather than attaching. Have you tried the latest version of FlexPDE? It is possible that the trouble you are haveing was solved in a later version. | |||||
| Alexej Kushunin (kaa_mmf) New member Username: kaa_mmf Post Number: 2 Registered: 06-2006 |
Here is script _______________________ SELECT ERRLIM=0.1 VARIABLES Q(threshold=0.1) Sigmax(threshold=0.1) Sigmay(threshold=0.1) Tauxy(threshold=0.1) DEFINITIONS MESH_SPACING =3 x0=3 y0=3 A=25 b=5 l=5 rad=0.1 Sxx=(Sigmax/1.05-1)/0.91+0.942*(1+Q) Syy=1.1*(0.95*Sigmay-1)+0.942*(1+Q) Txy=1.1*(0.95*Tauxy-1) beta=0.5*Atan2(2*Txy,Sxx-Syy) S1=(Sxx+Syy)/2+((Sxx-Syy)/2)*Cos(2*beta)+Txy*Sin(2*beta) S2=(Sxx+Syy)/2-((Sxx-Syy)/2)*Cos(2*beta)-Txy*Sin(2*beta) S3=0.3*(Sxx+Syy) INITIAL VALUES Q=0 Sigmax=0 Sigmay=0 Tauxy=0 EQUATIONS Q:DXX(Q)+DYY(Q)-Dt(Q)=0 Sigmax: DXX(Sigmax)+DXX(Sigmay)+DYY(Sigmax)+DYY(Sigmay)+0.14*(DXX(Q)+DYY(Q))=0 Sigmay:DX(Sigmax)+DY(Tauxy)=0 Tauxy:DY(Sigmay)+DX(Tauxy)=0 BOUNDARIES REGION 1 Start(l,0) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmax)=8.9445 line to (l,-b/2) fillet(rad) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmay)=-8.9445 line to (0,-b/2)fillet(rad) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmax)=8.9445 line to (0,-A) Natural(Q)=0 Load(Tauxy)=0 Load(Sigmax)=0 Load(Sigmay)=0 line to (2*A,-A) Natural(Q)=0 Load(Tauxy)=0 Load(Sigmax)=0 Load(Sigmay)=0 line to (2*A,A) Natural(Q)=0 Load(Tauxy)=0 Load(Sigmax)=0 Load(Sigmay)=0 line to (0,A) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmax)=8.9445 line to (0,b/2)fillet(rad) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmay)=8.9445 line to (l,b/2) fillet(rad) Load(Tauxy)=0 Load(Sigmax)=8.9445 line to CLOSE TIME 0 to 0.001 by 0.00005 MONITORS FOR t = 0 by 0.00005 to 0.001 Contour(Q) Contour(Sxx) ZOOM(3,-4,4,8) Contour(Syy) ZOOM(3,-4,4,8) Contour(Txy) ZOOM(3,-4,4,8) Contour(S1) ZOOM(3,-4,4,8) Contour(S2) ZOOM(3,-4,4,8) Contour(S3) ZOOM(3,-4,4,8) END | |||||
| Alexej Kushunin (kaa_mmf) Junior Member Username: kaa_mmf Post Number: 3 Registered: 06-2006 |
Here is script _______________________ SELECT ERRLIM=0.1 VARIABLES Q(threshold=0.1) Sigmax(threshold=0.1) Sigmay(threshold=0.1) Tauxy(threshold=0.1) DEFINITIONS MESH_SPACING =3 x0=3 y0=3 A=25 b=5 l=5 rad=0.1 Sxx=(Sigmax/1.05-1)/0.91+0.942*(1+Q) Syy=1.1*(0.95*Sigmay-1)+0.942*(1+Q) Txy=1.1*(0.95*Tauxy-1) beta=0.5*Atan2(2*Txy,Sxx-Syy) S1=(Sxx+Syy)/2+((Sxx-Syy)/2)*Cos(2*beta)+Txy*Sin(2*beta) S2=(Sxx+Syy)/2-((Sxx-Syy)/2)*Cos(2*beta)-Txy*Sin(2*beta) S3=0.3*(Sxx+Syy) INITIAL VALUES Q=0 Sigmax=0 Sigmay=0 Tauxy=0 EQUATIONS Q:DXX(Q)+DYY(Q)-Dt(Q)=0 Sigmax: DXX(Sigmax)+DXX(Sigmay)+DYY(Sigmax)+DYY(Sigmay)+0.14*(DXX(Q)+DYY(Q))=0 Sigmay:DX(Sigmax)+DY(Tauxy)=0 Tauxy:DY(Sigmay)+DX(Tauxy)=0 BOUNDARIES REGION 1 Start(l,0) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmax)=8.9445 line to (l,-b/2) fillet(rad) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmay)=-8.9445 line to (0,-b/2)fillet(rad) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmax)=8.9445 line to (0,-A) Natural(Q)=0 Load(Tauxy)=0 Load(Sigmax)=0 Load(Sigmay)=0 line to (2*A,-A) Natural(Q)=0 Load(Tauxy)=0 Load(Sigmax)=0 Load(Sigmay)=0 line to (2*A,A) Natural(Q)=0 Load(Tauxy)=0 Load(Sigmax)=0 Load(Sigmay)=0 line to (0,A) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmax)=8.9445 line to (0,b/2)fillet(rad) Natural(Q)=9 Load(Tauxy)=0 Load(Sigmay)=8.9445 line to (l,b/2) fillet(rad) Load(Tauxy)=0 Load(Sigmax)=8.9445 line to CLOSE TIME 0 to 0.001 by 0.00005 MONITORS FOR t = 0 by 0.00005 to 0.001 Contour(Q) Contour(Sxx) ZOOM(3,-4,4,8) Contour(Syy) ZOOM(3,-4,4,8) Contour(Txy) ZOOM(3,-4,4,8) Contour(S1) ZOOM(3,-4,4,8) Contour(S2) ZOOM(3,-4,4,8) Contour(S3) ZOOM(3,-4,4,8) END | |||||
| Alexej Kushunin (kaa_mmf) Member Username: kaa_mmf Post Number: 4 Registered: 06-2006 |
Sorry for double post, I've some problem with internet... | |||||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 699 Registered: 06-2003 |
1. You have specified LOAD boundary conditions for Sigmay and Tauxy, but the equations for these variables are first order, so the LOAD boundary condition is undefined (See "Natural Boundary Conditions" in the Help Index.) 2. If you clear up this trouble and create an equation set for which the solution converges, you will then run into trouble with the huge error tolerance (ERRLIM) you have selected. This request for sloppy solution will probably lead to oscillation and subsequent harsh timestep chopping. Use the default ERRLIM to start with. | |||||
| Ravindra Kashyap (kashyapastron) New member Username: kashyapastron Post Number: 1 Registered: 11-2006 |
Hello I thought of posing my question here only instead of starting a new thread as even my doubt is about time-dependent simulation. I am analysing a stripline to via transition in a multilayered PCB. I am doing a 2D analysis for this. I've attached the file also. What I want to see is the effect on the Ground reference planes when we excite the strip with a high frequency pulse.
I've following 3 queries. 1. Can I excite the middle conductor with a high frequency pulse (say a unit pulse with a high rise time or a step). 2. I tried exciting this middle conductor with a sinusoid of 10GHz. but the E contour plots are not coming properly. 3. I also want to see the voltage variations along the reference planes (to see the dependance of the variations of voltage on its relative distance from the via). I tried the 'feature' command and evaluated the line integral. but I don't think it is correct. Please Help me as soon as possible | |||||
| Ravindra Kashyap (kashyapastron) New member Username: kashyapastron Post Number: 2 Registered: 11-2006 |
Hello I thought of posing my question here only instead of starting a new thread as even my doubt is about time-dependent simulation. I am analysing a stripline to via transition in a multilayered PCB. I am doing a 2D analysis for this. I've attached the file also. What I want to see is the effect on the Ground reference planes when we excite the strip with a high frequency pulse.
I've following 3 queries. 1. Can I excite the middle conductor with a high frequency pulse (say a unit pulse with a high rise time or a step). 2. I tried exciting this middle conductor with a sinusoid of 10GHz. but the E contour plots are not coming properly. 3. I also want to see the voltage variations along the reference planes (to see the dependance of the variations of voltage on its relative distance from the via). I tried the 'feature' command and evaluated the line integral. but I don't think it is correct. Please Help me as soon as possible | |||||
| Ravindra Kashyap (kashyapastron) Junior Member Username: kashyapastron Post Number: 3 Registered: 11-2006 |
My apologies for that repeated posting. My browser showed me a problem after first upload, so i uploaded again | |||||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 703 Registered: 06-2003 |
You have specified time-dependent boundary conditions, but you have only a steady-state PDE. This provides no basis for a timestep control, and the problem will run with timesteps dictated by print intervals. There will be no time-dependent effects in the fields, because none are represented in the PDE. For a sequence of steady-state snapshots, you would do better with a STAGED problem. 1. You could use the UPULSE function, or a combination of SWAGEs, or a mathematical function such as the supergaussian to build a pulse in time. 2. You'll have to explain what you don't like about the E contour plots. They look ok to me. 3. You can use ELEVATION(V) FROM (a,b) TO (c,d) to examine voltage variations along a line. Or you can use an ELEVATION(V) ON "name" to plot variations on a boundary path designated by "name". 3. Why don't you think the integral is correct? | |||||
| Ravindra Kashyap (kashyapastron) Member Username: kashyapastron Post Number: 4 Registered: 11-2006 |
Hello My justification for using a steady state laplace equation for the governing equations was this Since I am analysing a 2 conductor transmission line, the system can support a TEM wave, but TEM waves have the form of a static field with its transverse components satisfying Laplace equation. So I thought that governing equation will be the Laplace equation itself. regarding the time-dependant boundary condition, since the field itself is static, cant we analyse the time-dependant behaviour as a static behaviour, if we freeze the time variable. That is if time instant is fixed, then the static field exists depending on the boundary condition at that point. How to use stage for this. Can we vary the voltage according to a function using stage. say the voltage to trace a sinusoid during each stages. Regarding the integral of norm(E), It shows a monotonic behaviour while I was expecting a peak near the central strip transition (as the energy pockets are present at the transitions). Thats why I thought it was not OK. | |||||
| Ravindra Kashyap (kashyapastron) Member Username: kashyapastron Post Number: 5 Registered: 11-2006 |
Hello Mr. Nelson I am really stuck with this problem. My problem is that, I want to do a full-wave analysis of the via structure explained above. My field will be a TEM so the longitudinal components will be solutions of laplace equations Ez:grad(div(Ez))=0 Hz:grad(div(Hz))=0 The boundary condn are that tangential components of Hz and Ez are zero at the metal interface. This can be implemented as "natural" and "value" boundary condition. But the problem is, where to incorporate the voltage excitation and how to study the frequency dependance. Please reply as soon as possible Thanks | |||||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 706 Registered: 06-2003 |
1. You have imposed a value boundary condition for V on the conductor. Why does this not satisfy your desire to impose a voltage on the conductor? 2. There is no frequency dependence unless the frequency appears somewhere in your equations. In your case, it does not. If you want to examine frequency dependence, then probably you want complex equations, which incorporate frequency factors. You should get a copy of Prof. Backstrom's book "Waves by Finite Element Analysis". There is a link on our Bookstore page www.pdesolutions.com/bookstore.html. You can also look at our applications manual on that same page, and the example problems in the Backstrom_Books folder in the FlexPDE distribution. 3. You can use a time-dependent boundary specification to drive steady-state equations as long as you provide a timestep control. One way to do this is to SELECT FIXDT and include a BY clause in the TIME section. | |||||
| Ravindra Kashyap (kashyapastron) Member Username: kashyapastron Post Number: 6 Registered: 11-2006 |
Hello Thanks for your useful replies for my question. Finally I have formulated my problem using the scalar electric potential. What I am doing is trying to solve an eigen value problem for the given transmission line structure with the specified boundary conditions. I have two queries though you might find them trivial 1. After obtaining the eigen values and finding the omega. How to use this information for exciting the stripline with a sinusoidal function of some "omega" (selected as one of eigen values) 2. We have solved the eigen value problem with some fixed voltages fed to the referrance planes. Now can i just solve the above wave equation again with omega being one of these eigen values, but this time with a periodic boundary condition. My idea is to consider the total length L of the stripline as say some fraction of the wavelength and use a boundary condition which varies sinusoidally with distance. In effect we have excited the center strip with a sinusoid. Is it logical. I have also attached my script file. I am able to observe the modes but the eigen values are coming out be too high.
Please reply as soon as possible. I am running out of time. | |||||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 707 Registered: 06-2003 |
Your script does not appear to be downloadable, so I cannot run it. If you solved the eigenvalue problem for Vr and Vi using the assumption that complex V=Vr*cos(omega*t)+i*Vi*sin(omega*t), then a fixed value BC of Vr really means a sinusoidal oscillation at angular velocity omega and amplitude Vr. A boundary condition sinusoidal in distance is expressed as VALUE(var)=Amplitude*sin(2*pi*X/wavelength) or some such specification. | |||||
| Ravindra Kashyap (kashyapastron) Member Username: kashyapastron Post Number: 7 Registered: 11-2006 |
Hello I think my problem could have been solved by just the solution of laplace equation but with the boundary condition which varies with distance. My logic is this. Suppose I solve the laplace equation with the center strip at a fixed voltage and the other two plates at ground. Then the transverse electric field will simply be pointing from the center strip towards the ground plane. Now If I vary the voltage along the strip so that the voltage between the ground planes and center strip is continiously varying along the strip, then the field will no longer be the same. Am i right I saw your reply for varying boundary condition, but I couldn't apply it. You have mentioned the distance as 'X' and have specified the voltage using the sine function, but when we draw the model, we never use any variable. What to put for 'X' Thanks Please reply soon |
