Discussion Forum

Posted: 1 decade ago 18 apr 2013, 01:16 GMT-4

Hi check your initial conditions, if you "solve for initial conditions" your V is probably =0, but if you have set it at t>0 to some other finite value, the results for t=0 will not be fully clean, as the time derivative must be respected and if you have a Dirac at "t=0" it will flow over in the final response. use rather a ramp of short rise time, and you will get V=0 at t=0 in your solutions too -- Good luck Ivar

Posted: 1 decade ago 18 apr 2013, 03:19 GMT-4

Hi Electric potential boundry condition is set to Vo = 30 V, in this instance in any time V non equal zero. Set Electric potential boundry condition Vo = 0 in t = 0 and enjoy. --- COMSOL Russian User Group http://vk.com/comsolmultiphysics

Attachments:

• 333.png

Posted: 1 decade ago 18 apr 2013, 04:41 GMT-4

[quote]I was assuming that at t=0, V =0. My initial value for V is zero.[/quote] This is a misunderstanding. The initial values you specify in the Comsol physics interface are simply start values for the solver. You can think of it as the "first guess", which the solver takes and then starts iterating until it converges to the solution (as defined by the partial differential equations and the boundary conditions). This should not be confused with the meaning of "initial value" in "initial value problem", which refers to solving a differential equation (usually an ODE) for a given initial condition (at t0, but not necessarily at t0 = 0). [quote]Also, at what value of t does the solver apply the electric potential?[/quote] If you have an "Electric potential" boundary condition in a time-dependent problem and enter a simple constant (as opposed to a function of time, using the built-in time variable "t"), then this potential is applied at [i]all[/i] times, i.e. starting at t = −∞. [quote]If I can't get around this, I would want to control the application of the electric potential to follow an exponential ramp such as: Vrf= (Vf -V0)* 1-exp(-t/tau) where Vf is the final potential desired, V0 is the initial potential at t=0, t = time, and tau = 1 sec.[/quote] Basically you have to enter the above formula in the "Electric potential" boundary condition and define all those constants as a "Global parameter".

Posted: 1 decade ago 18 apr 2013, 13:06 GMT-4

All, Thanks for so much valuable information about this subject. While I was waiting for a response, I was searching threads of this forum. I found something that is also relevant. In my model. Now I see that since I am solving the electrical problem in the frequency domain, the voltage plot should always look the same irrespective of the time step. There really is no t= 0 for that analysis, correct? In order to do the analysis for my purposes, I need to apply an AC voltage in the time domain so that it is included in the time dependent analysis. In all the threads, there were some incomplete suggestions but not concrete steps on how to apply an AC voltage in a time dependent study. One suggestion was to apply the voltage, Vo as Vo= 2* pi* f * time. No example file was posted. If this is how it's done, could someone post an example file? If this is not correct, could someone explain how to do this? Many thanks in advance, Art

Posted: 1 decade ago 18 apr 2013, 14:02 GMT-4

Hi in frequency domain swep you define the amplitudes of your AC voltage (or excitation signal) and/or phys with a phasor expression. but its for steady state results. You have also harmonic perturbation on a static DC level, as well as frequency time dependent (time dependent amplitude of a constant AC signal) (but no mix frequecy spectrum x time) this you obtain normally by superposition, assuming that the response is linear. In time series with 2*pi*freq*t you are really looking for the transient behaviour, take care with your initial conditions, else you will excite large swingings at t=0 that might not damp out and might mask competely your desired analysis ;) -- Good luck Ivar

Posted: 1 decade ago 18 apr 2013, 14:17 GMT-4

Hi Ivar, Thanks for your reply. Do you have more specifics on setting up this type of an analysis to prevent the kind of issues that you describe? Perhaps an example study? Thanks, Art

Posted: 1 decade ago 19 apr 2013, 01:02 GMT-4

Hi No I do not have any specific examples. Whatt I usually do, each time I change domain (Physics, and what I do frequently from the demand on modelling coming in ;) is to run through a few of the model library examples, then to make me some easy examples if I have any doubts. One thing, take some care with examples based on "cubes" in a cartesian corrdinate system: use rather a sphere or a skewed cube, as often you are studying an exception of the general theory that might show up when you are nicely aligned with the coordinate system For transient analyiss, the main disturbance comes from sudden BC values turned on at t=0. "My way" is to run a stationary case first (when possible) and use that as initial conditions, and supplied with time series equations when required. To mini mise the t=0 turn on transients. Sometimes, you might also start by turning OFF the inertial terms (d/dt^2 ones) or use the "creep flow" option, but go back once you have got the settings OK -- Good luck Ivar

Posted: 1 decade ago 19 apr 2013, 12:30 GMT-4

Hi Ivar, Thanks for the background; very useful. Another tidbit that shines some light on the subject comes from the AC/DC User's Guide for V4.3 on page 255: At the top of the page of the section "Theory for the Magnetic and Electric Fields Interface," is a note that says: "The Magnetic and Electric Currents interface only supports the stationary and frequency domain study types—that is, there is no transient formulation available." This would seem to say straight out that time dependent analysis for Electric Currents is not available, correct? Thanks, Art

Posted: 1 decade ago 19 apr 2013, 14:14 GMT-4

Hi I wouldnt say time dependent analysis is out in all ACDC, but the ACDC hypothesis is based on "small model compared to the wavelength", i.e. time dependent propagation of fields are instantaneous in some sens. But I have done time series due to capacitive discharge, inductive load etc. it is if you want to do high frequency field propagation with time series you need to use the RF. Check carefully the ACDC and RF hypothesis. there is a grey zone where both are valid or both are about as "wrong" ;) -- Good luck Ivar