Hi,
I am trying to combine Spice with MEF to simulate parallel resistors on a PCB at higher frequency where skineffect changes the behavior of the copper. The model solves correctly for low frequencies but simulation goes crazy for freq > 1MHz. This is a testcase presented. The real problem is way bigger.
I source 10V into one copper plane, that is connected through the two resistors (1mOhm and 1.5mOhm) to GND. As you can see in the attached image the solution does not converge (i stopped it after some minutes) and voltages are way out of range.
The behavior is independend of meshsize but dependend on frequency. the higher the freq. the worse the solution.
Usually i would expect the copper and the spice-resistor to act as a frequency dependend voltage divider, since the coppers resistance increases with frequency due to skineffect and inductive behavior.
Maybe there is a better way to excite the copper plane? Maybe the Electric Potential /Ground - BoundaryCondition is wrong here, but what else i can use?
I feel the problem is more systematic, because e.g. i cant use a currentsource instead of the voltage soruce. Since there is no GND the current can flow to, there is no solution found in the first attempt, because Comsol doesnt know that the current actually goes through the resistor since the amount of current is calculated in a iterative process.
Can someone suggenst how to solve these kind of problem?
I am trying to combine Spice with MEF to simulate parallel resistors on a PCB at higher frequency where skineffect changes the behavior of the copper. The model solves correctly for low frequencies but simulation goes crazy for freq > 1MHz. This is a testcase presented. The real problem is way bigger.
I source 10V into one copper plane, that is connected through the two resistors (1mOhm and 1.5mOhm) to GND. As you can see in the attached image the solution does not converge (i stopped it after some minutes) and voltages are way out of range.
The behavior is independend of meshsize but dependend on frequency. the higher the freq. the worse the solution.
Usually i would expect the copper and the spice-resistor to act as a frequency dependend voltage divider, since the coppers resistance increases with frequency due to skineffect and inductive behavior.
Maybe there is a better way to excite the copper plane? Maybe the Electric Potential /Ground - BoundaryCondition is wrong here, but what else i can use?
I feel the problem is more systematic, because e.g. i cant use a currentsource instead of the voltage soruce. Since there is no GND the current can flow to, there is no solution found in the first attempt, because Comsol doesnt know that the current actually goes through the resistor since the amount of current is calculated in a iterative process.
Can someone suggenst how to solve these kind of problem?