gix_d
Newbie
Hi all, (and thanks in advance! 
)
I am simulating a twisted pair going through different noise mitigation (low-pass) constructions in HFSS. The wires are 32 AWG, so about 200 μm in diameter, and with 13 μm thick enamel. Our plan is to run the twisted pair through the followings: 1) Highly lossy dielectric filling. 2) Pressing it between two metal plates, thus making the section through the plates capacitive. 3) Immersing it in a conductive epoxy. We expect the last one to effectively decouple the paired wires and form two "coaxial cables," where the inner and the outer conductors are the wire and the epoxy, and the enamel is the dielectric between them.
The problems I am having are generally due to our interest of kHZ~sub-GHz transmission. We understand this is where the quasi-static/lumped element approximation could apply to reduce the complexity/meshing, so we should probably use programs like Ansys Maxwell instead of HFSS, but we don't have Maxwell. So, our goal is to find potential solutions to simulate large/long-wavelength models in HFSS, such as invoking symmetry or smart boundary conditions.
Specifically, we are having the following problems. First, I think we need to make the wire rather long, so the whole model doesn't behave like a lumped element, otherwise I would expect the transmission to artificially ignore the "noise blocker" in the middle of the line. We have wave ports at the two ends of the line. We think, if we can set the ports to always match the impedance of the line, thus no reflection on the ports, the line is effectively infinite and allows us to make it short. However, we don't know how to achieve this, and we did find that our simulated S21 (and field visualization) had resonances corresponding to the distance of the ports. I should note that I already put on radiation boundaries overlapping the wave ports, but apparently the field still sensed the impedance-mismatched ports rather than the radiation boundary.
Second, my understanding is that the wave ports need to be substantially larger than the field of interest, in other words several times wider than the wavelength. This would require an extremely large port size in our case. However, my understanding is that the requirement is for exciting modes other than (quasi-)TEM, which actually oscillate within the area of the port. So, do we really need to satisfy this large-port requirement, since we are predominantly long-wavelength and thus likely dominated by equipotential, quasi-static transmission anyway? Alternatively, is there a way to invoke symmetry, e.g., some kind of boundary condition, to make the size of the wave ports effectively infinite but small in the model?
Finally, we do have COMSOL Multiphysics on our lab PC. I have never used it and am wondering if it'd be more appropriate for my application? A quick check on YouTube seemed to suggest it's useful for simulating quasi-static problems.
Thank you!
- Yen-Yung
)I am simulating a twisted pair going through different noise mitigation (low-pass) constructions in HFSS. The wires are 32 AWG, so about 200 μm in diameter, and with 13 μm thick enamel. Our plan is to run the twisted pair through the followings: 1) Highly lossy dielectric filling. 2) Pressing it between two metal plates, thus making the section through the plates capacitive. 3) Immersing it in a conductive epoxy. We expect the last one to effectively decouple the paired wires and form two "coaxial cables," where the inner and the outer conductors are the wire and the epoxy, and the enamel is the dielectric between them.
The problems I am having are generally due to our interest of kHZ~sub-GHz transmission. We understand this is where the quasi-static/lumped element approximation could apply to reduce the complexity/meshing, so we should probably use programs like Ansys Maxwell instead of HFSS, but we don't have Maxwell. So, our goal is to find potential solutions to simulate large/long-wavelength models in HFSS, such as invoking symmetry or smart boundary conditions.
Specifically, we are having the following problems. First, I think we need to make the wire rather long, so the whole model doesn't behave like a lumped element, otherwise I would expect the transmission to artificially ignore the "noise blocker" in the middle of the line. We have wave ports at the two ends of the line. We think, if we can set the ports to always match the impedance of the line, thus no reflection on the ports, the line is effectively infinite and allows us to make it short. However, we don't know how to achieve this, and we did find that our simulated S21 (and field visualization) had resonances corresponding to the distance of the ports. I should note that I already put on radiation boundaries overlapping the wave ports, but apparently the field still sensed the impedance-mismatched ports rather than the radiation boundary.
Second, my understanding is that the wave ports need to be substantially larger than the field of interest, in other words several times wider than the wavelength. This would require an extremely large port size in our case. However, my understanding is that the requirement is for exciting modes other than (quasi-)TEM, which actually oscillate within the area of the port. So, do we really need to satisfy this large-port requirement, since we are predominantly long-wavelength and thus likely dominated by equipotential, quasi-static transmission anyway? Alternatively, is there a way to invoke symmetry, e.g., some kind of boundary condition, to make the size of the wave ports effectively infinite but small in the model?
Finally, we do have COMSOL Multiphysics on our lab PC. I have never used it and am wondering if it'd be more appropriate for my application? A quick check on YouTube seemed to suggest it's useful for simulating quasi-static problems.
Thank you!
- Yen-Yung