A follow-up to https://github.com/awslabs/palace/pull/246, which removes the integration of the port area to calculate the geometry of uniform and coaxial lumped elements. This step was susceptible to discretization error, in particular for coaxial lumped ports with circular boundaries. These geometric properties can now be calculated directly using the element bounding box/sphere.
This change mandates rebaselining the coaxial regression tests. For the coaxial cable with inner and outer diameters of 1.6383 and 5.461 mm, respectively, the impedance per square for the 50 Ohm coaxial lumped port should be 260.9 Ohm/sq (R = Rs * ln(R_o/R_i) / 2π). The values computed by Palace and used internally for the impedance BC are:
main: 261.8 Ohm/sq
This PR: 260.9 Ohm/sq
Thus, this PR corrects a previous geometry processing accuracy issue which showed up for coarsely meshed curves. Looking at the port V and I time histories, it looks like the errors were on the order of 1% for this example.
A follow-up to https://github.com/awslabs/palace/pull/246, which removes the integration of the port area to calculate the geometry of uniform and coaxial lumped elements. This step was susceptible to discretization error, in particular for coaxial lumped ports with circular boundaries. These geometric properties can now be calculated directly using the element bounding box/sphere.
This change mandates rebaselining the coaxial regression tests. For the coaxial cable with inner and outer diameters of
1.6383and5.461mm, respectively, the impedance per square for the50Ohm coaxial lumped port should be260.9Ohm/sq (R = Rs * ln(R_o/R_i) / 2π). The values computed by Palace and used internally for the impedance BC are:main:261.8Ohm/sq260.9Ohm/sqThus, this PR corrects a previous geometry processing accuracy issue which showed up for coarsely meshed curves. Looking at the port V and I time histories, it looks like the errors were on the order of 1% for this example.