Open the interactive field viewer in a browser. The JavaScript dataset and Plotly library are included beside it; the viewer does not need an OpenEMS installation or external data service. This is a publication of saved results, not a simulation runner.
The no-fence control removes both the top ground guard and the via row. The sweep therefore compares the combined guard-plus-via arrangement against that control, rather than isolating only the effect of vias.
The viewer includes 8 mm, 4 mm, 2 mm, 20 mil and 10 mil centre-to-centre pitches, with no-fence and continuous-wall reference controls. Choose the left and right comparison, rotate either board, and move the field slices. Both boards and all cases use the same logarithmic magnetic-field scale. The bottom plane can be hidden for visibility; it remains part of every simulated model.
The prior coupling runs saved geometry, port traces and spectra, but not field volumes. The recorded field-export runs repeated each of the seven base cases with a frequency-domain H dump at 1 MHz and 20 MHz. The original coupling runs remain unchanged. These additional runs use the same physical geometry, material properties, ports, mesh, excitation and native iteration count as their respective original case. The XML comparison excludes only the added field dump.
The retained native HDF5 source volumes contain the native half-domain complex fields and mesh coordinates. The published verification record contains durations and raw-field hashes; full solver traces and raw volumes are not included here. H is divided by complex total source-loop current, giving units of (A/m)/A = 1/m. At a 1 A sinusoidal source-current amplitude, numerical H values correspond to A/m. This current is not DC converter load current.
The full board is reconstructed using the PEC x = 0 symmetry: H(-x,y,z) = [-Hx(x,y,z), Hy(x,y,z), Hz(x,y,z)]. Fields are not multiplied by two. The original full-board no-fence check validated port reconstruction; no new full-domain field solve was made. The field window is inside the absorbing boundary.
The original field-export workflow produced fourteen ParaView multiblock files, one for each of seven cases at 1 and 20 MHz. Their field grids retain the mirrored native samples, with complex real/imaginary vectors and the corresponding copper surfaces. Those large native files remain outside this public package. Geometry coordinates are in millimetres; normalized field components are in 1/m. Use one fixed color scale across comparisons.
The browser dataset is derived from those same native values. It retains every fourth x sample and every second y sample, with every z sample retained. Its smooth display interpolates between retained samples; a smooth image does not prove convergence or resolve sub-cell via gaps.
These are short-record field diagnostics, not independently converged steady-state 1 MHz fields. The simulation uses ideal perfect conductors, a lossless dielectric, a finite open guard without a lid, and no conducted supply or shared-ground noise path. Full residual-energy convergence and pointwise field convergence are not established. Reproducing the original pickup metric checks that adding dumps did not change that observable; it does not validate every field sample. The smooth fields and the comparison at 20 MHz are not proof of low-frequency convergence.
The base grid resolves a 0.127 mm post with only about two cells. Browser display keeps every fourth x sample and every second y sample, with all z samples; rendered surfaces interpolate between the retained samples. The ParaView grid retains all native samples. Fine detail between 10 mil vias is limited by the solver mesh and is further reduced in the browser display.
The displayed magnitude is the norm of the complex H vector. Optional equal-length arrows indicate instantaneous direction at the selected phase. They are not field-line trajectories, and phase stepping is not a transient propagation movie. For the pickup comparison, use the separate quasi-static coupling report; raw 1 MHz field values and that fitted port metric are different quantities.
The validation record documents source hashes, geometry/mesh equality, port-metric reproduction and run durations. All fourteen ParaView multiblock files were opened with the native VTK 9.1 reader, and their coordinates and complex vectors were checked against the HDF5 data. Browser display magnitudes were also compared against the native data. The original browser review exercised comparison, field mode, synchronized camera, slice, ground-visibility and phase controls. That review did not establish physical accuracy or convergence.
OpenEMS documents the field-dump types and interpolation caveats in its CSPropDumpBox reference, and the limitations of tiny-cell low-frequency FDTD in its mesh guidance.
This viewer contains the saved browser dataset, its renderer, the local Plotly dependency, the comparison CSV and field-export verification. It contains no solver executable, private native PCB project, raw solver archive or external data-loading dependency. Raw-volume hashes are provenance identifiers, not download links. No hardware measurement, EMI compliance or fully converged 1 MHz field claim is made.