Because the diagram shows only a few field lines, some areas appear completely devoid. This lack of lines does not make crosstalk zero within those zones. If you imagine a diagram with hundreds or thousands of lines, you can see that the crosstalk is never zero at any point, but it is substantially less in areas in which the lines have more distance between them, and it varies smoothly from point to point.
Imagine sliding via Pair A to the right, halving its distance to the central pair. In that new position, three field lines penetrate the space between the purple and the red elements of Pair A, substantially increasing the crosstalk it perceives. As you push more to the right, further decreasing the distance, crosstalk increases markedly. In general, differential-to-differential crosstalk varies inversely with the square of distance, much like the differential-to-differential crosstalk between PCB traces (ref ).
Now consider via Pair B on the right side of the figure. The axis of this pair lies parallel to the magnetic lines of force so that no lines penetrate between its purple and its red elements. Fabulous! With this orientation, the central via produces no differential-mode crosstalk at B. In a long row of differential via pairs, turning every other pair in this way suppresses all nearest-neighbor differential crosstalk—a good plan and one that I am surprised I don’t see more often.
Quadrature layout is not a new idea. If you look at multichannel analog circuitry from the mid-20th century, such as telephone exchanges or high-fidelity recording equipment, you’ll see oodles of audio transformers, mounted in rows, with every other transformer turned 90° with respect to its nearest neighbor. Such physical arrangements successfully mitigate differential crosstalk from nearby aggressors. Quadrature layout does nothing to mitigate common-mode crosstalk, but at least it nails the differential mode.
The magnetic-field diagram admits one additional interpretation: the 2-D scaler magnetic potential. This concept applies to all situations involving parallel conductors, such as long cables, PCB traces, and arrays of vias. The 2-D potential interprets the magnetic-field diagram as a topographic contour plot, with the purple signal via perching atop a hill and the red signal via sitting down in a valley. The lines of magnetic force represent contours of constant height, or potential. The key insight you can derive from this method is simple: Any two elements on the same contour, or at the same potential, exhibit zero differential crosstalk between them. At position C, for example, the two vias lie along the same line of magnetic force. That alignment eliminates differential-mode crosstalk from the central source (ref ). The way the contours work, no matter where you place a differential via pair, you can always rotate its alignment to mitigate crosstalk from a troublesome differential source.