@Tim Edwards In Kunals Udemy course on crosstalk I learnt that parasitic capacitance is even possible between the outer wires when there are wires in between. So I would assume that nearly every wire has a very small capacitance on every other wire of a design, the capacitance values are just very very small. Shouldn't the cthresh value for extraction be something bigger than 0? From that theory I would expect the extraction to generate O(n²) parasitic capacities.

Yes, the field lines extend over the top and under the bottom of the wire and will couple to wires far away. But the order of magnitude of the coupling drops considerably. Previous versions of magic would compute the coupling as if there were no intervening wire at all, which is a much worse error than assuming that there is no coupling. To properly compute the remaining coupling on the other side of a shielding object, I need more sophisticated modeling. I will be doing some of that modeling soon, to see what 2nd (and higher) order effects need to be captured to keep the parasitic error to a reasonable number, especially as the feature size goes to 90nm and below.

Thanks!