Hi:
I think rrumpf misunderstood myem's question. I think myem's question is that "when a structure is less than lambda/10, people normally can use quasi-TEM methods to get reasonable results". I think rrumpf considers the lambda/10 is normally the maximum grid size for numerical method to get reasonable results. Both cases are ok.
Why is lambda/10, this is not an absolute limit. It just says the size of the structure is much smaller than a wavelength. Staitc or quasi-static methods normally assume the propagation constatnt is 0 or close to 0. This assumption means that the structure should be much smaller than wavelength. How much is much smaller? I think it is dependent upon the real situation. For some structures, lambda.10 is already small and quasi-staitc solvers can yield reasonable results. For some structures, quasi-static solvers may not be able to get good enough results even with hte size smaller than lambda/100.
For numerical methods, the maximum grid size should be lambda/10. If it is larger than lambda / 10, the chance of significant numerical error is very high. However, for many structures, lambda / 10 may not enough. For some structures, it may require the grid size smaller than lambda / 100 to meet the high accuracy requirement. It is dependent upon the structures and the requirements of different designs. Regards.