I am making some simulations for a stripline filter in Ku band (17.2 GHz) using a DiClad 880 Substrate.
Dielectric constant variation should be as plotted below
However I see that in layout simulation the dielectric constant variation is significantly greater.
For example, the following is a plot of the filter response from schematics:
while the next one is the layout simulation:
To center the pass-band at 17.2 I have to manually tune the dielectric constant from 2.2 (schematics SSub) to 1.81 (layout substrate).
Could someone explain such a huge variation vs the datasheet plot ?