Of course C = eA/d, which shows no relationship with frequency but is there something more happening at frequencies above 100MHz? I understand that capacitive impedance changes as the parasitic resistance and inductance becomes more dominant. Even that the capacitor will appear like and inductor after a resonance frequency

I have been measuring capacitance of overlapping microstrip capacitors using a 4395A Impedance Analyzer up to from 10kHz to 100MHz. I am monitoring Cs and Rs. I'm measuring 15-40pF for different overlap lengths. As frequency increases I am seeing a decrease in capacitance of about 10% per decade until around 40MHz where the value turns to increase sharply before 100MHz (or course I cant see anything above 100MHz). The increase is sharper for greater overlap lengths, higher capacitance

Is there some limitation of my analyzer to truly separate capacitance and esr from the parasitics? Or is there something in the understanding of capacitance that I am missing?

 

All I know about capacitors came from Feynman Vol II. He conducts you on a MMT of both capacitors and inductors as the frequency goes from DC to blue light. It is well worth the price for just that one chapter. A bargain at $18.99 used!

http://www.amazon.com/Feynman-Lectu...7&sr=1-6&keywords=feynman+lectures+on+physics

 

Without seeing your physical test setup, we can only speculate, but I suspect you are seeing the effects of inductance. Your impedance analyzer should also display phase. At resonance (the minimum impedance frequency), the phase should change from ≈-90° to ≈+90° fairly abruptly.
Dielectric constant of most substrates is also a function of frequency, but wouldn't result in the dramatic changes you are seeing.

See the attached schematic and plot for an example of series resonance impedance.

EDIT: 100nH seems too high for any reasonable test setup and PC layout that I can imagine. Can you post a photo of your setup?