[SOLVED] microstrip impedance without considering the solder mask?

How big impedance error do I get if I calculate a microstrip impedance without adding the solder mask dielectric constant to the calculation?

If I should consider the solder mask also the impedance would be different. But how much? Maby more on thin routes.

I've agonised over this in the past and found that the solder mask has made *zero* difference to the expected performance of my designs (well, variations were within what I considered acceptable manufacturing tolerances anyway). I can't think of a single case where I could attribute a resonant frequency variation/impedance offset solely to an unmodelled solder mask.

This has been my experience on commercial PCBs (Er ~ 3.5) up to 5 GHz, and with coplanar waveguide/microstrip track widths down to 0.2mm at any rate. It's up to you, but I don't bother stressing anymore

[Not to mention some practical difficulties: the RF characteristics of solder mask are hard to find, and I've found (via microsectioning) that there are huge thickness variations (e.g. 3 - 20 um) in the mask across a board from a number of manufacturers.]

Second that, as microstrip parameters are generally calculated on the basis that half the wave is travelling through air and half through the PCB substrate, a little thickness of solder resists dosn't seem to make that much difference.