I suspect that the problem is four-fold:
1. As has been pointed out, there is a slight embedded microstrip
effect here (see the calculator for this effect.)
2. Er is a function of frequency, so at the frequency of interest,
Er might be mischaracterized
3. Er might also be mischaracterized simply because it often is not
exactly what you expect it to be or what it is spec'd at
4. Manufacturing processes cause a variation that is hard to control.
We have found variations as much as 3 or 4 ohms ALONG A GIVEN TRACE
and especially between boards in the same production run. I have found
that the practical accuracy for impedance vs spec is about 10%.
See the results reported in the article "The Effects of Vias on PCB
Traces" PCB Design Magazine, 8/96 for some real world, controlled examples
of how much variation there can be.
At 02:35 PM 12/9/97 -0000, you wrote:
>Polar Instruments, UK, have a useful little calculator program (public
>domain) that allows you to check this effect. Their web site is
>I've put John's figures into this, and it shows about a 5 ohm drop due
>to the mask.
>The calculator also confirms Kenneth Willis' comments about overplating
>being responsible for some of the impedance drop.
>Perhaps someone from Polar would like to comment, especially about the
>source for their equations?