kicad-developers team mailing list archive

[Werner Almesberger <werner@...>]

vladimir_uryvaev wrote:
> Ofcourse, we should take 20 nm (GCD for 0.1mil and 1mm) or less quantum
> to keep current accuracy. We may take 1 nm quantum to make KiCad accurate
> enough for some decades (and for chip design:)).

Considering that some of today's boards are much smaller than
yesterday's components, that may not be such a bad idea.

I would prefer a power of ten, to make it easier to spot
equivalences.

> Only analytic number representation can remove rounding problem completely. :)

If such an ability came from retaining more context of where
dimensions originate, that would actually be quite useful in
a number of cases. It's basically what I'm trying to do in
fped, where the underlying math and geometry can be made
explicit. I think the next generation library initiatives will
also run into this issue before too long.

- Werner