Mail Archives: geda-user/2015/10/08/09:54:50
On Thu, 2015-10-08 at 04:39 +0200, Kai-Martin Knaak wrote:
> Stefan Salewski wrote:
>
> > The toporouter as described in in the PhD thesis of Tal Dayan from
> > 1997 assumes that there is an empty area for routing, which is only
> > restricted by the size of pads and pins. So each already existing
> > trace is a problem.
>
> I suspected so. So this is worse than just a not yet coded feature.
> Unfortunately, doing some portions in advance manually is key to deal
> with many special cases. Examples:
>
> * Make sure, there are proper high current highways
>
> * Do length compensated thin wires for critical timing
>
> * Put down these large heat dissipating copper areas
>
> * Let high voltage lines keep an extra distance to everything else
>
> * Related: Make sure, no track crosses the boundary of the enclosure
>
> Most of my boards have at least one of these cases.
>
Your points are true, but most can be solved. Last both points are
nearly trivial, just give that traces an larger amount of clearance. And
you can protect arbitrary areas from being used by the router by putting
some dummy pad there.
Heat: Put a pad there. With some work we may support arbitrary shapes.
High current traces should be also easy. Length compensation should be
also not a large problem.
More problematic is well defined impedance and matches pairs and
differential signaling. I think a differential trace pair should be
doable by replacing it fist with one trace with double extent, so space
is reserved, and finally replace that one by final two real differential
pairs.
Also you can always partition your boards in distinct areas, routing one
manually. And limited manual fixing is always possible at the end.
But all that needs some thinking of course. Such points generally are
not covered in detail in papers and thesis. Coding that is not much work
in high level languages, what takes some time is thinking and testing.
But it is true, an autorouter can not do all kind of boards well. It is
like coding in C and assembler.
I have yesterday visited the TopoR homepage with the commercial russian
router again. Seems to be very quiet.
- Raw text -