Is there a way to control how far into an I/O pin the detailed router will draw a track? It looks like it stops pretty close to the pin boundary, and I'd like to have the track continue into the center. I don't see an option for this in the I/O-placer or global/detailed router documentation, but I could be missing something:

What would be the purpose of extending further? It is connected either way

It may not be necessary to produce a working design, but the eFabless MPW tapeout scripts include a DRC which seems to produce errors if there is not enough overlap between 'pin' and 'drawing' data types on a metal layer. I think that it might use the center point of the pin data type as a reference, but I'm not certain.

@User do you know about this rule and what it is trying to capture?

@User: The PDK defines the pin purpose as independent of the actual metal. That is, if you have a pin purpose there must be drawn metal underneath. We had some people submit designs where they had only the pin purpose, which was creating an open circuit in the mask layers. So we added the check to make sure people wouldn't do that.

@User but you shouldn't have to cover the whole pin, that's excessive. The picture is a valid connection and shouldn't be flagged.

It isn't too much trouble to add 'drawing' data types as special net tracks with the same dimensions as the corresponding pins in the post-routing DEF file. That seems to satisfy the pin-overlap check, but if the extended connection is required for taping out, it might be nice to have a way to influence these dimensions with an OpenROAD command.

I don't think the router should be changed when it is producing legal layout. There is no LVS or DRC issue with this connection.

@User: I wasn't addressing the issue, I was explaining why there is such a check. In the screenshot above, the pin should have metal + pin layers, and the route should be metal, and the route should be okay as long as it touches the pin. And it all looks okay to me. Perhaps it's just a misunderstanding?

@User "the eFabless MPW tapeout scripts include a DRC which seems to produce errors" is what @User said which sounds like a problem with the tapeout scripts.

It sounds like there is a reason for the checks, though: others have pointed out that the previous MPW masks did not include pin data types for one reason or another. So while the DRC/LVS checks pass on this design, the actual die would have a small gap between the metal track and the macro boundary where there is only a pin data type.