Net Length Report

[tradice]

A Useful report would be a net length report - my suggestion would be as follows:
a) Report sorted on Net Name
b) Component Pads that are part of this net (So far, this is right from the netlist)
c) IF the net has two components on the net:
   1) Straight line / Manhatten difference between centroids of pads
   2) Etch Length
   3) Etch / Manhatten ratio
   4) Number of vias
d) If the net has more than two components on the net
   1) Minimum straight line / Manhatten difference between centroids of pads
   2) Etch Length
   3) Etch / Manhatten ratio
   4) Number of vias
   5) If the members of a net go to polygons, what layer and which polygon

I would be interested if other users would want to add to this or comment - I am considering writing this as a utility to examine the netlist.

[Ineiev (ineiev)]

A hint on the starting point: there is Report(NetLength) action; it is not so advanced as you suggest, though. the relevant code resides in src/report.c.

I'm personally not very interested in this feature, but I think it's worth discussing on geda-user when you have a patch.

[rikster (rikster-users)]

For large boards I think this would be a good addition. I use Report(AllNetLengths) along with naming conventions to check, for example, that differential pairs have the same length to the tolerance I require. Also for differential pairs I may restrict the signal to route without vias. Your report would be useful in verifying that the via count on a differential pair is zero.

Finally, it would also be good for characterizing the quality of the autoroute. If more than 50% of the nets are over some threshold, say 150% of Manhattan distance, then I might investigate. I would certainly want to see any outlier nets that were routing above 200% of Manhattan. This might indicate that they had been routed along the edge of the board creating a huge inductive loop.