ID:	5118
From:	rn...@igor2.repo.hu
Date:	Mon, 13 Sep 2021 11:45:55 +0200 (CEST)
Subject:	[pcb-rnd] new: route-rnd 0.9.0
replies:	5119 from Evan Foss <ev...@gmail.com> , 5127 from pc...@cuvoodoo.info , 5136 from rn...@igor2.repo.hu , 5144 from rn...@igor2.repo.hu
Hi all,
 
I've been working a lot on route-rnd lately and I am proud to announce the 
initial, experimental release, 0.9.0. It is testable, but is in alpha 
state. This version is intended for early testers and packagers (for 
test-packaging) and potential users to test the workflow, to generate 
feedback. This release is not yet for production usage.
 
web: http://www.repo.hu/projects/route-rnd
 
svn: svn://svn.repo.hu/route-rnd/trunk
 
Demo renders: http://igor2.repo.hu/tmp/route-rnd/0.9.0/
 
The demo page features two small demo boards. The left one is with pins 
spaced far away, the right one with more densely placed pins where often 
only one trace fits between two pins.
 
First the rats nest shows random networks. 
 
The second render is the output of the horver method. This method 
impleenmts a real trivial "assign one layer to horizontal traces, another 
to vertical, then escape nets to the side(s), place vias and use 
perpendicular lines to make the connections on the other layer" algo. 
While it does route some boards, the main goal was not to have a 
practically useful router algorithm but to have something simple that can 
be used to boot up the route-rnd infrastructure. However, you can still 
use it in case you have only a few parts on a largish board, vias are 
cheap and you can afford to have the net dispatching areas around the 
edges. This router is based solely on custom ideas.
 
The third redner is from TRBS, which tries to route by allocating 
triangulation edge crossings. This router is partly based on the Tal Dayan 
thesis (the layer assignment part) and partly based on custom ideas. I had 
high hopes about it but it turned out to be tricky on some corner cases. 
At its current state it generally yields a conceptually okayish route but 
in details it fails and runs traces too close to obstacles. At the moment 
it's on hold, I hope CRBS will perform better.
 
The final render is from CRBS. It is mostly based on the Tal Dayan thesis. 
It already routes many nets properly but there is one major feature 
missing (concave/corridor routing). I am working heavily on this one, a 
better version is expected before end of october.
 
 
Usage:
 
- you need the svn version of pcb-rnd (at least r35671)
 
- you need route-rnd (either from svn or from release tarball)
 
- you need to install route-rnd; can be a $HOME installation, but the 
executable needs to be on $PATH
 
- run pcb-rnd, load a board or place some footprints. Make sure:
 
  + you have nets (rats nest)
 
  + your current routing style has the right track width and clearance set
 
  + your pins/pads have the right copper size and clearance so tracks with 
the current routing style can pass between them
 
  + for now, prefer thru-hole pins; smd pads are not always handled 
properly
 
  + if you use the horver router, make sure your pins are aligned to a 
grid, preferrably of 100 mil
 
 
- open the external autorouter dialog: Connects menu, automatic routing, 
external autorouter
 
- if you had route-rnd on $PATH, it will automatically list all routing 
algorithms offered by route-rnd
 
Best regards,
 
Igor2
Mailing list archives : pcb-rnd

Navigation
