X-Authentication-Warning: delorie.com: mail set sender to geda-user-bounces using -f X-Recipient: geda-user AT delorie DOT com Message-ID: <1389632128.2414.50.camel@AMD64X2.fritz.box> Subject: Re: [geda-user] New router pictures From: Stefan Salewski To: geda-user AT delorie DOT com Date: Mon, 13 Jan 2014 17:55:28 +0100 In-Reply-To: References: <1389363489 DOT 2427 DOT 5 DOT camel AT AMD64X2 DOT fritz DOT box> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.8.5 Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Reply-To: geda-user AT delorie DOT com Errors-To: nobody AT delorie DOT com X-Mailing-List: geda-user AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk On Mon, 2014-01-13 at 16:48 +0100, Kai-Martin Knaak wrote: > Stefan Salewski wrote: > > > Yesterday I put some new pictures on my page: > > http://www.ssalewski.de/Router.html.en > > Nice! > And most importantly: Good to see, the project receives serious attention. > I appreciate your efforts. Attention is low, as for most FOSS EDA projects. (OK, some exceptions exists, maybe raspberry pi, LUFA, and for KiCad there is at least some noise.) But I do not care too much about attention. > > Would it be possible to do only some connections? Or is the algorithm a > case of all-or-nothing? > Unfortunately the Router can be not used if there already exist traces on the layers it should use -- later we may define something like keep-out areas similar to large pads, but the router will not be able to interfere with manually drawn traces. That is very similar to Anthony's router and is a very basic habit of this type of routers. But of course it is possible to route the first part with the router and then manually modify the layout and insert the missing traces. Or it may be able to manually guide the router, i.e. by telling it some via positions or order of routing. Current routing strategy is straight forward, with a layer assignment algorithm as described in Tal Dayan's thesis. That one is good for preventing crossing of traces, but it takes not into account space blocked by pads and pins, which is necessary for dense populated boards. There exists many strategies and ideas to improve net ordering and via positioning -- all that is not difficult, but needs some thinking and testing. Maybe we should have a simple GUI which allows to insert additional rubber bands and to move vias -- that is very easy to do. As you may know I have staring a simple schematics editor two years ago http://www.ssalewski.de/PetEd.html.en a very simple application. I have just converted it to GTK3 (was easy) -- now I consider to extend it to basic layout editing. Unfortunately first I have to do very much cleanup work... I think you are one of the persons who have really used PCB program in the last 4 years (I did not, maybe DJ did) -- have you ever noticed the polygon bug reported some days ago by Gabriel Paubert? There seems to be no reply from other people, so my impression that no one is using PCB currently is supported unfortunately. I myself have no idea about polygon handling and gerber generation, it was my assumption that that was working correctly. Yesterday I found a problem report related to polygon dicer from 2008 http://t14292.cad-geda-development.cadtalk.us/yet-another-dicer-bug-t14292.html And do you have problems with OpenGL? I had a few times when I inserted layout data into an existing board, and also some days ago when I tested Gabriels bug. PCB was reacting very sluggish, even after restart of the program -- and also GLXGEARS was running very slowly after that. I had to restart my box. I hope this is only a problem for my box, maybe related to nvidia driver or Gnome shell. Can be hard to find. Yesterday I tested for 10 minutes without problems -- when I find a way to enforce that problem, I will compile PCB without OpenGL support, maybe with Lesstif GUI, and try again. Best regards, Stefan