kicad-developers team mailing list archive

[Wayne Stambaugh <stambaughw@xxxxxxxxx>]

On 3/30/2016 8:02 AM, jp charras wrote:
> Le 29/03/2016 15:35, Damiano Bolla a écrit :
>> Good morning you all
>>
>> Let me explain what happens to me
>>
>> I am using Kicad for the excellent schematics and PCB support and also using FreeRouter since it
>> allows free angle routing and an incredibly efficient push and shove
>>
>> What normally happens is that I place components on the board and then export import to/from Freerouter
>>
>> As far as I can see there are no issues on importing from FreeRouter to Kicad, however, the export
>> has one quirk than I am explaining now
>>
>> - Pads coordinates truncation: There are cases when pad dimensions/position is something like
>> 0.89916 and while exporting to DSN it gets "rounded"
>> When this is done Freeroute display an unconnected since the track does not end p exactly where it
>> should
>>
>> The question on Kicad side is this: Is there any of you top developers that think that this
>> could/should be fixed/addressed ?
>>
>> I could have a go at trying to address it, I just wish to avoid working if the pull is then
>> denied/ignored (happened before on other projects)
> 
> It could be worth to have a look at it.
> 
> What is strange for me is the fact the same coordinates are rounded for a pad location and not (or
> differently) rounded for a track end point.
> 
> Note also Pcbnew DRC does not request the pad location and the track end point have the same value.
> For many reasons, it requests only the track end point is inside the circle contained by the pad, to
> be seen connected.
> 
> Can you send us a board which exhibit this issue?
> 
>>
>> --------- Unrelated, if any of you are interested
>>
>> I am "adjusting" FreeRouting so the above is less of an issue, what I have now is a version that
>> allow the autorouter to "reconnect" the almost joined traces
>>
>> One key to know is that FreeRouting normally zap a point in a segment that is colinear , so a
>> segment has only a start-end point.
>> This is an issue if you have components that have pads on the same net and are colinear. Now, at
>> least, it tries to keep the further away points
> 
> Yes, because the track end point must be inside the circle contained by the pad, this is an issue
> for Pcbnew.
> 
> This constraint is due to 2 reasons:
> - calculation time.
> - practical reason: when moving/removing/editing a track segment, the fact the segment starts or
> ends on a pad avoid "strange" changes.
> 
>>
>> To join all the almost joined pads there is now a checkbox on autorouter that allows you to specify
>> if rip are allowed or not.
>> If you just do the autoroute pass with no ripoff what happens is that the autorouter reconnect the
>> points "almost connected" and does not mess up the board.
>>
>> What I am aiming to is to have an "automatic" push and shove on the clearance violations, should be
>> possible (it probably is 99% done), I just have to find a way to "enable" it. This is quite useful
>> whan you have to "adjust" traces and end up with some clearance violations in the board.
>>
>> ------------------------------------------------
>>
>> What is the feeling on the matter by you developers ?
>>
>> Thanks, Damiano
> 
> If you are working on Freeroute enhancements, thanks you and good luck.
> There are certainly a very few amount of guys who are able to work on Freeroute.
> 

I thought the reason that the Freeroute website was taken down was due
to licensing issues.  If memory serves, the Freeroute developer was sent
a cease and desist letter because he wrote the router code for a
customer.  I would be sure about the current Freeroute license before
spending a lot of time on it.  Your best bet would be to check with the
author before proceeding.