I’m not actually saving the older files before I do the rebuild. And if I delete the pours and then redo them, I get the same weird shapes and islands showing up.
I can’t for the life of me work out what’s going on.
Weird. I don’t see the reason, yet, but I have V8.0.9 and V9.0.6 on my machine and can confirm that redoing the zone pour in V8.0.9 seems to work AOK, and redoing them in V9.0.6 makes the same goobers you’re seeing. It’s like there are conflicting overlapping zones or keep-outs or … but I can’t see them. Looks like an interesting puzzle question that’ll I poke more at if I get the time later this evening. If someone else doesn’t unravel it sooner.
Seems to be something in the inductive sled tracks / guard rail feature. In the V8 file, this feature has been make into a group and if I move this grouped feature off board, then the inner-layer pours seems to pour fine in V9.0.6. Another clue is the size of the group box- It’s much larger than just the supposed sled track loop feature suggesting there is something hidden or invisible grouped into it along with the visible features.
Anyway, i can confirm the misbehaving of the gerbers (9.0.6, Linux).
For a moment i could see glimpses of the construction curves that causes the problem but
i suspect some tricky to trigger bug is here…
For who would like to try here is the zipped project: NorthPoleCircuit_PCB.zip (1.5 MB)
i’ve tried some tricks but nothing apparent happened…
Thanks @DaveL & @Claudio.Lorini - at least I know I’m not going bonkers, and its a legit “feature”. I might reach out to the designer(s) and see if they can shed some light
Whatever the odd root cause turns out to be, this smells like a V9.0.6 bug as it seems this represents a design that opens and yields valid in V8, but not in V9. I believe the intent is that V9 should open and yield valid any and all V8 designs?
Trying to reduce / isolate this odd puzzle piece further, I’ve reduced it to focus on just one goober on one zone pour In2.Cu. When the short trace segments on In2.cu on the sled track are removed, the zone pour on In2.Cu works as expected. When they remain, there’s (3) distinct circular voids in the pour zone. Quite odd. I suppose I’ll move on to looking at the ASCII s-expression raw text file to learn more- If someone else doesn’t distill an root-cause answer sooner.
@DaveL - there is a Python script in the repo that this design was originally spawned from. That was used to generate the track layout that provides the forward/reverse propulsion.
Wondering if that’s introduced something into the KiCad file that v9.0.5 isn’t dealing with.
Those weird curves seem to be defining the track path in some way.
I am guessing the other way around.
All the anomalies seem to be from circles or arcs. My guess is that the python script has some latent bug that inserts too many arcs onto the PCB, and that these arcs do not show up in KiCad V8 (maybe they are malformed?). And then that KiCad V9 interprets the arcs in a different way, which makes them visible.
OK, looks like it something that has occured in the transition from V9.0.0 → V9.0.5 (or possibly an earlier version).
I can open it up and get it to work correctly on V9.0.0. Things go pear shaped with V9.0.5.
@RaptorUK - those 900+ errors drop down to about 5 warnings, that are just to do with footprints not matching what’s in the custom library - was easy enough to sort out.
Still not sure what’s causing the issue in V9.0.5 or V9.0.6, but at least I can work around it with V9.0.0 on my machine
It still smells like a bug to me. No matter how ‘special case’ the sled tracks were created in V8, the fact that it works OK with version 8.0.9 and apparently 9.0.0, but fouls up with V9.0.5 and V9.0.6 colors it as a bug to me.
Staying with focusing on just one layer (In2.Cu), I reduced the trigger to just (3) short segments. Each one of these short segments seems to trigger each one of the (3) phantom circular voids in the zone pour. With all (3) removed In2.Cu pours just fine. When I replace these short segments with a short free angle route, it also now floods fine in V9.0.6.
I expect I can reduce and isolate similar the phantom flood goobers on In1.Cu also, but I won’t bother unless it has value to someone.
Can you walk me through what you’ve done here @DaveL ?
I’m keen to get this resolved for V9.0.6, as all my attempts at installing V9.0.0 seem to prevent me also keeping V9.0.6 installed (which I need for actual projects that generate me an income!)
Not sure what you are referring to when you say “trigger”
By ‘trigger’ I mean identifying the thing that causes something. It’s basic troubleshooting that progressively divides the problem into smaller and smaller parts or groups until you’re at the smallest thing you distill it to. In this case, seems it’s a few to several small arc segments with a very shallow arc radius relative to length that V9.0+ isn’t happy with, but V8 and V9.0.0 was.
By trial and error and whittling the problem to the smallest, then redoing those that trigger the issue, see the attached. Give it a go in V9. Ho Ho Ho.
I’m still on KiCad V8 myself, and can’t do much here. However I noticed that: PCB Editor / File / Board Setup / Design Rules / Constraints / Copper to edge clearance is set to zero. And that is not good. This can (and will) lead to smearing of the copper during routing. Some PCB manufacturers even refuse such PCB’s (and rightly so).
Similar to what I’ve showed I did on In2. Identified the segments that seem to trigger the large circular phantom voids then deleted them and redid them. Note I didn’t replace any of the short segments with arc traces, just short free-angle segments. All of the traces in the sled rail are defined as arcs, but the arc radius is so shallow they’re effectively straight anyway. Esp the short segments.
BTW, I think this is cool little tech gadget, but not so cool that I’d make one myself. If there’s a place I could just buy one for <~$75 USD, (with sled) I’d be interested.
