I’ve noticed an odd issue with walkaround on KiCad 9 (not sure if it was an issue on 8). It seems like when trying to route a trace that goes past other pads on the same side as the originating pad on some ICs (maybe a pin spacing thing?), walkaround fails and it refuses to draw the trace.
I’m not able to upload the video here (new users can’t upload files), so here’s a streamable link: Watch KiCad Issue | Streamable
You can see the router not able to route a trace up the left side of the RP2350 (and RP2040), but it can on all other sides and on other ICs.
Anyone know what’s going on here? Is there a setting I can change to fix this?
Changing my track clearance to 0.1mm seems to fix this for this specific IC. Looks like the issue is that the clearance is exactly the same as the pin spacing on the RP2350, and that’s causing issues. Possibly a bug?
Possibly. But these “exactly” problems with widths and clearances are extremely difficult to solve. As far as I can see the problem is inherent to how computers work: number crunching speed and exact, non-rounded non-integer values are incompatible. The logic is very easy to humans to see: 1/3 * 3 = 1, but computers calculate it like 0.33333333333 * 3 = 0.999… That’s why the computer program may decide that for example an item with certain width which should fit into some gap with exactly the same width, doesn’t fit.
Typical solutions involve making some clearance or width a tad below the neat full number. For example, set a clearance to 0.1999 instead of 0.2, or track width to 0.1999 instead of 0.2. This makes us feel very uncomfortable, but computers don’t care.
For the record, the manufacturers who set the “hard” limits don’t usually care either. If you keep your min. clearances and widths above their min. possible values, everything is OK. And they also should have some room for rounding errors in their software because their computers use same technology than ours.