Print Mystery

While making the parts for my Pepper Cannon Accessories, I noticed the loader funnel had a birds nest of loose filament strung around right after the layer change. I clipped it off and sanded it down and it works fine, but what was the cause?

I theorized that slic3r printing inside perimeters by default might have provided very little support for each new layer since the funnel diameter was reduced at each layer.

So I checked the box telling it to print outside perimeters first on the theory that they would be fully supported by the previous layers.

The inside of the funnel did indeed get nice and smooth, but now the problem moved to the outside, seemingly disfigured by greyscale.

Here's the initial print with 4mm retraction and no retract on layer change:

That looked like retraction problems to me, so I turned retraction completely off to see what happens:

OK, now I'm getting stubborn and really want to know what is going on, so I read about bad prints following layer change on the internet and decide to try some of the things I read. Here is retraction back to 4mm, but retract on layer change turned on and the extrusion "primed" with an extra 0.1mm following retraction:

Possibly a little less of the blank expanse before the greyscale starts, so try a prime value of 0.3mm:

Not very impressive change. Another internet recommendation is printing at a lower temp, so I switched from 250 to 245:

Not much improvement there either. The final recommendation was to make sure the part had a wall thickness of an even multiple of extrusion widths. In my case that would need to be a multiple of 0.48 and 1.92 is the closest to the existing part. It was a little tricky to work out how to do this, my funnel model wants to know the thickness of the wall I desire perpendicular to the funnel wall, and now I want to set the thickness of a horizontal slice. I have the OpenSCAD echo the slice thickness, then plugged in different wall thicknesses till that echoed 1.92 (which took me from 1.5mm to 1.65122mm). Printed this new funnel:

Holy Cow! What a difference that made. Only defect is the Z seam now and I might be able to reduce that (but I'm tired of printing funnels for now, work on that some other day).

I should probably add a utility function to round up the wall thickness in my OpenSCAD funnel library.