I'm having problems with thin walls where no infill is used.

I’m having problems with thin walls where no infill is used. The layers don’t stick together creating hollow walls. Can anyone give me a tip what I need to tune? Would making the filament width smaller (eg. 1.7 instead of 1.75) in the Slic3r make a difference? what would be a good value? 1.7? 1.6?

What version of slic3r are you using? The newer versions (i think starting with 0.9.7) try to fill those gaps. Decreasing filament width would indeed increase the amount of filament extruded, but would make for a quite messy and ugly print.
What i do to precisely calibrate the amount of filament extruded is this:
a) Set the “skirt height” to something like 15 layers. It should end up at least 2mm high.
b) Use a vernier/dial/digital caliper to measure the width of the skirt at a couple points and compare it to the “single wall thickness” that slic3r writes down in the beginning of every .gcode.
c) Adjust the “extrusion multiplier” to compensate for a non-ideal width. It’s a linear correlation: If the skirt is 0.6mm when it should be 0.65, set the multiplier to 0,65/0,6 ~ 1,08.

I’m using 0.9.9 but I don’t see a “single wall thinckness” value in the generated gcode. Is it “perimeters extrusion width”?

You’re right, it’s “perimeters extrusion width”.
Try 0.9.8, though, it should fill those gaps.

And you might want to check your axes for backlash, that infill looks like you have some in X or Y.

A single-walled object measured with a calibration tool such as a pair of calipers should help you dial in the perimeter extrusion width a little better.

I actually was wondering this the other day: does Slic3r account for speeds when calculating extrusion width? I’ve noticed if you go faster, extrusion changes due to die swell…

measuring a single wall (using the skirt trick Thomas mentioned) I get about 0.8 to 0.9mm instead of the 0.7mm Slic3r assumes. So we’re on the right track I guess. Now the question is: should I adjust the “extrusion multiplier” as Thomas suggested or directly set the “Default extrusion width” as Anthony seems to suggest?

@ThantiK The skirt essentially is a single-walled object. With a .stl, slic3r might try to make it the exact width of the walls in the .stl by adjusting flow, so there is the possibility that this would be somewhat inaccurate.The skirt will always be exactly one single wall.
I don’t think slic3r factors in speed, as die swell occurs a couple a fair bit away from the nozzle. With layer heights in the magnitude of 1/2 of the nozzle width, die swell should not be an issue. Additionally, slic3r makes the extrudion width a bit wider than the nozzle anyways - and of course, die swell does not affect the volume of plastic extruded.

@Andreas_Gohr the default extrusion width will not change the amount of filament extruded - if you change the extrusion width, the actual width will proportionally get thinner as well (volumetric logic). The gap you’re seeing will become wider as well, maybe even wide enough that the old slic3r will fill it with a single extrusion.
0,8 / 0,9 for a wall that’s supposed to be 0,7mm seems a bit extreme, though, and the exact opposite of what the infill appearance suggests.

@Thomas_Sanladerer yeah, I was kind of surprised by his calculated extrusion width myself as well. The gap in the wall, as well as the infill, seems to suggest that the width would need to go down, not up. (At least how I’ve always understood it)

I tend to believe that the optimal extrusion width is pretty much a fixed value for a given nozzle / hotend. But yeah, if Andreas wanted to adjust it to fill the center gap, it would need to be decreased to at least 1/3rd of the wall thickness of the object. Newer Slic3rs will not care about extrusion width and automatically decrease it for the center channel, while the older versions simply decide that the extrusion is too wide and would overfill it - hence the gap between the wall.
Edit: I just read your post again, @Andreas_Gohr , i misread it the first time for 0.9.0. Are you sure that you’re running 0.9.9, though?

I think what happens with Slic3r, is that it determines that the wall is too thin to get the nozzle into, and skips filling it in. I get this even on 0.9.9-dev from the github repo. Sometimes I fake my nozzle width down to 0.49 in order to get those areas filled.

@ThantiK that’s weird, since a couple of versions the thin wall gaps I’m seeing have drastically decreased. Maybe that’s a regression introduced after 0.9.8?

I’ve had no issues on my 0.35 nozzle extruding from 0.3 to 0.5, comes out fine. I’ve been using this to make strong thin walls. I also have up on measuring single extrude, if there is no gap left between the perimeters and the solid infill layers don’t make a rough surface with ridges pushed up or leave visible gaps, then it’s right.

I have a 0.35mm nozzle (well, my printer has, i don’t), and i did a test batch with 0.4, 0.45, 0.49 (default) and 0.55mm extrusion width at a 0.15mm layer height. Interestingly, the default setting came out as a winner in most of the tests, which included
-bridging 10-80mm
-overhangs 0-90°
-text quality of Mesquite 4-10mm
-4mm hole accuracy in the XY layer / XZ layer and 45° inbetween
-curling
-consistency of round parts
0.4mm clearly won the bridging test, but every other test either had a draw between widths or showed best results with 0.49mm.

Setting the extrusion multiplier to 1.02 seems to have fixed the problem. I’ll try 1.01 on my next print. Thanks for your help guys I learned something new today (for example that my calipers suck ;-))

From what I’ve seen, 0.5mm nozzles can make stronger parts than I can, but my gears come out amazing.

This is a slicing/skeining software bug. Some of them refuse to inject plastic into a gap that’s narrower than the nozzle.

I have this problem sometimes when I don’t calibrate my printer correctly before printing and the gap between the printhead and printbed is too high.

@John_Ridley Interesting! I just tried with a smaller nozzle size and the print came out perfect.

@Andreas_Gohr , do you mean smaller nozzle or thinner layer?