Slic3r settings

After analysing the Slic3r source code, I have tested a bit with different perimeter widths and layer heights, to see what form the plastic takes, and what effect that has on the real perimeter size. I reported on that earlier already in another thread, but after recent comments from Ray, I have done some more experiments, and I thought it would be better to start a new thread on it.
Many people have run the thin walls test described here, in order to set the extrusion multiplier of Slic3r. In my opinion, that test is fine, but it fails to mention that you need to set other parameters correctly before running the test. Especially the perimeter width is important, because the default perimeter width of Slic3r is not correct. Read on for more details on that.

In order to run the experiment, I printed a simple test piece consisting of just a thin wall. Thin walls are printed with a single perimeter width by Slic3r, the same perimeter width as is used for normal perimeters. Slic3r lays down the perimeter line half a configured perimeter width from the specified side. Therefore, if there is a difference between the configured perimeter width and the real perimeter width, you will not get correct sizes of your objects.

Here are the results of the tests. All have 20mm/s speed and are printed with the PLA that comes with the machine:

Layer   Nozzle Extrusion  Configured    Real         Notes
height  Width  Multiplier  Diameter   Diameter
0.25     0.5       1         0.4        0.55         Will be used when you specify 160% as perimeter width (it's 160% of layer height)
0.25     0.5       1         0.5        0.60         The Slic3r default perimeter width (equal to nozzle diameter), taken when you specify 0
0.25     0.5       1         0.75       0.75
0.25     0.5       1         1.0        1.00
0.25     0.5      0.83       0.5        0.50         The wall has tiny holes, indicating the extrusion could not keep up with the movement
0.25     0.5      0.83       0.75       0.70
0.25     0.5      0.83       1.0        0.90
0.25     0.3       1         0.5        0.55
0.25     0.3       1         0.75       0.75
0.5      0.5       1         0.5        0.65          Very fragile, the layers hardly stick together, small gaps between layers here and there
0.5      0.5       1         1.0        1.00

This shows that it is hard to achieve 0.5mm perimeters. The only sample that achieved that has not enough extrusion to keep up with the movement speed; it forms tiny holes where the filament is missing. The 0.4 perimeter sample had the same problem, by the way.
As you can see, I have experimented with nozzle-width as well, because the calculations in Slic3r are based on nozzle width. The shape of the plastic that is extruded is changed by it.

Of course, these are my results, and they can differ between different machines, different plastics, etc, and you should definitely run your own tests. But these results strengthen my advice:

• Set the perimeter width to at least the nozzle diameter + the layer height.
• Do that certainly before running the recommended thin wall test to determine the extrusion multiplier
• Set the extrusion multiplier to 1. Don't change it to fix other problems than incorrect extrusion amounts. The extrusion multiplier is a catch-all parameter that affects everything you print, and should be treated with respect.

Have you tried measuring your nozzle diameter? Mine prints fine at 0.5mm, but if I try printing finer than that, I also get holes/overthinning - my nozzle fits a 0.5mm twist drill pretty tightly (I had to use a drill to clear it once).

Also have you tried running at different temperatures? I just ran some tests today keeping the same slicing settings, but varying the nozzle temperature (the behaviour of the extrudate varies with temperature because of the effects on viscosity and elasticity), running ABS at 220°C, I get wall thickness of 0.55-0.60, and running at 250°C with the same slicing and speeds I get a wall thickness range of 0.44-0.51. I normally run at 235°C and the wall thickness with my standard settings at 235° is pretty much spot on (thinnest measurement 0.48, highest 0.53). I measured today by excising two or three layers and measuring them in a micrometer, to minimise banding effects and to allow a consistent pressure.

I also tried measuring the extrudate thickness (as in the slic3r manual, extrude in the air slowly and measure the extrudate diameter then use this for nozzle diameter), which was 0.56mm at 250°C , and I used this measurement for nozzle diameter with 1.0 as extrusion multiplier and got, running at 250°C, layer widths ranging between 0.48 and 0.51 (which was better than when using my default 0.5 for nozzle width in earlier runs at this temperature). The width of the extrudate increases with a decrease in temperature by the way (at 220°C I measure it at 0.64mm), because of the viscosity and elasticity, and presumably this extra thickness would make it harder for the printer to print to size, I'd be interested if knowing if your second test (in the table above) shows an increase in accuracy when you raise the extruder temperature by say 10 and 20 degrees (I haven't repeated the tests with PLA today but it seems more viscous and rubbery when printing than ABS and may show a greater temperature dependency than ABS shows for me).

Ray

Good input, Ray, I had not realised that the temperature might have an influence on the shape of the line that is extruded. I will certainly try with different temperatures as soon as I have the chance. I can only try with PLA, though, and currently I run that at 190 degrees.
The nozzle width is specified by RRP as 0.5mm, I did not think that it might be very far from that, but I can measure it by the method that you specified. But a small deviation from 0.5mm should not make a large difference in the end.

Quote
rayhicks
...and I used this measurement for nozzle diameter with 1.0 as extrusion multiplier and got, running at 250°C, layer widths ranging between 0.48 and 0.51

I assume you mean wall/perimeter width here. I am surprised that you can get 0.5mm width even with 1.0 extrusion multiplier, because of the way Slic3r calculates:

1. When perimeter-width is larger than nozzle-width + layer-height, it will assume the cross section of the extrudate will be a rectangular with two half circles on both sides, and a height equal to the layer height. It will extrude enough to make such a shape with the specified width
2. When the perimeter width is equal to the nozzle width it will assume the shape is a pure rectangle with the width of the nozzle and height of the layer (I have never seen a rectangular blob of plastic myself...)
3. When the perimeter is somewhere inbetween those values, it will use an extrusion amount somewhere inbetween as well.
4. Slic3r is not really prepared for a perimeter-width that is less than the nozzle-width: it simply uses the same algorithm as for 3), which cannot be correct

I do suggest you try some tests with 0.75 or 1.0 perimeter width. Due to the way the calculations are done, I would expect 1.0 width to be more reliable and consistent than 0.5 width. BTW, which layer width did you use in those tests?

Hi Flyskyhy, I used a layer height of 0.24mm for all of the tests (and a width of 0.5). Since the volume of extrudate is what the slicers try to control, I'd say that diameter errors (both filament and nozzle) are more important than length errors (since the volume is proportional to the square of the diameter, but only directly proportional to length).

I agree with your comments on the extrusion rate criteria (without looking at the code), the rectangle doesn't crop up as a cross-section on perimeters I've seen very often - are the amounts you refer to relative to x/y travel?

Ray

Quote
rayhicks
Hi Flyskyhy, I used a layer height of 0.24mm for all of the tests (and a width of 0.5). Since the volume of extrudate is what the slicers try to control, I'd say that diameter errors (both filament and nozzle) are more important than length errors (since the volume is proportional to the square of the diameter, but only directly proportional to length).

Fair enough, and the filament diameter is obviously important. But the nozzle diameter does not actually influence the volume of extrudate, except in the limited way I described above.

Quote
rayhicks
...are the amounts you refer to relative to x/y travel?
Ray

Not sure what you mean by that. I described the cross-section of the line that is extruded. When traveling faster in x/y direction, it will extrude faster as well to keep the same cross section of the line.