Slicers

I started off using Slic3r with my M90. I got reasonable results pretty quickly but, with using different plastics and more demanding models, things started getting difficult. In particular, I found it tricky getting the right amount of infill in the right places. Gaps would be left around holes. Some areas would be overfilled even though E steps and Extrusion Multiplier were set up perfectly.

I tried Skeinforge. I worked pretty hard at it but after 2 days (on and off), I still had problems, found it difficult working with ratios and found some things useless (cooling small x sections, orbit). Lots of time spent and not a lot to show for it.

Then I tried Cura (the latest, non-SF version). I transferred most settings from Slic3r. Suddenly things worked much better. A few minor tweaks and the gaps and overfilling disappeared. It also slowed down nicely for small x sections so things stopped getting into a gooey mess. Prints were generally much cleaner. It seems Cura controls the extruder much better than Slic3r for slow and/or fiddly bits.

Some things are not good though:

1. Skirts are printed inside out meaning that the object sometimes doesn't bond well to the inner perimeter of the skirt, negating the whole point of a skirt.
2. Internal and external perimeters seem to get printed in a "mad" way (the fan makes it impossible to see what's really going on). I'm not sure this makes any difference to the print though.
3. Filling of some perimeter areas depends on the width of the model's wall, the number of perimeters and the extrusion width. You can't set the extrusion width (as you can with Slic3r) but, from inspection, I guess the width is about 0.4mm). This is frustrating sometimes as you have to go back to the model and adjust wall dimensions to get the perfect slice.
4. Out of the box, Cura's bridges were worse than Slic3r. Now they're pretty good; I've no idea why as there are no settings specific to bridging that I can see.

I think I shall stick with Cura for a while as the results can be so impressive, without the time-consuming trial and error necessary with SF.

Anyone else trodden the same path?

I have been playing with Cura. It does make nice looking prints. For example the android printed with 0.2mm layers looks much nicer than the 0.3mm layer one I distribute. It does infill at the same speed as outlines which improves the quality.

Objects don't come out with the right outer dimension though. For example a 20mm cube comes out about 20.5 even though the flow rate gives perfect fill. Looking at the code it seems to use the flow rate for outlines as it infills, which is wrong. Oddly holes seem to come out right. I haven't had time to do a thorough investigation.

---

[www.hydraraptor.blogspot.com]

nophead Wrote:
-------------------------------------------------------
> I have been playing with Cura. It does make nice
> looking prints. For example the android printed
> with 0.2mm layers looks much nicer than the 0.3mm
> layer one I distribute. It does infill at the same
> speed as outlines which improves the quality.

Why does infilling at the same speed as the outline improve quality? Is it just because it slows it down?

Because there is no pressure change when switching from infill to outline. I actually ran it at 50mm/s so the outlines were faster rather than the infill slower. However that works less well on objects with sharp corners as there is a bit of ringing.

I think if I set up skeinforge to do the same I would see the same quality and have it dimensionally correct.

---

[www.hydraraptor.blogspot.com]

@nophead
For example a 20mm cube comes out about 20.5
Interesting... I tried a 1mm walled 20mm cube. No fill. ABS. It printed at about 10mm/s.
Result: x=19.87 y=19.87 z=19.7 The walls were smack on 1.0mm.

Looking at the code it seems to use the flow rate for outlines as it infills
Surely that's what you'd expect if Infill Speed is set to 0 (ie: use print speed for infill) Or had you set it to something else?

Edit: I don't know about SF but printing at 0.1mm layer height at 40mm/s or less is very impressive with Cura; sharp detail and no extraneous dobs or blots anywhere as with Slic3r. The one item I printed with a 10mm hole was about 4% undersized.

Edited 1 time(s). Last edit at 08/21/2013 08:45PM by threedyprinter.

Feed rate should be the same but low rate for outlines should be slightly less to get the right dimensions.

Mine was a solid cube with two outlines. Same with the Mendel90 calibration object.

---

[www.hydraraptor.blogspot.com]

Feed rate should be the same but low rate for outlines should be slightly less to get the right dimensions.
Maybe I misunderstand you but as the wall thickness comes out absolutely right (for me at least), I would think the feed rate / print speed is spot on. The external cube dimensions are, again for me, 6.5% under. I assumed either the thing had shrunk (doesn't appear so) or that the nozzle positioning was wrong, improbable though that may be. I'm off to look at the gcode and try a solid cube.

This is the maths I use to get things the right size: [hydraraptor.blogspot.co.uk].

Outlines are not constrained to be rectangular so they come out wider than they should be if the volumetric calculation assumes they are rectangles. This effect becomes less when layers are thinner compared to their width. I can only assume the rest of the world get things the right size by calibrating the thin wall size, using thin layers and putting up with slightly less than 100% infill. I have never measured the wall thickness since I wrote that article. I put that maths in my own software together with a scaling factor for shrinkage and things come out the right size. I put the same numbers into Skeinforge and the come out the right size with Marlin as well. All other slicers I have tried don't produce the same size objects with the same firmware configuration.

---

[www.hydraraptor.blogspot.com]

That's a great write-up of the maths, thank you.

I have never measured the wall thickness since I wrote that article.

So when you change filament you just measure the new thickness; extrude, say, 100mm and adjust some multiplier according to whatever is actually pulled off the reel, and that's it. No need to faff around with cubes, the calibration suggested by Slic3r ought to be redundant?

I poked my nose into the gcode of the 20mm cube. This is one layer of the outer wall.
G0 F5460 X-9.80 Y9.80
;TYPE:WALL-OUTER
G1 F2100 X-9.80 Y-9.80 E2.99806
G1 X9.80 Y-9.80 E3.19402
G1 X9.80 Y9.80 E3.38999
G1 X-9.80 Y9.80 E3.58595

The coordinates are as expected but can you tell me why the extruder lengths are different for every segment which are all the same length?

Yes all I do is enter the filament diameter and calibrate the E steps per mm for each machine and plastic. But as far as I can tell Skeinforge is the only slicer where you can do that as you can set the outline flow rate different from the infill. You have to change it manually to match the filament W?T ratio though.

The E values in your listing are absolute so the difference between them is constant.

---

[www.hydraraptor.blogspot.com]

Thanks for that info. Chris.

Just to complete the extrusion picture; am I right in guessing that Marlin takes the length of filament demanded, then computes an acceleration-hold-decelerate envelope for the x,y motors, then computes a similarly shaped envelope for the extruder motor that gives the correct feed rate to print speed ratio (line width) at all phases of the movement?

Yes it computes the acceleration to be used and keeps all the motors in sync. All moves are potentially 4 axis so I presume it works out the best case acceleration taking all axes limits into account.

It doesn't necessarily keep the flow in sync because there is lag in the extruder due to the liquid plastic being springy. There was an attempt to model that and compensate by Matt Roberts IIRC, but I don't think that is used much and I don't think the maths was correct.

---

[www.hydraraptor.blogspot.com]

Yes...

Right. So should the slicer be involved at all in setting extruder feed rates? We have your math for calculating feed rate/print speed, we have existing Marlin code for doing the fiddly motor driving. Wouldn't it be more efficient to put your math into Marlin and just have the slicer say "draw a line (or circle, arc...) of x mm width from A to B at speed V. Rather like Postscript I guess. I'm guessing that's just the way things happened rather than were planned?

The firmware would have to know the layer height, filament width and whether it was doing infill or outline and also do extra square root computations so better to keep those details in the higher lever software.

---

[www.hydraraptor.blogspot.com]

I guess so.

I just like the idea of lumps of s/w doing what they do best; slicers to slice and firmware to get the best out of a particular printer/filament combination.