Accuracy vs Settings

[KeithA]

I have a new issue that has popped up. When I first started printing with my machine I was amazed on the accuracy of the prints. I was able to print with very little tolerance and pieces designed to fit together did without issue. I have since messed around with multiple settings in an attempt to tweak the setting for the best possible prints. I am still trying to settle on what works best overall but it seems that some of the items that used to just fit together are a little tight and no longer want to go together. Are an of the setting tweaks know to affect accuracy? Could it be something else? I have leveled the bed maybe it is ever so slightly closer to the tip as I used recommendations I found in this forum. How do you validate settings, is it the torture test from thingiverse or something else? Does one color print differently than another? I have noticed that I get a slight stringy effect more than I used to but haven't noticed if it was certain colors or in general. If anyone cares to share their setting file I would appreciate it. I am still printing in PLA using Makergear products. I need to settle on one and quit messing around and print.

Thanks,

Keith

[Toby]

For parts with tight tolerances the key is slowing the prints down. Once you've got bed-leveling and z-height done, I think that's the single most important factor. The default acceleration in S3D is crazy high. If you haven't already done it, add a line to your start script like:

M201 X900 Y900 Z200 E10000 ; limit acceleration

Other people use higher values, like 1500, but this works fine for me.

I don't mess with coast and extra restart settings. I think they just confuse the print. I have retraction at 1.15, extrusion multiplier at .94, bed temp 60, nozzle temp 215 (for makergear PLA). Unless my print absolutely demands very low layer height, I use .2 or .25 mm. If I do need .1mm I try to use multiple processes and isolate it on only that part of the print that needs it. Most of the time for me, problems in the print are fixed by slowing it down.

I'm also starting to believe that rapid changes in print speed aren't good. The problem is that filament in the hot end is fluid and can't slow down and speed up instantaneously the way a solid can. If print speed changes, you get pressure inside the fluid that does strange things to the motion dynamics of the filament. If you want to make sure a print is good, slow it down to 1500-2800 mm/min (depending on part size, larger prints faster than smaller one), and print with perimeter and infill at the same speed. With that as a baseline you can adjust infill up and perimeter down to make trade-offs between quality and speed.

Here's a print that just came off the bed. No adjustment of any kind. There are three parts that had to fit together: the inner cylinder is in two parts, and the outer shell snaps around and holds them together. The curves on this thing are funky. Each piece bends in all 3 dimensions (more mold making fun ).

I'll tack on the factory file for this print, but you always have to adjust things like layer height, infill and print speed to whatever you're printing at the moment.

trefoil.JPG (41.28 KiB) Viewed 14268 times

[jimc]

in addition to what toby said, over and under extrusion will cause either too large or small of a print. there are many calibration objects downloadable from thingiverse. the most common being a simple 10mm cube. print one and see where your at with a pair of calipers. lowering or raising the multiplier and even a larger difference is made by the filament diameter setting.

[jsc]

I've found designing mating surfaces with a .1mm clearance gives a tight fit and .2mm gives a looser but still snug fit. I know Toby doesn't believe in that whole "clearance" thing, but I believe he has a set of files....

Here is a blog post about hole and post errors (with graphs):
http://softsolder.com/2013/04/21/makerg ... t-objects/

Round holes and curves will tend to be undersized because of the nature of 3D printing. Circles are actually inset polygons, so are cutting corners on the circle. A better way is to use low count polygons for your hole needs.

[KeithA]

For parts with tight tolerances the key is slowing the prints down. Once you've got bed-leveling and z-height done, I think that's the single most important factor. The default acceleration in S3D is crazy high. If you haven't already done it, add a line to your start script like:

M201 X900 Y900 Z200 E10000 ; limit acceleration

Other people use higher values, like 1500, but this works fine for me.

I don't mess with coast and extra restart settings. I think they just confuse the print. I have retraction at 1.15, extrusion multiplier at .94, bed temp 60, nozzle temp 215 (for makergear PLA). Unless my print absolutely demands very low layer height, I use .2 or .25 mm. If I do need .1mm I try to use multiple processes and isolate it on only that part of the print that needs it. Most of the time for me, problems in the print are fixed by slowing it down.

I'm also starting to believe that rapid changes in print speed aren't good. The problem is that filament in the hot end is fluid and can't slow down and speed up instantaneously the way a solid can. If print speed changes, you get pressure inside the fluid that does strange things to the motion dynamics of the filament. If you want to make sure a print is good, slow it down to 1500-2800 mm/min (depending on part size, larger prints faster than smaller one), and print with perimeter and infill at the same speed. With that as a baseline you can adjust infill up and perimeter down to make trade-offs between quality and speed.

Here's a print that just came off the bed. No adjustment of any kind. There are three parts that had to fit together: the inner cylinder is in two parts, and the outer shell snaps around and holds them together. The curves on this thing are funky. Each piece bends in all 3 dimensions (more mold making fun ).

I'll tack on the factory file for this print, but you always have to adjust things like layer height, infill and print speed to whatever you're printing at the moment.

trefoil.JPG

Thanks for your response. I have messed around with acceleration but havent added it to a startup script which I have now done. I had retraction at 1 so I changed it along with the adjustment of the extrusion multiplier from .9 to .94. What do you use for extrusion width? How do I use the factory file you attached to this post? I wanted to look at it thinking I can see the remainder of the settings you typically use.

Thanks

[Toby]

I've been using the auto extrusion width. I haven't played with setting it manually, but I believe others have recommended setting it at a hair over your nozzle diameter, like .4- .42 for a .35 mm nozzle.

The factory file is what you load into S3D with ctrl-o or in the file menu, "Open." It's the whole S3D environment- models and settings. I don't usually save .fff files separately anymore, other than one or two basic settings for each type of filament. Instead I organize my work by the models I'm printing, and save as I go using ctrl-s in S3D. It's wasteful of disk space if your models are large, but it also allows you to keep customized things like how you orient the models on the bed and the start point locations if you set that numerically in S3D. I'm finding to get the seams where I want them I have to rotate the models around a bit, so it's useful to have that saved along with the other settings for those models.

[Toby]

I've found designing mating surfaces with a .1mm clearance gives a tight fit and .2mm gives a looser but still snug fit. I know Toby doesn't believe in that whole "clearance" thing, but I believe he has a set of files....

lol, it's not a religious conviction. hard to generalize a guess, but in many situations I've tried, no clearance works and is easier to model. tolerances don't play nice when you scale your models. but for the backscratcher I needed a .1mm clearance to get the snap fit joints to work.

[jsc]

Regarding extrusion width, I've found that setting it explicitly gives me more control when designing small parts. If you know that your extrusion width is going to be, say, 0.4mm, you can design your walls to be an integer multiple of that so they are made of solid perimeters instead of having a gap with some machine gun infill. Any value between 0.4 and 0.5 should work with a 0.35mm nozzle, with thicker values reducing print time but also reducing the resolution of the finest details in the x/y plane.

Regarding retraction and extrusion multiplier, I think there is a wide latitude of values that will work. I have retraction set to 1 and it works fine, no stringing. It will vary depending on material and temperatures. Extrusion multiplier, especially, varies by material. You should set your filament diameter correctly, then use extrusion multiplier as a slop factor to dial it in: low enough so that you don't get ridges in solid layers, but high enough so that neighboring filaments touch.

[ednisley]

How do you validate settings

This works for me...

Print the thinwall open box (http://softsolder.com/2014/04/18/revise ... on-object/) while adjusting things until:

• The wall thickness matches your preset thread width (0.40 mm, in my case)
• The height comes out at 5.00 mm
• It adheres firmly to the platform during & after printing
• It prints slowly enough to not slump into a puddle

Slic3r no longer does retractions on thinwall objects, but if you distribute five thinwall boxes across the platform (center & four corners), you can check overall platform leveling while also forcing retractions. The wall should be solid and there should be no stringing.

When that works, print Coasterman's (http://www.thingiverse.com/thing:5573) 20 mm solid box (http://www.thingiverse.com/download:17274) with about 25% infill while adjusting other settings until:

• All the dimensions match
• Bottom and top infill comes out solid
• Minimal perimeter zittage

Print the same box with 100% infill, which should produce the same results.

I used to use Coasterman's perimter-wt tester (http://www.thingiverse.com/download:17277) as a show-off piece, but the STL isn't manifold and, even after repairing it, Slic3r produces bogus G-Code. I should rebuild that with OpenSCAD.

At that point, you've got good dimensional accuracy and haven't spent half a lifetime screwing around with test pieces, which I count as a win.

I'm running Slic3r, but much of this should transfer to S3D:

• Force 0.20 mm thickness, 0.40 mm width (0.35 mm nozzle)
• Perimeter shells >=2, auto-add as needed
• Top/bottom layers = 3
• First layer 25 mm/s
• Infill & solid infill 150 mm/s
• Top infill 50 mm/s
• Internal perimeters 150 mm/s
• External perimeters 50 mm/s
• Small perimeters 50 mm/s
• Non-print moves 250 mm/s
• Retract 1.0 mm @ 60 mm/s
• Minimum 10 s/layer, fan if < 30 s, minimum speed 4 mm/s = 240 mm/min

That's with a somewhat tweaked M2 with a rigid platform, so the numbers aren't directly transportable to a stock M2. They should give you some idea of what works, anyhow.

To be fair, I'm a lot more concerned with dimensional accuracy and a lot less bothered by surface finish imperfections than the real craftsmen around here; they put my blocky stuff to shame!

[jsc]

Ed, all your settings seem perfectly reasonable to my eye, with one caveat: the differential between infill/internal perimeter and external perimeter speeds is quite high. I read on nophead's blog that he has moved to printing infill and perimeters all at the same intermediate speed, to avoid having large pressure changes in the hot end that may cause issues at the transitions. Have you ever noticed this to be an issue?