Extrusion Width vs Nozzle Width

[Ryan Carlyle]

I'm in the camp that extrusion width should be the same as nozzle orifice width, unless you have a compelling reason to do otherwise, such as making thin-wall prints where wall thickness is close to the nozzle size and need to convince the slicer to fill the space. But LOTS of people think extrusion width should be larger than nozzle size. Why? What's the rationale?

[whosawhatsis]

It's about the shape of the cross-section of the extruded thread. No matter what you do, it's not going to be rectangular. If your extrusion is as wide as the nozzle diameter plus the layer height, you will get a rectangle the width of the nozzle plus a semicircle with the radius of the layer height on each side. Any thinner than that, and the shape will be less certain, and will have a contact area between layers that is less than the nozzle width.

In theory, you could do your first perimeter and infill at this width, then make subsequent passes thinner as long as they are adjacent to the first one, but it would probably require better modeling of the extrusion characteristics than current slicers have.

Empirically, I find that it makes prints that are stronger, more dimensionally accurate, and (particularly with translucent filaments) more aesthetically pleasing.

On Thursday, April 16, 2015 at 18:40, Ryan Carlyle wrote:

I'm in the camp that extrusion width should be the same as nozzle orifice width, unless you have a compelling reason to do otherwise, such as making thin-wall prints where wall thickness is close to the nozzle size and need to convince the slicer to fill the space. But LOTS of people think extrusion width should be larger than nozzle size. Why? What's the rationale?

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/b79269f1-c84f-4c57-9eac-29818018e4bd%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Petr Ptacek]

Let's assume, that you will set up extrusion as wide as the nozzle diameter plus the layer height, and you will get a rectangle the width of the nozzle plus a semicircle with the radius of the layer height on each side. Well, now let's assume you will print with 100% infill. Two thinks may happen. Either slicer is dumb and will still print with nozzle width and you will be over-extruding, which makes it totally useless. Or slicer is smart (S3D) and will honor wider extrusion, but then you will get same "gaps" as before between semicircles. Yes, your extrusion is wider, but does it really matter so much? I don't know. I'm not so convinced yet. I clearly see benefit in vase mode, but that's pretty much it.

[whosawhatsis]

None of the modern slicers are dumb enough to ignore the extruded volume when calculating the spacing of lines for solid fill.

Also, I've read arguments among slicer developers about how they model the shape of extrudate when it is thinner than nozzle bore + layer height. Some of them are definitely wrong, and it's likely that none of them are right.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/b28d501e-86bd-40f7-91f1-4bf828b48ee9%40googlegroups.com.

[Ryan Carlyle]

What makes you think it comes out as a semi-circle, as opposed to any other type of rounded bulging shape? It sounds like that assumption is wildly over-simplifying the polymer rheology and physical system complexity. This isn't laminar flow of a Newtonian fluid between two stationary plates, it's transition flow of a viscoelastic polymer with crystallinity and memory effects, that just experienced a massive drawing strain through the nozzle and thus is trying to relax residual axial strain, and is being squeezed into a roughly crescent-shaped gap at the strand front between a cold stationary surface and a hot moving surface. That's an INSANELY complex shear stress/strain environment.

Then, assuming it does come out roughly semi-circular, why is that a good thing? The more the cross-section bulges, the less likely you are to fill the corner voids when the adjacent strand goes down. More bulging = larger corner voids.

I can completely believe that wider extrusion widths are better for part strength, but to my mind, you'd be much better off using a larger nozzle to accomplish that, so you don't have as much side bulging or nozzle back-pressure.

[Ryan Carlyle]

I think this issue is crying out for some empirical test data. The flow regimes are too complex to intuit or predict the answer. (Unless somebody's got a supercomputer model based on Dan's PhD thesis that I'm not aware of.)

Any thoughts on the best way to cut a print cross-section for examination? The academic papers I've seen on extrusion cross-sections printed a part with constant fill angle (eg all strands stacked aligned), cut fracture initiation sites, froze the part in liquid nitrogen, hit it with a hammer to get a clean break, and used a scanning electron microscope on the exposed crack face. (A buddy of mine worked on this paper: http://www.stratasys.com/~/media/Main/Files/FDM%20Test%20Reports/Process-induced%20Properties%20of%20FDM%20Products.pdf)

This was printed on an old Stratasys Fortus and it's surprisingly weak -- our modern hobbyist slicers, when calibrated properly, overlap the strand pitch so they get much more edge contact than this. And rotating the strand alignment further helps fill the corner voids quite a bit.

Another option for examining porous microstructure is to wick or vacuum fill a colored glue/epoxy into the void spaces, then precision mill the face down. We do that in the oil industry to look at rock microstructure. Sounds a lot harder than the "freeze and hammer" option though. 

[Ryan Carlyle]

Turns out, PLA is brittle enough that I can cleanly "score and snap" a small test sample without any difficulty. I'm trying different colors now to see what gives good contrast under a microscope. Black has terrible contrast but you can kind of see how little void space there is: 

This was sliced with S3D, printed with a Replicator 2 style all-metal hot end and QU-BD Makerbot style nozzle, 80mm/s feed, 0.4mm nozzle, 0.4mm extrusion width, 0.2mm layer height (giving 6.4 mm^3/sec), 100% infill, 0% fill/shell overlap, black Protoparadigm PLA, printed at 212C, with some light blower cooling. Extrusion volume was calibrated for part strength, ie slightly over-extruded. 

[Ryan Carlyle]

Brown is a lot better. Same print settings, except no print cooling.

White shows the voids pretty well. It also has way more/bigger voids, because its diameter is a little smaller and I didn't bother reslicing with proper volume calibration.

[Ryan Carlyle]

Aaaand here's the exact same print settings, except with 0.8mm extrusion width and 40mm/s feed. Filament flow rate is nominally the same.

The doubled extrusion width causes half as many corner gaps but they're larger. I'm prepping some side by side comparisons. 

[Ryan Carlyle]

Ok. These are all a bit under-extruded to make the shape of the void spaces really clear. All are printed at 6.4 mm^3/sec, through a 0.4mm nozzle, Esun white PLA, 212C, infinite perimeters. Closeup shots are rotated so "up is up."

0.2mm layer, 0.4mm extrusion width:

0.2mm layer height, 0.8mm extrusion width:

0.4mm layer height, 0.4mm extrusion width (yes, this is a bad idea, I just wanted to know what it looked like)

[whosawhatsis]

Just to clarify, by my formula, the extrusion width for a .2mm layer with a .4mm nozzle should be .6mm, not .8mm. The same should hold true as long as the extrusion width does not exceed the layer height plus the diameter of the flat on the end of the nozzle, but of course making the extrusion unnecessarily wide impares the ability to print small features in the X/Y plane.

These images are showing pretty much what I expected. When your extrusion width is less than my formula demands, the plastic comes out as an inverted dome. The nozzle-facing side is pretty flat, but the other side is rounded so that it has only a small contact area with the previous layer. This is bad for layer adhesion, for optical properties when using translucent filaments, and the less predictable bulging of the sides is bad for dimensional accuracy of the surfaces.

Seeing these images makes me wish even more that we had slicing algorithms smart enough to print perimeters on alternating half-layers so that they would pack hexagonally rather than rectangularly.

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/848fb2d0-2f54-4d4a-bd5c-e59ee27f8bfe%40googlegroups.com.

[Petr Ptacek]

Ryan, what would be useful to do is to pick two different nozzles, let's say 0.3 and 0.5, then print 0.2mm layer with 0.3mm nozzle with 0.5mm extrusion width and with 0.5 mm nozzle 0.5mm width. That would be the test I want to see to make any conclusion.

But from what I see with 0.8mm extrusion somewhat corresponds with what you said in the beginning. There is going to be very little difference between printing 0.2mm layer with either 0.4mm nozzle and 0.6mm extrusion or 0.6mm nozzle printing 0.6mm extrusion.

Other then using stock nozzle to print wider extrusions to gain strength at the cost of speed and possibly more warping, I still do not see benefit so far.

[Ryan Carlyle]

On Friday, April 17, 2015 at 6:40:23 PM UTC-5, Whosa whatsis wrote:

Just to clarify, by my formula, the extrusion width for a .2mm layer with a .4mm nozzle should be .6mm, not .8mm. The same should hold true as long as the extrusion width does not exceed the layer height plus the diameter of the flat on the end of the nozzle, but of course making the extrusion unnecessarily wide impares the ability to print small features in the X/Y plane.

These images are showing pretty much what I expected. When your extrusion width is less than my formula demands, the plastic comes out as an inverted dome. The nozzle-facing side is pretty flat, but the other side is rounded so that it has only a small contact area with the previous layer. This is bad for layer adhesion, for optical properties when using translucent filaments, and the less predictable bulging of the sides is bad for dimensional accuracy of the surfaces.

Seeing these images makes me wish even more that we had slicing algorithms smart enough to print perimeters on alternating half-layers so that they would pack hexagonally rather than rectangularly.

Right, I just wanted to get the extremes documented before looking at intermediate values. Here's 0.2mm height by 0.6mm width, it looks qualitatively identical to the 0.2x0.8 print:

[Ryan Carlyle]

I think I actually prefer the inverted dome shape you get with width=nozzle over the squashed oval with width>nozzle. The domes cause more overlap between adjacent strands. That should inhibit cleavage between aligned perimeters because it's more like a zipper.

To my eye, the interlayer contact percent is about the same, if not actually better for the width=nozzle case. (These two pics are normalized to the same scale.)

[whosawhatsis]

Remember that you're intentionally under-extruding here. That overlap is caused by the adjacent line intruding into the current lines's area, which should not be happening when under extruding like this. Your shape will push the whole extrusion to the side then this happens, which will cause the line to be too far to the side, with the same gap under the rounded edge. With the .6 profile, the plastic will touch the layer below before it touches the side, and has a better chance of filling in under the adjacent line's curve, as well as above it.

Either way, as long as the total width is less than the width of the flat on the end of the nozzle, sufficient plastic will be pressed to the sides enough to fill in the cleavage on the top surface (and more than that will create artifacts sticking up), but my way will have fewer/smaller internal gaps.

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/01578a3c-292b-4835-8394-3aa0ae8da1c0%40googlegroups.com.

[Ryan Carlyle]

I'll try it again with "proper" volume calibration tomorrow and see how it looks. I don't think it'll be drastically different though. This isn't actually under-extruded by much, relative to the typical calibration range.

Can't do any more tonight because my printer has been co-opted by an 8 hour Cookie Monster print. Family requests trump experiments :-)

[Ryan Carlyle]

On Friday, April 17, 2015 at 7:44:56 PM UTC-5, Whosa whatsis wrote:

Remember that you're intentionally under-extruding here. That overlap is caused by the adjacent line intruding into the current lines's area, which should not be happening when under extruding like this. Your shape will push the whole extrusion to the side then this happens, which will cause the line to be too far to the side, with the same gap under the rounded edge. With the .6 profile, the plastic will touch the layer below before it touches the side, and has a better chance of filling in under the adjacent line's curve, as well as above it.

Another data point... the 0.2x0.4 print is 7.93mm from a nominal 8.0mm. The 0.2x0.6 print is 5.91mm from a nominal 6.0mm. I also did another 0.2x0.4mm print with 4% more extrusion volume and it came out 8.02mm from a nominal 8.0mm. Visually, that last print is exactly what most people call perfect for volume calibration. You'd have to over-extrude to fit any more plastic into the corner voids. 

One thing I suspect is that you need to calibrate extrusion volume differently when you change extrusion width. Which makes sense, from a back-pressure standpoint. More extrusion width means you're pushing more rapidly-cooling plastic away from the nozzle, even though I'm normalizing feedrates for constant flow rate. So you should expect to have a bit more back-pressure and extrude a little less volume for the same nominal flow rate. 

[whosawhatsis]

That flow multiplier vs. back-pressure relationship is highly dependent on the drive gear you use. I've seen filament pulled from an Ultimaker (knurled drive) that clearly had about a 2.5x difference in the spacing of the tooth marks with only extrusion speed changed, but I've done similar tests with Deezmaker drive gears and was unable to get any measurable effect before the back-pressure got high enough to make the motor skip.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/5a87189f-eb40-4401-a21a-e354d9bc9fa5%40googlegroups.com.

[Ryan Carlyle]

I'm using a genuine Makerbot mk7. It does lose a little volume at high rates, like all drive gears/hobs (simply because of filament elasticity), but I don't think it's anywhere near as much "slip" as the UM knurled bolt. That thing is just silly. 

That's another dataset I'd like to rigorously measure at some point. Compare hob geometry vs slip. Really though, the best way to do that would be with a constant-force rig, not pushing through a nozzle. Too many minimally-controlled variables when you push through a nozzle. I'm imagining an extruder pulling on filament in open air to lift a weight. Like, tie 2,4,6,8 kg weights to the end of a couple meters of filament and see how much filament length actually gets lifted vs the commanded linear travel. 

On Friday, April 17, 2015 at 9:38:37 PM UTC-5, Whosa whatsis wrote:

That flow multiplier vs. back-pressure relationship is highly dependent on the drive gear you use. I've seen filament pulled from an Ultimaker (knurled drive) that clearly had about a 2.5x difference in the spacing of the tooth marks with only extrusion speed changed, but I've done similar tests with Deezmaker drive gears and was unable to get any measurable effect before the back-pressure got high enough to make the motor skip.

On Friday, April 17, 2015 at 19:30, Ryan Carlyle wrote:

On Friday, April 17, 2015 at 7:44:56 PM UTC-5, Whosa whatsis wrote:

Remember that you're intentionally under-extruding here. That overlap is caused by the adjacent line intruding into the current lines's area, which should not be happening when under extruding like this. Your shape will push the whole extrusion to the side then this happens, which will cause the line to be too far to the side, with the same gap under the rounded edge. With the .6 profile, the plastic will touch the layer below before it touches the side, and has a better chance of filling in under the adjacent line's curve, as well as above it.

Another data point... the 0.2x0.4 print is 7.93mm from a nominal 8.0mm. The 0.2x0.6 print is 5.91mm from a nominal 6.0mm. I also did another 0.2x0.4mm print with 4% more extrusion volume and it came out 8.02mm from a nominal 8.0mm. Visually, that last print is exactly what most people call perfect for volume calibration. You'd have to over-extrude to fit any more plastic into the corner voids. 

One thing I suspect is that you need to calibrate extrusion volume differently when you change extrusion width. Which makes sense, from a back-pressure standpoint. More extrusion width means you're pushing more rapidly-cooling plastic away from the nozzle, even though I'm normalizing feedrates for constant flow rate. So you should expect to have a bit more back-pressure and extrude a little less volume for the same nominal flow rate. 

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+unsubscribe@googlegroups.com.

[whosawhatsis]

Yeah, that'd be the way to test it.

The Ultimaker's knurled drive does apparently have the advantage that the teeth don't clog with plastic when it strips, and thus need to be cleaned out. You don't want it to strip, but if the bits don't stay in the teeth, it's easier to recover and requires less maintenence.

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.

To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/05a5778d-723d-472a-b0a0-8f0643244b14%40googlegroups.com.

[Ryan Carlyle]

Alright, more photos. This time, I calibrated extrusion volume to the level I would typically use for optimal part strength. Basically the most infill you can use without messing up the part. The 20x20x10 calibration box tops were bulging a bit. I added 2% volume after printing these: (left is 0.4, right is 0.6)

Then I printed the test beams. All perimeters, no fill, constant 6.4 mm^3/sec flow rate for both.

• Nominal test beam dimensions: 9.6m wide by 10mm tall (this is an integer multiple of strands for both prints)
  
• 0.2mm layer, 0.4mm width: 9.73 x 10.32 mm
  
• 0.2mm layer, 0.6mm width: 9.71 x 10.31 mm
  

I'd call that identical, for all practical purposes. There's more external variation within each test piece than there is between test pieces. I measured in the same location on each test sample. They both got a little wider with height as the over-extrusion built up. 

The problem I ran into is that I couldn't get a clean break on the 0.6mm width piece in two attempts. The layers delaminated and bent rather than making a clean break. I printed both with identical settings, scored both the same way, broke them at the same temp, and used the same tools and bending technique. This was a pretty consistent trend across all the test pieces -- wider extrusion test bars broke less cleanly.

This means either one of two things is true:

• Maybe wider strands have more strand toughness, so they resist fracture propagation during brittle tension failure (perhaps due to a different degree of polymer alignment / crystallinity -- PLA is weird for this sort of thing)
• Maybe wider strands have less interlayer adhesion, so they delaminate easier

Between the two, I'm strongly leaning towards the latter. The layer bonding just feels weaker to me. My proposed mechanism for this is that the plastic cools rapidly as it touches the existing lower layer, so any extra material squeezed out the side is too cool for good diffusion bonding. But I'm breaking these by hand so it's hardly a rigorous test. This is where a four-point test machine would REALLY come in handy. 

What's really weird is the interlayer surface where the 0.6mm piece delaminated. The top surface of each layer is very convex. This is true all the way to the top of the print. It's printing a banana shape, not a squashed oval. My best guess is that surface tension is rounding off the top layer after the nozzle moves away. 

There is zero visible gap between those strands. That's high in the print where the over-extrusion had accumulated a bit. But they delaminated remarkably easily. I'm trying to break apart the pieces right now, and the 0.6mm print comes apart way easier than the 0.4mm print. It's got to be some kind of filament temp difference or something. 

0.2mm x 0.4mm pics:

0.2mm x 0.6mm pics:

Scaled side by side shots:

Reposting the under-extruded version for side by side comparison:

[Chris P]

I don't have the technical background to comment intelligently on this, so this could be a dumb idea...

It seems plausible to me that forcing a wider extrusion bead out of the nozzle is causing more shear stress in the molten material.  Could higher levels of internal stress cause it to bond more poorly to the neighboring strands?

[Ryan Carlyle]

Well, it seems possible. PLA is a really goofy material for stress/strain effects and crystallinity and such. (Look at the heat capacity curve from DSC experiments and you see multiple small crystallinity phase changes with history-dependent hysteresis.) It does form localized zones of aligned polymer strands that have crystal-like behavior when it's strained. So it's possible the added shear from higher extrusion width causes more alignment and that's somehow bad. Fewer free polymer ends available to diffusion-weld the layers together? But that's extremely speculative. Could be something exactly opposite... Maybe printing with width=nozzle keeps the polymer strands MORE aligned because they merely have to take a right turn out the nozzle, so you get more alignment between layers, and that's somehow good. Who knows.

[whosawhatsis]

Keep in mind that plastic extruded into air will swell to a wider cross-sectional area than the nozzle bore (how much larger depends on a lot of factors, including extrusion speed), but if you're printing with nozzle width and a layer height less than pi/4 times that width (which is about the maximum layer height you would ever think of printing), you're stretching it to a smaller cross-sectional area than the nozzle. Logically, you would think there would be the least stress in the plastic if the cross-sectional area of the thread you lay down is equal to extruding at the same speed into air, which would make an extrusion wider than the nozzle better. This, of course, would be particularly true with amorphous materials and ones that are prone to warping, like ABS.

On Saturday, April 18, 2015 at 16:17, Ryan Carlyle wrote:

Well, it seems possible. PLA is a really goofy material for stress/strain effects and crystallinity and such. (Look at the heat capacity curve from DSC experiments and you see multiple small crystallinity phase changes with history-dependent hysteresis.) It does form localized zones of aligned polymer strands that have crystal-like behavior when it's strained. So it's possible the added shear from higher extrusion width causes more alignment and that's somehow bad. Fewer free polymer ends available to diffusion-weld the layers together? But that's extremely speculative. Could be something exactly opposite... Maybe printing with width=nozzle keeps the polymer strands MORE aligned because they merely have to take a right turn out the nozzle, so you get more alignment between layers, and that's somehow good. Who knows.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.

To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/8a4b1a74-6b8f-4678-8aa6-c6182f41ea8e%40googlegroups.com.

[whosawhatsis]

Come to think of it, it makes a lot of sense that the thinner extrusion width would snap while the wider one would bend. Because the thinner extrusion is being stretched as it leaves the nozzle, the polymer chains are being pulled into alignment. When you pull on these, they reach maximum elongation very quickly, and then snap. With a wider extrusion, the polymers have some curve, and thus have more elasticity because they have to be pulled straight before reaching maximum elongation. When you bend the print, the lines inside the bend are being compressed while those on the outside are being stretched, so the less elastic print will snap off more cleanly.

Of course, this difference has more to do with the cross-sectional area relative to the nozzle bore than cross-sectional shape, or whether it's been pushed out to the sides of the nozzle, so printing thinner layers should have a similar effect to printing narrow lines.

These characteristics could be useful. There may be some prints that you would want to be more stiff and brittle, and others that you want to be more flexible and resilient, and adjusting the extrusion width would allow you to choose between those properties.

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/a7b14548-d4a9-4c88-99e4-d23157126143%40googlegroups.com.

[whosawhatsis]

Come to think of it, this probably also explains why I sometimes see filament that is flexible in some places, but will snap if you look at it funny in other places. It's could just be a difference in how much it was stretched leaving the extrusion die.

OTOH, I still suspect that the difference has to do with the variable concentration of the plastic from the colored pellets (which are not PLA/ABS) that are added to color the plastic (this is supported by the fact that I don't recall ever having this problem with natural-colored filament).

[Ryan Carlyle]

On Saturday, April 18, 2015 at 6:29:02 PM UTC-5, Whosa whatsis wrote:

Keep in mind that plastic extruded into air will swell to a wider cross-sectional area than the nozzle bore (how much larger depends on a lot of factors, including extrusion speed), but if you're printing with nozzle width and a layer height less than pi/4 times that width (which is about the maximum layer height you would ever think of printing), you're stretching it to a smaller cross-sectional area than the nozzle. Logically, you would think there would be the least stress in the plastic if the cross-sectional area of the thread you lay down is equal to extruding at the same speed into air, which would make an extrusion wider than the nozzle better. This, of course, would be particularly true with amorphous materials and ones that are prone to warping, like ABS.

Hmm, good point. Small extrusion strands have to accelerate the linear velocity as they leave the nozzle causing drawing out of the polymer, while big extrusion strands decelerate. Both are probably doing some interesting things as far as shearing/straining. These round number height/width combos are roughly "equal area" with the orifice of a 0.4mm nozzle:

• 0.3mm high, 0.4mm wide
• 0.25mm high, 0.5mm wide
• 0.2mm high, 0.6mm wide
• 0.15mm high, 0.8mm wide

Question is, should we care about the relative areas? Is it better for mechanical properties to use under-sized extrusion strands, equal-area strands, or over-sized extrusion strands? It's not obvious to me what kind of polymer alignment and residual stretching/shear would be good or bad for the resulting print strength, elasticity, etc. It's a little more complex than simple property anisotropy, like you'd see in something with a traditional aligned grain structure like wood, because PLA exhibits crystalline behavior that might actually strengthen it when the polymer gets aligned. 

[whosawhatsis]

That is interesting, and PLA may or may not crystalize in a way that increases strength when you draw it out, but it seems unlikely that ABS would, and it should be stronger and less prone to warping (as well as things like constricting holes) when the polymers are stretched less, right?

Even if PLA does crystalize when you draw it out, the higher flexibility of a larger extrusion area is probably preferable for certain applications.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/1e6a441e-ad63-4d77-94f4-aed82153966d%40googlegroups.com.

[Ryan Carlyle]

One super interesting aspect of high residual shear stress potentially being locked into extruded strands is that it wouldn't just increase or decrease ABS warping... it would skew the stresses. If we assume the extruder always does perimeters in the same direction, any locked-in nozzle shear stress will be aligned with / amplify thermal contraction stress at the first half of each straight line segment, and counteracting thermal contraction stress at the second half of each straight line segment. 

Of course, this assumes the shear stresses sum up nicely. I don't know if that would happen. 

[whosawhatsis]

Well, for the most part, stresses in the first two diagrams should be perpendicular to one another rather than colinear, but to the extent that they are colinear, it would seem to suggest that you should see a very slight twisting of the print when the perimeters are all printed in the same direction but not if you alternate perimeter directions.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/97e27273-ff97-4fdc-909c-9b6d0d28674d%40googlegroups.com.

[Ryan Carlyle]

On Saturday, April 18, 2015 at 8:23:08 PM UTC-5, Whosa whatsis wrote:

That is interesting, and PLA may or may not crystalize in a way that increases strength when you draw it out, but it seems unlikely that ABS would, and it should be stronger and less prone to warping (as well as things like constricting holes) when the polymers are stretched less, right?

Even if PLA does crystalize when you draw it out, the higher flexibility of a larger extrusion area is probably preferable for certain applications.

Hmm, this is all very plausible, but I think I'd like to see some rigorous test data before drawing any hard conclusions. There are a lot of confounding factors here. For example, cooling rate is hugely important to whether residual stresses get locked in or shake out before hitting the glass point. Even just letting the PLA prints sit for a few days might make a big difference, since it creeps at room temp but that creep strain relaxation actually counts towards the elongation at break required for ultimate failure. (Absolutely terrible mechanical property, by the way. PLA is simply god-awful for highly stressed applications.) Would these properties appreciably change if I ran my HBP at a lower or higher temp? What about turning on the blower fan? Natural vs white PLA?

It WOULD be really neat if you could control the part's toughness/stiffness by changing extrusion height/width. 

ABS may or may not behave similarly. It's an odd duck as far as printing filaments go, since it's more or less a solid suspension composite of stiff acrylonitrile-styrene plastic and flexible butadiene rubber. (PLA/PHA is the same type of microstructure and it's wildly different from vanilla PLA.)

[Ryan Carlyle]

From the test data I've seen, the thermal contraction is primarily axial along the strand. Likewise the large-scale print's young's modulus is much higher axially along strands.. The combination of thermal contraction and material stiffness is what generates warping stress. So the contraction shear stress is primarily along the strands, not necessarily inward to the center of the print.  

[whosawhatsis]

It might be interesting to try your break test with two sets of prints, one of which has just cooled after printing while the other has been sitting around for a week, but is otherwise identical.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/323d96ce-6ef9-4163-a062-597a1bb11113%40googlegroups.com.

[Ryan Carlyle]

I'm reprinting the same test beams at 25% speed now. So, should be 1.6 mm^3/sec. It's a 1hr 45min print, so I probably won't do a lot of these...

I printed these yesterday and forgot to break them. So they're aged about a day. No appreciable difference in break strength. 

02x05, under-extruded:

02x07, under-extruded:

[Ryan Carlyle]

Just because I think it's nice to have them side by side, here's a shape comparison for all the different widths:

[Ryan Carlyle]

One week followup... no qualitative difference in breakage behavior for "aged" samples. The thin extrusion width samples still snap somewhat cleaner than the thick extrusion width samples.

The "print really slow" test bars didn't do anything exciting enough to post them last week. They were visibly more over-extruded at 1.6 mm^3/sec nominal than the 6.4 mm^3/sec nominal samples, which isn't terribly surprising. The 0.2x0.4mm low-speed bar broke a bit messier than expected. I don't know if this is just "single sample" noise, or if it indicates less residual strain due to the lower print speed. I could believe either. 

02x04 at 25% speed:

[Ryan Carlyle]

Really need a quantitative way of measuring strength...

[Whosa whatsis]

There's also the aesthetic issue I was printing the Benchy model, and I decided to try one with extrusion width = nozzle diameter. I attached some pictures (wider extrusion is on the left). The pictures aren't great, but in person you can really see that the wider extrusion has more consistent color, and is more clear in the thin areas, where the thinner extrusions show signs of air pockets unevenly distributed, making some parts look lighter and breaking the optical clarity. Some of the overhangs and bridges are clearly better on the one with the wider extrusion, but others, while they have different failure modes, are hard to call which is better than the other.

[Ryan Carlyle]

That's an interesting distinction that I hadn't thought of. Optical properties of translucent filaments do seem like they should be better with wider extrusion due to the more oval cross-section. Although there was no real difference in total quantity or consistency of air pockets in my test pieces. The shape distribution was certainly different.

Which slicer is this? It looks to me like your .6mm width print is actually over-extruded a bit, judging by the look of the floors by the base. That could be causing some of the color consistency differences. Over-extrusion = smaller gaps.

[Joseph Chiu]

Keep in mind that for some colorants used in filament, the temp and dwell time in the hot end may affect the color of the output it as it is being extruded.

On Apr 30, 2015 10:31 AM, "Ryan Carlyle" <temp...@gmail.com> wrote:

That's an interesting distinction that I hadn't thought of. Optical properties of translucent filaments do seem like they should be better with wider extrusion due to the more oval cross-section. Although there was no real difference in total quantity or consistency of air pockets in my test pieces. The shape distribution was certainly different.

Which slicer is this? It looks to me like your .6mm width print is actually over-extruded a bit, judging by the look of the floors by the base. That could be causing some of the color consistency differences. Over-extrusion = smaller gaps.

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/32f13fe9-bd0b-4eed-837a-e82dfc0c5e72%40googlegroups.com.

[whosawhatsis]

Sliced with Cura. I rushed the tramming after doing some modifications on the machine, and the first layer was a little over-compressed. That's why it look over-extruded.

+1 what Joseph (who I go the filament from, BTW) said. I've also noticed that, at least with this particular filament, the color has some directionality. The bottom layers in particular, which is a significantly thicker layer with significantly wider extrusions, looks different from different angles. Alternating lines will look darker, and they switch when you turn the part around, indicating that it has something to do with the direction the nozzle was moving when it was printed relative to the viewing angle.

On Thursday, April 30, 2015 at 10:31, Ryan Carlyle wrote:

That's an interesting distinction that I hadn't thought of. Optical properties of translucent filaments do seem like they should be better with wider extrusion due to the more oval cross-section. Although there was no real difference in total quantity or consistency of air pockets in my test pieces. The shape distribution was certainly different.

Which slicer is this? It looks to me like your .6mm width print is actually over-extruded a bit, judging by the look of the floors by the base. That could be causing some of the color consistency differences. Over-extrusion = smaller gaps.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.

To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/32f13fe9-bd0b-4eed-837a-e82dfc0c5e72%40googlegroups.com.

[tray]

I wonder if annealing would erase some of those differences by relieving internal stresses.

[Ryan Carlyle]

Probably would. I did some annealing tests with polycarbonate a while back, and it ended up WAY stronger. (Bend more before breaking, less delamination.) Problem is, you have to exceed the glass point to anneal an amorphous polymer, which means it's only really feasible with shapes that can support their own weight without sagging (eg flat shapes). And yes, you CAN dualstrusion print support in a high-Tg material and then anneal the print with support intact, but Stratasys has a patent on that. 

[whosawhatsis]

If you did it by burying the object in hot sand, you wouldn't have to worry about drooping, though you might have to worry about the sand trying to push into the infill areas, or even sticking to the surface if you got hot enough...

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/6919b638-798a-40f9-81af-e68e17e88415%40googlegroups.com.

[Ryan Carlyle]

THIS is interesting:

Read the full article: http://engineerdog.com/2015/09/02/mechanical-testing-3d-printed-parts-results-and-recommendations/

Strength maxima at width = nozzle and width = 2x nozzle? That's weird. I suspect he's seeing artifacts in the way Slic3r does volume calculations as much as real physical effects, but it's certainly an interesting dataset.

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+unsubscribe@googlegroups.com.

[Chris P]

Interesting doesn't do it justice.  Wish he'd repeat the testing with a slicer that doesn't do weird inexplicable things.

[Ryan Carlyle]

I seriously need to build a test jig like he has. But it's just a huge time investment to both build the jig and collect the data. 

The three big challenges with FDM print properties testing are:

• Disentangling slicer-specific behaviors and printer calibration quirks from real results
• Applying strains at rates that don't misrepresent material strengths
• Controlling and documenting non-obvious inputs like print speed and room temperature and filament pigment load

I have yet to see anybody publish test results that do any of these correctly. 

The academic FDM properties literature is mostly either Stratasys machines or uncalibrated consumer-grade printers. Comparing a Slic3r print on a RepRap to a Makerware print on a 5th Gen without trying to normalize for slicer behavior and calibration differences is next to useless. None of the data can be replicated because the people doing the testing don't seem to understand the calibration parameters they're ignoring. 

And when you load low-Tg amorphous filaments like PLA at high strain rates, they DRAMATICALLY over-perform their low-speed failure loads. PLA can creep to ultimate failure under load in a day or two. But that kind of polymer behavior isn't covered in university Strength of Materials courses, which focus on linear-elastic homogeneous isotropic materials (metals). So even real mechanical engineers usually don't know how to test polymers unless they have specific industry experience with them. 

Things as simple as repeating the same test in summer and winter or switching filament brands can significantly change results. Which makes any particular set of test results extremely unlikely to apply to anybody else's conditions. There are so many variable permutations that it really isn't practical rigorously test any meaningful subset of them... but there are statistical techniques that can extract trends from big, sloppy, multi-variate datasets like this IF you can quantify and record all the variables for inclusion in the database. 

[whosawhatsis]

Given the peculiarities of material distribution in an FDM print, I'm not sure the strength/unit weight is the most useful number for the vertical axis. I'd also really like some more data on what factors he controlled for, as well as more pictures of what the failures looked like.

The one failure that is shown seems to demonstrate that the printer used was under-extruding significantly in the long stretches, while over extruding slightly in corners. This makes all of the perimeters except the innermost one basically useless (though I wouldn't be surprised if the two apparent spikes in strength were cases where the adhesion to adjacent lines was better, for one reason or another), which means that increasing extrusion width effectively decreased the width of the test piece (thus actually decreasing the volume of the part of the piece that is actually contributing to strength while increasing the weight denominator). If I had seen the failure mode shown after doing this test, I would have thrown the results out, fixed my printer/settings, and started again.

A test piece with the sides sloped (say, a hexagonal profile) would be more resistant to this type of problem.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/f0897c4f-587f-46a4-a3f5-fb44771cb48a%40googlegroups.com.

[Ryan Carlyle]

Agreed, there are some test protocol issues. We could use a standard for this kind of testing.

[whosawhatsis]

There are also only 6 data points on the extrusion width vs. strength/weight graph, and no obvious line or curve that they fit (the closest fit would be some kind of sinusoid, which wouldn't make any sense). There are barely enough samples here to average them all together to get a single value for the approximate strength/weight ratio of a printed part. I think this is a case of more noise than signal, and you'd have to do a lot more tests to even begin to draw conclusions.

On Friday, November 6, 2015 at 07:00, Ryan Carlyle wrote:

Agreed, there are some test protocol issues. We could use a standard for this kind of testing.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.

To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/19e59407-116f-4220-9645-0da629f1018d%40googlegroups.com.