Raised Edges

[gr5 2,269]

Okay - I've seen this since I first started making prints - an overhang edge lifts too high and potentially the print head hits it and knocks it over. The worst ever was when I printed a propeller which was very difficult to print because of this but I prevailed and got a very nice print eventually.

 

Anyway I always thought it was caused due to shrinking sides pulling and rotating the overhang upwards. But that theory makes no sense with this print. Every shape you see is identical to the layer below but shifted. The problem is occuring as usual on the leading edge of overhangs. This is very zoomed in by the way, these are .1mm layers.

 

I've seen this on many of my own prints and also on other people's prints.

 

Can someone explain what causes this and how do you fix this? If this print fails I'm going to set up an experiment and play with fan/temp/etc. But meanwhile - does anyone have a theory as to why the edge hanging over space would stick *up* instead of droop down?

 

You might want to skip right to post #37 where we all started figuring out finally what is really going on!

 

[gr5 2,269]

Oh - and before anyone mentions it:

I have bed at 70C, head 220C, this is about 20mm off the bottom. This print is huge - this is a tiny area of the print. Each layer takes several minutes to print so by the time the print head comes back to one of these shapes the part has had a few minutes to cool. Fan was on but I just turned it off to see if this makes it better or worse.

edit: with fan off it seems better. We shall see...

 

[chrisp 10]

I've had similar results when the nozzle was leveled too close to the buildplate. It slowly creeps up with the build.

 

[IRobertI 521]

I'll give it a think. But the first thing that catches my attention is that it looks like you're over extruding quite a bit? Looking at the biggest top surface it looks like there's quite a bit of excess plastic. I wish my top surfaces were that solid so I could dial it back...

 

[gr5 2,269]

It's always overextruding on the lip because the lip is sticking up. But the plastic ends up going somewhere and it should fix itself after a few layers. I guess maybe because the part bends a little - that must have something to do with it - it bends down and then the nozzle leaves and when it comes back it's stuck up even further.

So I can see how it doesn't correct too quickly but it seems like it should still self correct after a few layers.

 

[gr5 2,269]

Maybe I should reduce flow to 90% for a few minutes?

 

[foehnsturm 970]

That's funny, what a coincidence!

Today I thought about starting a topic on this, but didn't have the time: "The tiny leg issue" or "Meshmixer support worries"

My understanding so far:

- The extrusion process applies some pressure to the structure underneath

- slanted structures, if tiny enough, sooner or later will bend due to that pressure

- bending causes thicker extrusion, especialy on the most overhanging edges

- ...

- if structures are rigid enough, backpressure increases and things even out by themselves

- if not, more bending, more thickened edges ... and finally the nozzle breaks down the structure or knocks it over

Therapy: I don't have a clue ...

Extrusion *has* to apply some pressure. Small structures *will* give in.

Thinner layers should increase the problem as there is less room for the extruded filament and less clearance for the print head.

[Dim3nsioneer 558]

(Sorry, I have to... :wink: ) UM1 or UM2?

I guess it's on the UM2... so this cancels out the direction of the air flow.

Maybe I didn't get it right, but your question is why it goes up? I guess it's really just the shrinking as you thought. Why shouldn't it go up with this print? The top of the new layer cools first becoming a tiny bit smaller than the bottom of the new layer. So the direction of the curl is up... and that's why you see an influence without fan...still you have an air-flow (e.g. from the heated bed). Don't underestimate thermally induced air flows; they are very strong actually...

Maybe some additional air flow below the overhang might change the situation. If you do tests, you could also play around with the bed temperature...

 

[gr5 2,269]

This explanation (plus the overextrusion explanation) both make sense.

I've been using 6mm diameter supports in meshmixer - partly for this very reason. Interesting.

The lack of fan may have made no difference because I got to a different section of the print then where parts "came together".

I'm going to lower the temp from 220 to 210 to hopefully underextrude a bit.

I want to set up a full scale test at some point but right now I'm in a multi-day print. I guess I could test on the UM1 but I am "working" right now. So maybe tonight.

 

[gr5 2,269]

When I was printing a V shaped part where the middle of the V was filled in, I got this. The overhang of the V arms I assumed were shrinking and pulling/rotating at the base of the V and causing this problem.

But I like foehnsturms explanation much better now. The PLA doesn't want to come out in a perfect rectangular stripe of filament. It would rather be skinnier and thicker but normally the head is pushing down on the part slightly so it flows into the proper rectangular shape.

But if the part isn't well supported it bends and lets the PLA be every so slightly thicker than desired (maybe 5%). So you get .105mm of filament where you expected a height of .100. This adds up over a few layers and eventually the pressure is enough for everything to equalize but it means you have these "raised edges". Sometimes the head hits these and knocks over a part.

To lower the pressure one might do better with thicker layers as you might need less pressure to spread out a layer of PLA. Or maybe underextruding by 5% will help. I really don't know the solution. Yet.

 

[gr5 2,269]

(Sorry, I have to... :wink: ) UM1 or UM2?

 

It happens on both. But the photo is of a UM2 which is doing a 36 hour (maybe much longer - maybe 48 hour?) print.

 

[foehnsturm 970]

One thing I will check out, is nozzle shape.

The UBIS I received today has a super sharp nozzle with almost no flat area. Eventually there is more room for the filament to "escape".

[Dim3nsioneer 558]

That's funny, what a coincidence!

Today I thought about starting a topic on this, but didn't have the time: "The tiny leg issue" or "Meshmixer support worries"

My understanding so far:

- The extrusion process applies some pressure to the structure underneath

- slanted structures, if tiny enough, sooner or later will bend due to that pressure

- bending causes thicker extrusion, especialy on the most overhanging edges

- ...

- if structures are rigid enough, backpressure increases and things even out by themselves

- if not, more bending, more thickened edges ... and finally the nozzle breaks down the structure or knocks it over

Therapy: I don't have a clue ...

Extrusion *has* to apply some pressure. Small structures *will* give in.

Thinner layers should increase the problem as there is less room for the extruded filament and less clearance for the print head.

 

I guess it's even worse: the backpressure pushes the printed part against the nozzle. It decreases the gap for the new layer even further and if print speed is slow enough the printed part is heated up again and expands...

If you have a retraction in addition creating a tiny tip then the knock-over is scheduled...

I have the curling effect quite strong on the UM1 when printing an Ultimaker robot at the right ear (face to front). I always get the buttom of that ear curled up. Not for the left ear which is strongly cooled by the stock fan. I hoped to get better results one day with a different fan setup.

So do you still have the same strong curling effect with the crossflow fan?

 

[illuminarti 18]

I recognize that print... :-)

Another issue relates to the slicer path: I think you'd get far better quality on overhanging parts if the slicer made sure to approach and leave a print island from a supported rather than overhanging part, wherever possible. E.g., on those diagonal struts, arrive on the side opposite to the overhang, where there is a good thickness of material underneath the nozzle. That would reduce the likelihood of the the head catching on the part on approach and flexing it out of the way. If the head catches on the lip on arrival, then the thin piece is vibrating and moving around under the nozzle, so its less likely that the new extrusion will go where it is supposed to - making the alignment and quality problems even worse.

(Which is a different suggestion from my other one that Daid said he was working on, which was making sure that unrelated islands don't get crossed/hit while moving around on complicated multi-island prints like Meshmixer supported ones).

More generally I've though for a while now that there could be a lot of quality gains from more carefully considering not just the order in which islands are printed, but whereabouts on each island the extrusion starts and stops.

 

[owen 19]

Not sure if someone has said this but in different terminology.

Normally you are squeezing out a round shape and flattening it out into a rectangle shape and sticking it onto a complete surface below.

In this instance one side is flattened out and stuck down as usual while the outside (that doesn't have a surface underneath) doesn't get squished flat and comes out as a thicker circle which goes higher and lower than the rectangle half (shape of a keyhole on it's side). On top of this there is no adhesion from underneath to prevent the normal warping and lifting of the non stuck down side.

 

[takei-naodar 0]

I know this type of behavior from metal..... you can bend metal if you heat just 1 side..... other side stays hard as befor but the heated sid will shrink when cooling down.

how to fix:

Way 1: Enclosed Printer with atleast 40-50°C Air temp (Could be problematic with larger builds)

Way 2: Decrease Heat insertion into the product (lower layer height, lower print temp. )

Edit:

What i forgot... it could also be that the cohesion force is pulling it into a more stable shape....which is normally a sphere....

how to fix:

lower layer height (less overhang that curls up)

 

[gr5 2,269]

Okay, I ended up pausing that print for 16 hours for a few reasons but it continued nicely. Now about 12 hours into the print I have to disagree with foehnstrum's idea. The supporting structure in this photo is solid as all hell. Even that "lip" is sharp enough to cut a finger on (I didn't though). It is very strong for it's thin edge and the supporting leg in the lower part of that photo is incredibly strong. I don't think there is even .001mm deflection of the leg when the head hits that lip. I mean it's not bending down anymore than say a solid cube would bend down.

I think I could pick the printer up by that leg and swing it around the room - it's on there very solidly.

 

[gr5 2,269]

I reduced flow to 90% and it didn't help - it made gaps between all the lines but still had a lifted edge. I did this for about 10 layers.

 

[foehnsturm 970]

Sometimes theory doesn't match reality so well :wink:

I'm still evaluating and making closeup, slow motion videos. I looks like it is all about print speed and - indeed - cooling.

Not about cooling down the part before the next layers starts, but immediate cooling in the moment of extrusion. The edges raise when the nozzle returns one or more times within the next fractions of a second and the spot is not cooled down enough.

Most common scenario: Tiny supporting poles with 2 perimeter lines. After the first perimeter the nozzle comes back within a few tenths of a second to print the second perimeter. If there is considerable overhang, trouble is likely to follow.

Is there that tiny supporting infill at the lips?

A short video:

 

 

[foehnsturm 970]

Another thought:

You see the nozzle is sitting there, heating up the part and waiting for priming/retraction to be finished. This takes a considerable amount of time compared to the short printing time in-between. Faster retraction (higher speed and/or less distance) would be very helpful. Direct driven extruders should have an advantage here. As far as I remember, Meshmixer documentation somewhere mentions a very small support diameter as the Makerbot default (of course it will move and print slower as well).

 

[Dim3nsioneer 558]

Very nice video!

If I look at the video I see the main issue is that already printed layers are partially liquified again. When the nozzle moves away, this liquified part is picked up by cohesion of the plastic fluid...

Why does it happen just on the overhang side? Well it doesn't. The effect is just stronger. You get the same thing everywhere if you print small structures too fast (e.g. minimum layer time below 5s). But on the overhang side the amount of material to be heated up again is limited or smaller compared to the other side. So the material temperature after contact with the nozzle is higher on the overhang side.

It's quite difficult to distinguish cooling phases. I expect the temperature to drop on some kind of exponential curve starting at the very moment the material leaves the nozzle. Re-heating certainly changes that temperature slope (EDIT: and active cooling by a fan as well). I think the crucial point is that the printed plastic is below glass temperature and stays there when the next layer is printed.

 

Another thought:

You see the nozzle is sitting there, heating up the part and waiting for priming/retraction to be finished. This takes a considerable amount of time compared to the short printing time in-between. Faster retraction (higher speed and/or less distance) would be very helpful. Direct driven extruders should have an advantage here. As far as I remember, Meshmixer documentation somewhere mentions a very small support diameter as the Makerbot default (of course it will move and print slower as well).

 

What if at least part of the priming after rectraction is done during the travel move? The idle time contacting the printed material would at least be smaller... But the priming process would have to be split up between the travel move and the original priming. Both gcode commands would have to be altered. It might be possible to do this with a Cura plugin.

 

[foehnsturm 970]

at 50 mm/s it's approx. 0,3 sec sitting, waiting and cooking the part and 0,3 sec printing ...

 

[Dim3nsioneer 558]

at 50 mm/s it's approx. 0,3 sec sitting, waiting and cooking the part and 0,3 sec printing ...

 

The 0.3s idle time seems quite long, but not impossible. Assuming you have to move the filament by the usual 4.5mm with 30mm/s, a quite conservative value, you would just need half of that time. Acceleration is set to 3000mm/s^2 on my Ultimaker, I think, so it takes 0.01s to reach the 30mm/s. So total time needed for priming might be something like 0.16s taking the distance driven during acceleration and deceleration into account. Is that compatible with whatever method you used to determine the approx. 0.3s? If not, where is the additional time lost?

Some things I rather like not to have 'al dente', e.g. already printed PLA... :shock:

 

[foehnsturm 970]

just doubled the value: the first time sitting and waiting for priming, the second time waiting for retraction.

 

[Dim3nsioneer 558]

just doubled the value: the first time sitting and waiting for priming, the second time waiting for retraction.

 

of course... :rolleyes:

I didn't see the forest, there were too many trees... :wink: