How to prevent "overshoot" corners or "round edges"

[Bossler 8]

Hi All!

 

I have an issue with the print quality of corners on my UM3e.

They always become sort of rounded/overshoot as one can see in the pictures below.

 

Now, I thought this must be due to the rather fast print speed I used.

So I did print a test cube (2x2x2cm) at 70 mm/s and a second at 40 mm/s.

 

40mm/s

 

70mm/s

 

To be honest - I do see rarely a difference as far as the "rounded corners" are concerned.

In both cases the overshoot is about 0,25mm in height.

 

Interestingly the overshoot is on both sides of the corners.

The other dimensions of the test cube are fine, nearly 100% to the point (so 20mm=20mm in X- and Y-direction).

So I would exclude overextrusion from the list of suspects...

 

What can be the root cause? Mechanical problems of the printer?

How is the "corner"-performance of your UM3's?

Anybody with the same issue?

[SandervG 1,521]

I would be curious to see if you increase the acceleration what happens. In Cura you can enable 'Acceleration control', and in particular you could increase wall acceleration to 1500. 

[SandervG 1,521]

I also read a tip from @tinkergnome that changing the jerk settings to a minimum of 10 or 12 could have a positive affect on your corners as well. 

[Brulti 177]

Had the same problem recently, changing printcores for a newer one made it disappear... Yes, it is strange.

[gr5 2,228]

The bowden pushes on the filament like a spring and there is a roughly constant pressure in the nozzle as you print so every time the print head slows down it over extrudes.  and it slows down on corners.  The lower the acceleration and the lower the jerk, the more time it spends on the corners.  The simplest solution is to set the print speed to the jerk speed.  That way it won't slow down.

 

The term "jerk" is a bad term here because physicists have a very different definition for that word.  What it means for Marlin (the firmware in all UM printers and in 90% of printers out in the world) is the speed the head slows down to on corners.  Marlin takes the velocity vectors for the movement going in and out of a vertex (a 3 dimensional velocity) and subtracts the vectors and takes the magnitude of that.  for 90 degree corners that means the speed at each corner can not exceed square root of 2 times jerk speed (so if jerk is 20 then 14mm/sec is max corner speed).  For very gradual corners like in a circle with many segments - that speed approaches infinite.  Typically a very slight change in direction might have a jerk limited junction speed of 1000mm/sec (basically unlimited).  

 

If you increase jerk too much your printer will skip steps and your prints will have layers above that don't match the layer below basically ruining a print.  The easier solution is to just slow things down.  25mm/sec should be more than enough.   You can compensate by doing thicker layers.  It's fine to print the infill much faster than this.  Please try this as a test and let us know.  You can even do this test live - if you slice for 100mm/sec then in the TUNE menu the % speed will match the speed in mm/sec and you can play with the speed in the TUNE menu without doing any math.  Or you can slice at any speed you want and do the math to figure out what speed you are getting when you multiply by the %.

 

Anyway consider printing the cube and in the TUNE menu try 100% speed for 10 layers, then 50%, then 25%, then 10%, each test for a few mm of the print and mark the changes with a sharpie maybe and keep notes and let us know what you discovered.

[gr5 2,228]

By the way - other people disagree with me - so I'd love to see you do the experiment to prove me right or wrong.  I just print everything around 30 to 35mm/sec and don't see so much rounded corners - but maybe I don't print cubes much so...

 

[Bossler 8]

Thanks for the hints!

 

gr5 - will try that "live" test.

First one will be going down from 100mm/s, acceleration & jerk on standard settings - let's see.

Btw. - I do normal print other things than cubes as well;-)

 

Did not take the bowden-setup into account when thinking about this issue.

My pervious printer was direct extruder type.

There the solution was to print slower - while slow was around 50mm/s.

As I thought that a bowden setup is way lighter than a direct extruder, 

I did expect it to be even faster...

[gr5 2,228]

It is faster.  I've printed certain items at 200mm/sec.  Let's do the experiment first.

 

Plus the acceleration, at least on the UM2 is much higher then printers with head mounted extruders.  The UM3 has much lower acceleration due to it's heavier print head.

 

So the UM2 at 50mm/sec can beat other printers at 100mm/sec because they are printing something where the head rarely  actually gets above 50mm/sec.

[Bossler 8]

Well, it might be possible to print at 200mm/s - as long as you do not have any edges, right?

Anyway.

 

I did the test:

 

From bottom to top the speeds where (in mm/s):

100, 80, 70, 50, 30, 25

 

Looks like at 30mm/s it starts to become relatively okay.

 

So this is definitely something the R3D N2 could handle better:

This was printed at 70mm/s, No-Name PLA.

(and the N2 I have owned did have a lot of issues, far from perfect)

 

Shouldn't it be possible to take the direction change into account during slicing?

So slowing down extrusion before slowing down head speed?

I think I have been reading something about that issue with bowden type extruders and how to cope with that -

but can't remember, where...

 

 

 

[yellowshark 153]

Another complementary solution is to put a chamfer into your corners so it is not a 90 degree bend - of course if you want a 90 degree corner that will not work very well

[SandervG 1,521]

10 hours ago, Bossler said:

So slowing down extrusion before slowing down head speed?

 

 

I think that is actually what is happening too, perhaps on the background. Let me see if I can find some proof for my statement. 

 

In the meantime, did you change anything to the acceleration / jerk?

[Bossler 8]

15 minutes ago, SandervG said:

 

I think that is actually what is happening too, perhaps on the background. Let me see if I can find some proof for my statement. 

 

In the meantime, did you change anything to the acceleration / jerk?

 

Hi Sander,

thanks, would be very interesting to learn whether one can finetune this aspect.

 

I did not yet do a try with changed acceleration but will start a test for that right away.

[geert_2 558]

13 hours ago, Bossler said:

Well, it might be possible to print at 200mm/s - as long as you do not have any edges, right?

Anyway.

 

I did the test:

 

From bottom to top the speeds where (in mm/s):

100, 80, 70, 50, 30, 25

 

Looks like at 30mm/s it starts to become relatively okay.

 

So this is definitely something the R3D N2 could handle better:

This was printed at 70mm/s, No-Name PLA.

(and the N2 I have owned did have a lot of issues, far from perfect)

 

Shouldn't it be possible to take the direction change into account during slicing?

So slowing down extrusion before slowing down head speed?

I think I have been reading something about that issue with bowden type extruders and how to cope with that -

but can't remember, where...

 

 

 

This is an interesting test. Maybe you could replace the X by a couple of vertical lines of same depth, so they show up in every layer? On my UM2 (non-plus) it generally looks like the areas from 50mm/s and slower in your test. But I only print between 25mm/s (for fine details) and 50mm/s (general prints). Ringing is more visible than the widening of corners.

 

[ultiarjan 1,223]

Also, be aware comparing print speeds has become virtually impossible since The Ultimaker3 and cura2/3.

for the UM3 if you set 80mm/s it's only the the infill,  the actual print speed is 40ish...

 

 

Default UM2 profiles have this effect a lot less;

 

I would suggest to always communicate the used wall speed

 

@Bossler I assume the speeds you mentioned are the infill speeds... so actual wall speed i less that half...

 

 

 

[neotko 1,417]

7 minutes ago, ultiarjan said:

Also, be aware comparing print speeds has become virtually impossible since The Ultimaker3 and cura2/3.

for the UM3 if you set 80mm/s it's only the the infill,  the actual print speed is 40ish...

 

 

Default UM2 profiles have this effect a lot less;

 

I would suggest to always communicate the used wall speed

 

@Bossler I assume the speeds you mentioned are the infill speeds... so actual wall speed i less that half...

 

 

 

 

They still don’t understand the importance of a no-tricks mode?

 

sad...

[gr5 2,228]

I get similar results.  On a UM2 which has higher jerk and acceleration I can print faster and get the same corners but it's the same basic trend.  This is one reason why Cura likes to do faster infill speeds, then on the inner shell it slows down to get a perfect outer shell.

[cjs 34]

I think the problem, which can be seen in the thread, isn’t the speed in particular. It has to do with the fact that Ultimaker still misses a pressure advance algorithm.

The concept was sucesfully introduced around 2011 on makerbots thanks to the sailfish team! Prusa added such an algorithm a few month ago as well to his printers.  Maybe Ultimaker will acknowledge it soon or show us the research they did which suggest otherwise. 

[gr5 2,228]

Well the "advance" feature definitely helps at higher speeds.  It's a bit hacky and does not deliver the correct advancement of extra and less extrusion when slowing down and speeding up.  But it does help a bit.  It was popular on non-bowden printers that had lower accelerations but now that the UM3 is out and has lower accelerations than UM2 maybe it's time to turn this feature on for the UM3.

[Bossler 8]

1 hour ago, cjs said:

I think the problem, which can be seen in the thread, isn’t the speed in particular. It has to do with the fact that Ultimaker still misses a pressure advance algorithm.

The concept was sucesfully introduced around 2011 on makerbots thanks to the sailfish team! Prusa added such an algorithm a few month ago as well to his printers.  Maybe Ultimaker will acknowledge it soon or show us the research they did which suggest otherwise. 

 

Makes sense.

I also suspect the coordination of movement and extrusion feed to be the root cause for this artefacts.

Unfortunately. 

Because this means to me that there is nothing I can do against it - expect printing at ridiculous slow speeds.

 

When comparing prints at various speeds (and yes, the speeds mentioned by me are "the headline numbers"),

one can see that there is literally no difference from 100mm/s down to 40mm/s.

Going down to 30 or 25mm/s is where the corners start to look like corners and not like a lory turning around a corner...

Reflecting neotko's remark 30mm/s "headline number" print speed means 9 mm/s (!!) outer wall speed -
which is somewhat like using a 3d-pen ...

 

[gr5 2,228]

9mm/sec!  Yikes!!!  You should be able to do better than that!  Maybe it's time to play with the temperature next?  Try colder.  Much colder.  Like maybe 180C and start speeding it back up again.

 

I have to say it's a bit annoying that everyone wants the UM3 because... well... "3 is better than 2".  But in reality the UM2 has much higher acceleration and jerk values giving you better quality at higher speeds.  The UM2go - the worst selling in the UM2 series - is the best printer of all of them due to it's extra stiff frame and extra short bowden.

[Bossler 8]

Well, for me the UM2 was no option since I need a dual extruder and did not want to 

experiment with tinkering on the UM2 to get dual extrusion.

 

Compared to the printhead of my former printer, the R3D N2, the printhead of the UM3 is still a lightweight!

Just the Direct Extruder of the N2 weights close to 800 gramms...

 

So I would think that the UM3 should be able to go much faster.

 

Is the printer/Cura somehow protected against to high acceleration/jerk values?

Because I would like to see how the prints look like when I increase them -
but I am cautious not to kill the machine...

[tinkergnome 927]

22 minutes ago, Bossler said:

So I would think that the UM3 should be able to go much faster.

 

Sorry, but I think, that's a wrong expectation. Because the printhead is lightweight (compared to direct drives) it is also less sturdy (or more wobbly - if you want so...).

 

You can even switch off the acceleration and jerk control in Cura entirely, the printer will then use the default values that are defined by the firmware (which are probably quite high).

 

I assume you will get much more ringing or other drawbacks. You'll have to find a compromise that works for your specific models and requirements.

 

[Bossler 8]

9 hours ago, tinkergnome said:

 

Sorry, but I think, that's a wrong expectation. Because the printhead is lightweight (compared to direct drives) it is also less sturdy (or more wobbly - if you want so...).

 

You can even switch off the acceleration and jerk control in Cura entirely, the printer will then use the default values that are defined by the firmware (which are probably quite high).

 

I assume you will get much more ringing or other drawbacks. You'll have to find a compromise that works for your specific models and requirements.

 

 

From what I saw in the prints I would agree regarding ringing. 

Yes, that is for shure as good or bad as it was with the R3D N2.

 

What a pitty.

[Creat0r 0]

Hey there,

I had this problem too, in case of an Ultimaker there are two ways to solve it:

1 Disable"Accelerations control".

2 Enable "Accelerations control" and set the print acceleration to something like 6000 or higher (You may want to leave the acceleration for tasks with higher feed rates like travel or infill just low as it was - to preserve your bearings and other printer parts).

 

So basically having a faster acceleration at the corners should do the job.

Let me know if it was helpful. :)

[DracieVajko 0]

On 2/10/2020 at 7:05 PM, Creat0r said:

Hey there,

I had this problem too, in case of an Ultimaker there are two ways to solve it:

1 Disable"Accelerations control".

2 Enable "Accelerations control" and set the print acceleration to something like 6000 or higher (You may want to leave the acceleration for tasks with higher feed rates like travel or infill just low as it was - to preserve your bearings and other printer parts).

 

So basically having a faster acceleration at the corners should do the job.

Let me know if it was helpful. :)

Hi, I was thinking about this too to just disable acceleration or set it up to huge numbers. The thing is, I need acceleration especially on outer walls, because I get a lot of ringing\ghosting. without it. So thats kind of a problem. I usually print at 70mm/s for outer and 120mm/s for inner walls and infill. Acceleration is around 150 and 250 at max, higher values again shows too much ringing. But im thinking about trying some jerk control towers with different settings if it would help somehow. I have Artilerry Sidewinder X2