Why am I having to run with an extrusion multiplier of 60%?

Avatar

By elmoret 22 October 2016 03:30

Member · 158 comments

The tracks don't lay down like bricks - in practice, there's a tiny bit of air that gets trapped. It takes less than 1 cm^3 of filament to make a 1cm^3 "solid" cube.

Any chance your filament diameter exceeds what you've set in the slicer?

Avatar

By tomwang256 22 October 2016 06:54

Member · 6 comments

I am having similar issues with my corexy machine as well with titan. using the e3d recommended steps per mm i have to reduce the extrusion multiplier to 84% for 200micron layer height to not over extrude like crazy. When I was using just a regular bowden extruder the steps per mm was spot on and i had no issue with leaving extrusion multiplier at 100%. This is all with simplify3d as slicer.

Avatar

By deckingman 22 October 2016 09:03

Member · 702 comments
elmoret wrote

The tracks don't lay down like bricks - in practice, there's a tiny bit of air that gets trapped. It takes less than 1 cm^3 of filament to make a 1cm^3 "solid" cube.

Any chance your filament diameter exceeds what you've set in the slicer?

Understood and I can accept a few percent but 20% worth or air? I think that would be visually noticeable. Filament diameter is spot on, if anything slightly under. All rolls tested come out at 1.74 to 1.75 depending on where they were measured and slicr3r is set to 1.75. Thanks for the thought though.


Avatar

By deckingman 22 October 2016 09:24

Member · 702 comments
tomwang256 wrote

I am having similar issues with my corexy machine as well with titan. using the e3d recommended steps per mm i have to reduce the extrusion multiplier to 84% for 200micron layer height to not over extrude like crazy. When I was using just a regular bowden extruder the steps per mm was spot on and i had no issue with leaving extrusion multiplier at 100%. This is all with simplify3d as slicer.

Interesting. However, from all the testing I've done so far, I'm now convinced that the Titan is doing what it should. i.e. you ask it to deliver x amount of filament and that's exactly what it does. In a way, I'm pleased that I'm not the only one but it's frustrating to not know the cause.


Avatar

By deckingman 22 October 2016 09:27

Member · 702 comments

Quick update. I set the nozzle width to 0.6 in slc3r and tried a test print with extrusion factor set to 1.00. Severe signs of under extrusion - gaps between the lines, poor layer adhesion. Bang goes another theory but there something very weird going on which I'll photograph later when the bed cools down.


Avatar

By deckingman 22 October 2016 12:38

Member · 702 comments

This is the weird thing I was talking about. Sorry, I don't know how to get pictures in posts but here is a shareable link to my google drive folder.

https://drive.google.com/drive/folders/ … sp=sharing

It's 2 cubes created in OpenScad as a single object. They are 80mm square. This single file was then rendered and saved as an stl. It was then opened in slic3r, the layer width was changed from the 0.5 that I had been using to 0.6 then the gcode files exported.  Then it was printed but the print aborted after the first layer. It is under extruded and that is due to the change in layer width. But why the weird pattern on the left hand cube compared to the right? The infill direction was 45 degrees. This means that for BOTH cubes, only one of the XY motors was running to give the same 45 degree movement, from near left of the build plate to far right. 

The build plat is flat and level - trust me. It's a 10mm thick machined aluminium tooling plate on top of which is 6mm flat glass. Probing the bed at the 3 points where the screws are, shows all 3 point to be within 0.05 mm and that's on a 400mm square. The object was centred on the bed.

All explanations and theories gratefully received.

Ian


Avatar

By DjDemonD 22 October 2016 15:19

Member · 407 comments

Hi Ian,

Apologies if I have been negative that's not my intention, its just the extrusion issue whilst annoying is not going to, and hasn't stopped you from printing well it just means changing a setting that makes no sense i.e. set steps/mm to 300 instead of 400. But then if it works why be so concerned about it?

As for the images there is clearly a big difference between one motor turning (the 45 deg rotated object), compared to when two motors are turning. They are both underextruded, as that is a first layer you would expect filament to pile up due to first layer squash-down (unless you are deliberately setting your nozzle to bed gap higher, "laying" down the filament rather than squashing it down). Because you're underextruding the filament is being pulled thinner than it should be. In the non-rotated object Its almost like the extruder is pulsing, like a direct drive mk8 or similar with low steps/mm but with titan being 3:1 geared pulsing isnt normally noticeable. Also I cannot explain why the pattern appears when two motors are turning rather than just one.

Simon.

Avatar

By dc42 22 October 2016 15:52

Administrator · 2,256 comments
deckingman wrote

This is the weird thing I was talking about. Sorry, I don't know how to get pictures in posts but here is a shareable link to my google drive folder.

https://drive.google.com/drive/folders/ … sp=sharing

It's 2 cubes created in OpenScad as a single object. They are 80mm square. This single file was then rendered and saved as an stl. It was then opened in slic3r, the layer width was changed from the 0.5 that I had been using to 0.6 then the gcode files exported.  Then it was printed but the print aborted after the first layer. It is under extruded and that is due to the change in layer width. But why the weird pattern on the left hand cube compared to the right? The infill direction was 45 degrees. This means that for BOTH cubes, only one of the XY motors was running to give the same 45 degree movement, from near left of the build plate to far right. 

The build plat is flat and level - trust me. It's a 10mm thick machined aluminium tooling plate on top of which is 6mm flat glass. Probing the bed at the 3 points where the screws are, shows all 3 point to be within 0.05 mm and that's on a 400mm square. The object was centred on the bed.

All explanations and theories gratefully received.

Ian

That's a very interesting pattern. Send me the gcode file and I'll try it on my Ormerod in CoreXY simulation mode.

What happens if you change the infill direction to 0 degrees?


Duet WiFi hardware designer and firmware engineer
http://www.escher3d.com
https://miscsolutions.wordpress.com

Avatar

By deckingman 22 October 2016 16:45

Member · 702 comments
DjDemonD wrote

Hi Ian,

Apologies if I have been negative that's not my intention, its just the extrusion issue whilst annoying is not going to, and hasn't stopped you from printing well it just means changing a setting that makes no sense i.e. set steps/mm to 300 instead of 400. But then if it works why be so concerned about it?

As for the images there is clearly a big difference between one motor turning (the 45 deg rotated object), compared to when two motors are turning. They are both underextruded, as that is a first layer you would expect filament to pile up due to first layer squash-down (unless you are deliberately setting your nozzle to bed gap higher, "laying" down the filament rather than squashing it down). Because you're underextruding the filament is being pulled thinner than it should be. In the non-rotated object Its almost like the extruder is pulsing, like a direct drive mk8 or similar with low steps/mm but with titan being 3:1 geared pulsing isnt normally noticeable. Also I cannot explain why the pattern appears when two motors are turning rather than just one.

Simon.

Hi Simon.

No probs. It's just that I don't like mysteries. There has to be a logical explanation and it's bugging me that I can't find it.

The under extrusion is due to me changing the setting in slic3r from 0.5 to 0.6. When I run 0.5, I get over extrusion (or some effect that looks exactly like over extrusion). It's looking like the sweet spot will be somewhere between the 0.5 and 0.6 but I need to do more tests to be sure. In any case, it seems that, for whatever reason, my machine is very sensitive to how the nozzle width is set in slic3r.

As for the patterns, the only time 2 motors turn is when it lays down the perimeter. For infill, it was near left to far right of the build plate for BOTH of the "cubes". That is to say, the left hand motor tuned but the right hand motor was stationary when doing infill in BOTH cases. It seems that when you set infill direction in slic3r it is the direction on the build plate, regardless of how the object perimeters are orientated.  Don't forget that as far as slic3r is concerned, this was single object created from a single stl file.

So the mechanics/dynamics of the machine were identical (apart from the fact that one cube is physically offset to the right of the other). That's really freaking me out.

Ian


Avatar

By deckingman 22 October 2016 16:52

Member · 702 comments
dc42 wrote

That's a very interesting pattern. Send me the gcode file and I'll try it on my Ormerod in CoreXY simulation mode.

What happens if you change the infill direction to 0 degrees?

Hi David,

Thanks for the offer. How do I get the gcode file to you? Can't see how to attach a file by contacting you via this forum or by your web site contact form.
Cheers
Ian


Avatar

By peridot 22 October 2016 16:54

Member · 122 comments

Is it possible your extruders are skipping steps under some circumstances? When I had RAMPS electronics, skipped steps made a very obvious and recognizable noise, but on the Duet it is much less clear. And when I simply switched electronics and transferred settings, I did indeed get lots of skipped steps on the extruder when trying to do my "extrude 100 mm and see how much actually gets fed in" test, which put my steps/mm settings all over the place. (I am using an unnamed extruder that came as part of a kit, based on a full-size stepper with a metal planetary gearbox. It doesn't make manual feeding possible, it clogs easily, and tightening it down enough that it doesn't strip jammed filament raises divots that jam in the Bowden tube. But it prints okay most of the time.)

The "am I extruding enough?" test I find most helpful is to print a series of cubes, one perimeter, no top or bottom, and with different levels of infill: 10% (see how isolated lines of infill look), 90%, and 100%. When I have it right, the 90% infill should be visibly not-quite-solid, while the 100% should be solid without being bulgy. It's also a quick test that you can adjust to see what over-extrusion looks like.

I have also found, over three manufacturers, that "1.75 mm" filament is pretty consistently close to 1.72 mm in diameter; I think that's what they aim for. Of course this difference isn't enough to produce the error you're seeing, but I am surprised you report 1.75 mm actual diameter for your filament.

Avatar

By peridot 22 October 2016 17:06

Member · 122 comments
deckingman wrote

No probs. It's just that I don't like mysteries. There has to be a logical explanation and it's bugging me that I can't find it.

The under extrusion is due to me changing the setting in slic3r from 0.5 to 0.6. When I run 0.5, I get over extrusion (or some effect that looks exactly like over extrusion). It's looking like the sweet spot will be somewhere between the 0.5 and 0.6 but I need to do more tests to be sure. In any case, it seems that, for whatever reason, my machine is very sensitive to how the nozzle width is set in slic3r.

As for the patterns, the only time 2 motors turn is when it lays down the perimeter. For infill, it was near left to far right of the build plate for BOTH of the "cubes". That is to say, the left hand motor tuned but the right hand motor was stationary when doing infill in BOTH cases. It seems that when you set infill direction in slic3r it is the direction on the build plate, regardless of how the object perimeters are orientated.  Don't forget that as far as slic3r is concerned, this was single object created from a single stl file.

So the mechanics/dynamics of the machine were identical (apart from the fact that one cube is physically offset to the right of the other). That's really freaking me out.

Ian

In an effort to resolve the mystery, would weighing objects help? I mean, you can weigh a piece of filament to find out its density (optionally linear density), and then weigh a printed object to see whether the weight of the printed object equals the weight of the filament you told the extruder to put into it. Unfortunately my kitchen scale is only good to the nearest gram, so I'd have to print something big (and my extrusion multiplier isn't doing anything weird) but it would answer whether the amount of filament coming out is really the amount the slicer thinks should be coming out. You can also, of course, measure the length of filament extruded against the amount the slicer thinks needs to be extruded.

I got some weirdness when I had troublesome retraction settings: normal extrusion would work, retraction would work, but un-retraction wouldn't. so I'd get strange under-extrusion problems, and ultimately I got a hot end jam as the molten plastic was pulled back into the heat sink. This isn't what you're describing, but could retractions be failing to actually retract? That would produce extra extrusion, depending on the model.

What difference does nozzle size actually make to this? All the plastic that goes into the melt zone has to come out somewhere, and unless it's squeezing out the top (keeps happening as my E3D v6 unscrews itself) it's got to go into the print. Similarly, there's no extra plastic coming from anywhere. The nozzle size just sets how thin you can expect an extrusion to come out okay.

Avatar

By deckingman 22 October 2016 17:39

Member · 702 comments

@ Peridot,

I know this is becoming a long thread so maybe you haven't read it all through but if you go back a bit, you'll see that I've done everything you suggest (apart from weighing object that is). On printing a 1cm cube, slic3r calculates that it will need 417.8mm of filament which is exactly right when you do the maths for 1.75 mm dia filament. On printing that object, my initial steps per mm where a little high and it actually used 437mm of filament. Adjusting the steps per mm from 424 to 405 has got it very close. That is to say, when I print the object it uses very close to 417.8 mm of filament. That's with retract on layer change enabled so it's a good "real world" test which gives me confidence that the extruders are doing exactly what hey are supposed to do. The freaky thing is that the resultant cube has all the attributes of an object where the extrusion setting was too high (bulging corners, poor finish, lumps and blobs and dimensionally slightly too large). If I then print the same cube with an extrusion factor of 0.8 then it uses 80% less filament (around 330 mm) but the resultant object looks perfect, (although it is dimensionally a little under sized).

Ref difference in nozzle size? It beats me. When sliced, the calculated filament usage and object volume are the same, yet the print quality is totally different. 0.5 gives the effect of severe over extrusion (as detailed above), whilst 0.6 gives the effect of under extrusion - gaps between the lines, poor adhesion etc. My only thought is that the spacing between lines must be different. e.g for an object that is say 30mm across, if the slicer thinks the extruded filament is 0.6 mm wide, it will take 50 lines of filament to span that width, but if the slicer thinks the filament width is 0.5 mm, it will take 60 lines. Of course, the 50 lines should have more filament per line so it ought to cancel out.

I'm planning on doing a series of prints with the same object sliced using nozzle widths between 0.5 and 0.6 in 0.01 increments just to see.

After that, I think I need to find a nice sun bed, somewhere hot and sunny, next to a pool, in order to restore my sanity.


Avatar

By dc42 22 October 2016 18:16

Administrator · 2,256 comments

Hi Ian,

You can store the gcode files on dropbox or a similar cloud storage site and post a link.

You said that your cubes are a little under sized. By how much? If your X and Y axes are not moving enough, that will lead to  over extrusion.

David


Duet WiFi hardware designer and firmware engineer
http://www.escher3d.com
https://miscsolutions.wordpress.com

Avatar

By deckingman 22 October 2016 19:12

Member · 702 comments
dc42 wrote

Hi Ian,

You can store the gcode files on dropbox or a similar cloud storage site and post a link.

You said that your cubes are a little under sized. By how much? If your X and Y axes are not moving enough, that will lead to  over extrusion.

David

1. I'll put the file on my google drive and post a link then.

2. Cubes are a little undersized when extruding at 80% (about 0.2 to 0.3 mm)  and a little over sized when extruding at 100% (again, about 0.2 to 0.3 mm). That's regardless of the size of the object so on a 10mm cube its 2 to 3 % but on a 100mm it's 0.2 to 0.3%


Avatar

By deckingman 22 October 2016 19:16

Member · 702 comments

Herewith link to gcode file

https://drive.google.com/file/d/0B_MwtH … sp=sharing

Permission is set to anyone with link can view. Let me know if you want me to change that to "edit".

Ian


Avatar

By peridot 22 October 2016 20:09

Member · 122 comments
deckingman wrote

@ Peridot,

I know this is becoming a long thread so maybe you haven't read it all through but if you go back a bit, you'll see that I've done everything you suggest (apart from weighing object that is). On printing a 1cm cube, slic3r calculates that it will need 417.8mm of filament which is exactly right when you do the maths for 1.75 mm dia filament. On printing that object, my initial steps per mm where a little high and it actually used 437mm of filament. Adjusting the steps per mm from 424 to 405 has got it very close. That is to say, when I print the object it uses very close to 417.8 mm of filament. That's with retract on layer change enabled so it's a good "real world" test which gives me confidence that the extruders are doing exactly what hey are supposed to do. The freaky thing is that the resultant cube has all the attributes of an object where the extrusion setting was too high (bulging corners, poor finish, lumps and blobs and dimensionally slightly too large). If I then print the same cube with an extrusion factor of 0.8 then it uses 80% less filament (around 330 mm) but the resultant object looks perfect, (although it is dimensionally a little under sized).

Ref difference in nozzle size? It beats me. When sliced, the calculated filament usage and object volume are the same, yet the print quality is totally different. 0.5 gives the effect of severe over extrusion (as detailed above), whilst 0.6 gives the effect of under extrusion - gaps between the lines, poor adhesion etc. My only thought is that the spacing between lines must be different. e.g for an object that is say 30mm across, if the slicer thinks the extruded filament is 0.6 mm wide, it will take 50 lines of filament to span that width, but if the slicer thinks the filament width is 0.5 mm, it will take 60 lines. Of course, the 50 lines should have more filament per line so it ought to cancel out.

I'm planning on doing a series of prints with the same object sliced using nozzle widths between 0.5 and 0.6 in 0.01 increments just to see.

After that, I think I need to find a nice sun bed, somewhere hot and sunny, next to a pool, in order to restore my sanity.

Sorry about that. I see you have been pretty darn thorough - I just don't like mysteries either.

It's starting to sound like a slicer problem of some kind - if the extruder is putting out just what it was asked to, then the slicer must be asking for too much plastic, surely?

The nozzle diameter does affect what thicknesses of filament and layer heights are a good idea. Specifically, you can't easily lay down tracks much thinner than the nozzle size. So for slicers that automatically determine the width of infill, a bigger nozzle will lay down thicker lines (and fewer of them). So there will be differences in what the print looks like. Some of these differences will be non-obvious: the cross-section of a piece of extruded filament is a complicated rounded-edged shape that depends on the layer height, the nozzle width, the intended width, whether there are any other lines nearby, and even whether the path is curving. So the prints should look different, including whether there are air gaps inside them, but the total amount of plastic should not change.

The sun bed is starting to sound like a good idea.

Avatar

By deckingman 22 October 2016 21:03

Member · 702 comments
peridot wrote

It's starting to sound like a slicer problem of some kind - if the extruder is putting out just what it was asked to, then the slicer must be asking for too much plastic, surely?

Nothing wrong with that logic. However, when I slice a 1cm x 1cm x 1cm cube, with 100% infill and no skirt or brim, then check the gcode file, slic3r reports the volume to be 1cm^3 and that it will take 417.8 mm of filament to produce the object. If you do the maths, using 1.75 mm dia filament you get (1.75/2)^2xPi x 417.8 = 1000mm^3 (or 1cm^3). So on that basis, the slicer is asking for exactly the right amount of filament (and I now know that the printer will deliver exactly what it is asked to deliver).

When I print the object, it has the appearance of being printed with too high an extrusion rate and if I print it with 80% of the correct amount of filament, it has the appearance of being printed with the correct amount of filament. Note that I am now using the expression "appearance of" rather than stating that it is over extruded. I think there is a distinction somewhere between the two - but I'm not sure yet what that distinction is. 

Ian

Not that this is driving me crazy but I have to go - there is a little man wearing a white lab coat knocking on the door........


Avatar

By dc42 22 October 2016 21:35

Administrator · 2,256 comments

Have you tried a different filament?


Duet WiFi hardware designer and firmware engineer
http://www.escher3d.com
https://miscsolutions.wordpress.com

Avatar

By bot 22 October 2016 22:11

Member · 197 comments

I wonder if your printer is not printing in the correct x/y scale? Have you verified this? If your 10x10 cube with the correct volume of plastic is trying to take up 9.5x9.5 area, it would be squishy. Maybe incorrect Z scaling, too?

Last edited by bot (22 October 2016 22:12)

Avatar

By deckingman 22 October 2016 22:23

Member · 702 comments
bot wrote

I wonder if your printer is not printing in the correct x/y scale? Have you verified this? If your 10x10 cube with the correct volume of plastic is trying to take up 9.5x9.5 area, it would be squishy. Maybe incorrect Z scaling, too?

Hi bot
As per one of the above answers, cubes are a little undersized when extruding at 80% (about 0.2 to 0.3 mm)  and a little over sized when extruding at 100% (again, about 0.2 to 0.3 mm). That's regardless of the size of the object so on a 10mm cube its +/- 2 to 3 % but on a 100mm it's +/- 0.2 to 0.3% depending on the extrusion amount I use. Given that it can be under or over, I'm reasonably certain that the steps/mm in X Y and Z are correct.
Ian


Avatar

By deckingman 22 October 2016 22:30

Member · 702 comments
dc42 wrote

Have you tried a different filament?

I only have PLA - that's all I've ever used to date. I get the same result with "solid" Black and White and translucent Red, Yellow, and Clear, so 5 different reels but all PLA. They were from 3 different sources and bought at different dates about 4 months between the earliest and newest. I keep the filament in evacuated plastic bags with silica gel in the bag - the colour changing stuff that tells me when it's becoming saturated (and it's not).


Avatar

By deckingman 24 October 2016 12:06

Member · 702 comments

Latest (and probably final) update. Firstly here is a link to a picture of the issue.

https://drive.google.com/file/d/0B_MwtH … sp=sharing

It's exactly the same part, the only difference is that the one in the foreground was printed using an extrusion factor of 1.00 and the (better) one in the background was printed using an extrusion factor of 0.80. When the printer puts down the first layer, with extrusion factor of 1.00, the edge of the extruded filament is "rippled" - as if there is too much filament with nowhere to go, but with an extrusion factor of 0.80 it's a nice smooth "tube" of filament. When the vertical sides are being printed, these large blobs appear using 1.00 and I get stringing as if there is too little or no retraction, but there is no sign of any of this using a 0.80 (80%) extrusion factor. It's almost as if there is a build up of pressure at the nozzle tip caused by too much filament trying to be forced out through too small a gap and then suddenly is just "explodes" and releases the pressure but causes the blob. It looks like it can happen at any point, either on the perimeters or in the solid infill.

My thoughts at this moment in time are that it has to be over extrusion and that there is nothing else weird happening that is being masked buy under extruding. There also seems to be quite a bit of discussion on the "interweb" with many users of slic3r version 1.2.9 having the same problem. Some have said that rolling back to an earlier version fixes it. I know that if I slice a 1cm^3 in slic3r it generates exactly to correct amount of filament to fill that space. Maybe that's the problem. As someone has said, because of the way the filament is laid down, it's not "cube" shape so there will be gaps between individual threads at some points. So, it should take less filament to fill a given space to account for these gaps (but 20% or more seems an awful lot).

I've looked at other slicers but don't want to pay money in case that's not the problem. Other free ones don't have the features I need (like support for firmware retraction and multi colour parts for example). So, much as it pains me, in the interest of saving what little sanity I have remaining, I'm going to give up and just run with slic3r and a low extrusion factor, or maybe try an earlier version.

(Unless of course, someone comes up with a better idea...........).
Ian


Avatar

By iDevelo 24 October 2016 12:31

Member · 182 comments
Avatar

By deckingman 24 October 2016 14:04

Member · 702 comments
iDevelo wrote

You tried this? https://www.raise3d.com/pages/ideamaker

Thanks for the thought - had a quick look- seems to be specific to their machines - apart from anything else, it only support 2 extruders and I have 3.