Share:
Share
Notifications
Clear all

Fixing over extrusion

Page 1 / 2
Shinomori
(@shinomori)
Member

Hi,

I have a railcore ZLT with:

Bondtech extruder

Mosquito magnum 

duet wifi

It was printing pretty well and very precisely.  A print estimated at 700g would be dead on 700g.

 

I had a massive print failure resulting in needing new hotend, bltouch, and duet board.  I've reloaded my config and it is now over extruding.  I checked the extrusion setting and it is extruding what is requested within .25mm consistently, but the prints are terrible and overweight unless i set the extrusion factor to 98% on the machine.   The filament seems to be consistent diameter.

I've run out of ideas on what to look for.

Thoughts on where to start?   

Thanks

Quote
Topic starter Posted : 04/06/2021 1:24 am
Shinomori
(@shinomori)
Member

Here is a 3mf from prusaslicer that had the problem for sure.   Ugliness on the top layer, Build up on nozzle resulting in blobs on print, weight more than estimate(I'm not actually too concerned with this one other than it is telling my too much plastic is being extruded which is causing other problems).

 

ReplyQuote
Topic starter Posted : 04/06/2021 6:10 pm
Shinomori
(@shinomori)
Member

Here is a 3mf from prusaslicer that had the problem for sure.   Ugliness on the top layer, Build up on nozzle resulting in blobs on print, weight more than estimate(I'm not actually too concerned with this one other than it is telling my too much plastic is being extruded which is causing other problems).

 

ReplyQuote
Topic starter Posted : 04/06/2021 6:11 pm
Luke
 Luke
(@lmf5000)
Expert Moderator

Hi Shinomori,

Welcome to the community! 🙂

When your e-steps are perfectly calibrated as you say yours are, it's actually normal to need an extrusion multiplier of 95-97% to achieve perfect printing. It would be very difficult to isolate what is configured differently in your new build compared to the old one to be causing this problem, so the best approach is to tackle whatever quality issues are evident in this new build.

Could you start by setting your extrusion multiplier to 98% and printing a 20mm calibration cube? Post pictures of all sides of the cube, in particular I want to see the texture of the top and bottom to analyse how your printer is behaving.

Looking forward to your feedback 🙂

 

P.S. Before this, have you had a successful print of the object in your attachment? I notice you're trying to print a curved bridge without support (unless you had your support enabled and the setting simply didn't get transferred over the .3mf you sent me).

image

It is impossible to print such a line without support - you just end up with a straight line rather than one that follows that curve (because obviously it's printed in mid-air and the plastic just follows a straight-line path between the starting point and the nozzle) so the result is a lot of sagging plastic on the inside of the part where the bottom fill didn't meet the perimeter that didn't print. If you're actually printing with support on then disregard this observation 🙂

 

Regards,
Luke

ReplyQuote
Posted : 04/06/2021 10:05 pm
Shinomori
(@shinomori)
Member

Interesting.  I always figured if all settings were dead on it should print correctly and the under extrusion alot of people do was to make things a bit prettier, but you were losing strength since it is actually under extruded.

 

I'll get a 98% cube picture when this current print finishes.

This is the first time running this project and you are correct.  The undersides are terrible. This will be 10 pieces this size and will be mounted to a wall so ugly underside that will never been seen to save 2kg of plastic is worthwhile trade. Hehe it does eventually fill in by the time it gets to infill so the top is solid.

ReplyQuote
Topic starter Posted : 04/06/2021 10:35 pm
Shinomori
(@shinomori)
Member

Here are some pictures of a cube with extrusion set to 98% on the printer.

PXL 20210607 000915893
PXL 20210607 000848539
PXL 20210607 000841565
PXL 20210607 000829748.MP

 It came out a little heavy running 97% on current print.

ReplyQuote
Topic starter Posted : 10/06/2021 6:00 am
Luke
 Luke
(@lmf5000)
Expert Moderator

Hi again 🙂

Can you post the .3mf file of the cube? Normally when evaluating calibration cubes we print with two perimeters and rectilinear or monotonic top infill. The gaps between the adjacent lines and between top infill and perimeters gives a good indication of the extrusion multiplier. Your print appears to have either a large number of perimeters or concentric infill so it's harder to judge, but it does seem to be over-extruding slightly.

Could you re-print with the settings changed as described, and focusing on the texture of the top layer?

With regards to the final weight not matching the predicted weight, did this issue only start after changing the parts? I understand your desire to get things back to the level of precision that they were before, but even without the part changes being a contributing factor, I would expect some slight variation due to differences in filament diameter and even variation in the material from spool to spool.

After the extrusion multiplier is perfectly dialled-in, you can achieve higher accuracy by tuning the "filament density" parameter in the slicer so that the estimated weight becomes closer to the reality.

Looking forward to your feedback 🙂

 

Regards,
Luke

ReplyQuote
Posted : 10/06/2021 11:41 am
Shinomori
(@shinomori)
Member

@lmf5000

Hi Luke,

3mf was in the post above.  It does have 4 perimeters.  Oh!  I forgot about monotonic. Need to change that in my profile.  It does look nicer.

I'll run it again as soon as the printer is free.

My other printers almost always produce parts that are less than the estimate even if the material density matches the tds of the material.   Not a shocker since I know many printers default to slight under extrusion from the factory to make prints look nicer.   This printer was surprisingly almost always dead on (to the 1/10th of a gram if under 100g and within the gram even up to 1kg prints).  If I changed the extrusion factor in the slicer to adjust the part a bit it still matched.  So the printer was giving me exactly as much material as I asked for.   Which was a surprise but also what I would think would be the goal with a dialed in bondtech. 

I know it seems I'm crazily focusing on the weights.  If the only thing off was the number on the scale I would just change the material density in the slicer and call it good, but there is also a severe over extrusion appearance on the prints that matches pretty well with how far off the weights are.  So knowing that the weights were on before and now I have over extrusion that matches the over extrusion on the scale, weight seems like good data point to look at is all.  We can just ignore it if you'd prefer.   

Be back when I have another cube.

Thanks!

ReplyQuote
Topic starter Posted : 10/06/2021 6:22 pm
Luke
 Luke
(@lmf5000)
Expert Moderator

Hi Shino,

So as I understand it, all your other printers are dead-on with this same spool of filament? In that case the error is squarely due to the printer and not the filament. By conservation of material, you're only getting the filament on the print if the extruder has extruded it, so you can further focus your efforts on this subsystem of the printer.

Firstly, I know you mentioned earlier that the actual extrusion length matches exactly with what is requested, but can you double-check your esteps again using the same filament you are using for your test prints? The filament itself will affect the results slightly because different compressibility of the material results in a different effective radius when the extruder gear squeezes it. This is explained in this graphic -

image

If you've calibrated all your printers with this filament then this is a non-issue - but if you calibrated the rest a while ago with a different filament and they are spot-on now, but you're calibrating your new one with a different spool, that introduces uncertainty.

It's actually fine to be looking at the weight as an indication of extrusion multiplier - your logic is mostly sound there, though one thing that will throw you off is that slicers have different multipliers for different parts of prints, for example bridges can be done with a much lower extrusion multiplier so that the filament remains under tension during the bridging move and results in a less saggy bridge. Other optimizations under the hood of the slicer will tend to muddle the results a bit as well. So you might be over-extruding by 5% but only end up overweight by 2.5% so to speak, so the normally suggested method is to use visual inspection to determine whether extrusion multiplier is correct, or alternatively if you want a more quantitative approach, print a single-walled cube and use a micrometer to measure the wall thickness of the last few layers and correct it until it matches the extrusion width programmed into the slicer for those layers. The problem with this latter method is that inaccuracy in measurement will result in large differences in correction, so in practice it's actually much much easier to tune by looking at the top surface of a calibration cube.

Here's what I look for:

image

The way forward is as follows:

  1. Check very carefully that your e-steps are still calibrated. Your extrude length should be in the region of 100 or 120mm (G1 E100) - the longer the length the more accurate the measurement (but not too long because you'll hit the firmware extrusion limit)
  2. Print a calib cube, or just the top 5mm of one if you want to save time and material. Compare to the above image and adjust. My algorithm is normally to lower EM until I just start seeing gaps between the zigzags of the top solid infill, then raise by 1% and check that the top surface is entirely sealed (no more gaps).
  3. Measure the weight of the resulting cube and verify that it is satisfactorily close to the slicer estimate.
  4. If not, adjust to make it match the weight, and then look at the physical samples and see whether you're happier with the print from this point or the one from point 2, and fine tune accordingly.

The reason I'm suggesting to use cubes (or parts of cubes) is because it's a small, quick print, and being all the same you can compare and very clearly see the results of your settings changes after just a few minutes.

For the sake of convenience, I'm attaching an STL of a plane-cut calibration cube that will cut your time and material usage by 75% compared to a full cube (note that the lower mass means that weight becomes more unreliable, depending on the precision of your weighing scale).

Let me know how this goes! 🙂

 

Regards,
Luke

ReplyQuote
Posted : 10/06/2021 7:15 pm
Shinomori
(@shinomori)
Member

Not quite.  Other printers (mostly prusas) prints tend to be less weight than estimate with almost all materials.   But lets ignore those.

 

This printer:

Railcore II ZLT

bondtech extruder

mosquito magnum

duet2 wifi w/ duetx5

running reprap 2.05

 

before:

e-steps calibrated so 100mm extrusion extruded 100mm of filament.  Extrusion factor on machine set to 100%

Sliced estimate weight would match print weight exactly.  If the print looked over or under extruded I would change extrusion multiplier in the slicer.  Print weight would still match estimate exactly but over or under extrusion would be solved.   This was true over dozens of different petgs(what I am currently running and my main material on this machine)   and also very surprisingly Nylons filled and unfilled,PLA filled and unfilled,PP, and TPUs

 

Replaced with identical new parts:
BLTouch

Heatbreak

Duet2 wifi mainboard but moved sdcard over from old board so firmware and config is the same

 

Now:

e-steps calibrated with current filament so 100mm extrusion extrudes 100mm of filament.(didn't change)

Using the same slicer profiles

Print weight is 2-3% higher than the estimate from the slicer.  Part looks over extruded.  If I set extrusion factor on the machine to 97% part looks correct and the weight matches.

 

So it appears that the weight estimate is not off by the same amount as the machine is over extruding.  Which leads me to think the estimate is not off, but there is something causing the machine to over extrude even though e-steps are correct.   Filament diameter appears fine over several meters I checked and problem has persisted through 2 spools and several kg.

 

What am I missing?

ReplyQuote
Topic starter Posted : 10/06/2021 8:49 pm
Luke
 Luke
(@lmf5000)
Expert Moderator

@shinomori

This is a tough one 🙂

Slicers work internally by calculating filament volume, which they convert to a length measurement to send to the extruder as gcode (unless you've enabled volumetric E). The weight estimate is just a courtesy the slicer gives you by converting volumes to weights using the density you provide (in the sense that it's not used for internal calculations).

At this point the old config is lost so you can't compare to it. You say your extruder length is perfect - have you actually double-checked after my post? It's not the first time something goes out of spec on a machine after you've set it and checked it and documented you've checked it (happens even with $300k robots on the production line where I previously worked 😆 ).

I've done some tests and observed that changing extrusion multiplier changes the weight estimate of the print (i.e. the used filament (g) value). So it's odd that the estimated weight begins to match the real weight by changing the extrusion multiplier, because PrusaSlicer revises the estimated weight down in the same ratio when you reduce extrusion multiplier. Something non-proportionate is at play here ;).

The next thing to do is to open two instances of PrusaSlicer, and compare your Prusa profile with your Railcore profile side-by-side. Check all the pages and all the submenus. You might find a slight difference in the filament settings if the Prusa uses Prusa Filament settings and your Railcore is using the generic materials. Particularly filament density and filament diameter.

image

If PrusaSlicer is using more realistic values here then that would explain the discrepancies.

I know it seems unlikely, and doesn't make sense that hardware changes would cause this kind of issue - but this is really the only avenue left. The slicer controls extrusion amount by sending G1 Exxx commands to the printer, i.e. filament lengths. If the physical E-measurement exactly matches what the slicer requested then your hardware is as calibrated as it can possibly get. Another possibility is that the Prusa printers have e-steps set to slightly under-extrude - have you checked their extruder physical calibration recently? Might be useful to check a Prusa when you check the railcore too.

The only other small contributor could be the nozzle-to-bed distance - but having this too low would primarily cause overextrusion for the first few layers, it wouldn't increase the net weight of the print.

Let me know how it goes! 🙂

Regards,
Luke

ReplyQuote
Posted : 11/06/2021 9:01 am
Shinomori
(@shinomori)
Member

Hi Luke,

Here are some cubes. From left to right 100-95% extrusion set on on the machine same gcode otherwise.  3mf also attached.

PXL 20210611 202423586

 

The estimate does go down if the extrusion is changed in the slicer.  I was changing it on the machine.

I checked the extrusion it was off by .1-2 mm which may or may not be beyond the accuracy of my $30 calipers.  😆 

ReplyQuote
Topic starter Posted : 11/06/2021 9:47 pm
Luke
 Luke
(@lmf5000)
Expert Moderator

Hi Shino,

Sorry for the delay in replying.

Now I understand why you are able to change extrusion multiplier and look at the difference in weight! 😮 I would recommend against changing the extrusion rate on the printer menu - that setting overrides all axis commands going to the extruder, so it also affect sretraction distance, retraction speed and retraction acceleration among other things. It is best to adjust it in the slicer settings. The flowrate override in the LCD menu is mainly intended for emergency setting partway through a print if you find that you need to tweak extrusion multiplier one way or another.

Due to the colour of the filament and the lighting, it's hard to make out the lines on the surface of your cubes, but it seems like the central one (97%) might be best overall. However your print settings are very different from what I would normally recommend for general printing (and for gauging cubes), so I will reply in a few hours with an updated profile and an explanation of the rationale behind my changes.

Talk to you soon!

 

Regards,
Luke

ReplyQuote
Posted : 14/06/2021 2:58 pm
Luke
 Luke
(@lmf5000)
Expert Moderator

As promised, here's what I recommend you adjust to improve speed, material use and quality:

  • First layer height - 0.2mm to match the rest of the layers
  • Solid layers top/bottom - 3 and 3 to save time and material
  • Minimum shell thickness top/bottom - set to 0.6mm so PrusaSlicer automatically adds top/bottom layers for you when printing with finer layer heights (eg. this brings you back up to 6 top/bottom layers if layer height is 0.1mm but is done automatically - no need to change top/bottom layer heights manually; for larger layer heights it respects the 3-layer top/bottom minimum set in the previous point)
  • Ensure vertical shell thickness - you can turn this off to save some material
  • Avoid crossing perimeters - activate and set to 300% to avoid stringing in undesired places
  • Detect thin walls - disable, it prints thin walls better with this disabled in my experience
  • Seam position - aligned - makes it easier to sand the seam if it's aligned on shapes with no sharp corners like cylinders
  • Infill density - reduce to 15%
  • Infill type - change to adaptive cubic - huge material and time savings on large prints because the center prints hollow and the infill is concentrated near the walls and top/bottom surfaces. With gyroid and other infill patterns the infill is equally distributed throughout the print. Infill in the core of the print doesn't really have a beneficial effect so it's wasted. You want it close to the outer surfaces for maximum benefit.
  • Skirt and brim - I have these disabled for my prints. I've left them on for you, but consider removing them to make cleanup easier. The intro line should be sufficient to prime the nozzle
  • Speed - your speeds are a lot higher than my cartesian printers. I've left them unchanged but if you want more quality you can change yours to the following:
image
  • Advanced - I've changed the extrusion widths to "0". This makes PrusaSlicer calculate the optimal widths for all features. The plus is that if you change the nozzle diameter, you just need to change one value in the Printer Settings -> Extruder 1 -> Nozzle Diameter box, and all the extrusion widths will automatically update to match the new nozzle size. No need to maintain multiple profiles and change half a dozen boxes to make this change.
  • Infill/perimeters overlap - reduced to 25% to reduce over-extrusion at the weld between infill and perimeters (plus saves some material)
  • Filament settings - set your extrusion multiplier to 0.97 based on estimated best quality print from the photos (change this to suit after examining physical samples)
  • Cooling - minimum print layer time reduced from 15 seconds to 10 seconds for more speed (PETG can handle quite a bit more heat than PLA so you don't need to be so conservative here - on my printers I use 100% cooling fan and have minimum layer time at 5 seconds, but for your profile we won't push it that far for now)

In general, I'd recommend a lift-z of 0.2mm, retraction speed of 70mm/s, and disabling "retract on layer change" and "wipe while retracting" for more speed and better quality. But since you have these set as both a filament override and as base parameters under the print settings, I will leave it up to you to decide if you want to try these, since you will have to amend or disable the filament overrides for all your filament profiles to make these changes stick.

Here's the updated 3mf:

Material used has decreased from 2.27g to 1.56g and time has decreased from 8m to 6m on this small sample. Benefits will be greater on your full-sized prints 🙂 - though as always test and make sure you're happy with the results and revert settings back or fine-tune as needed before moving on to bigger prints!

Let me know what you think 🙂

 

Regards,
Luke

ReplyQuote
Posted : 15/06/2021 7:38 am
Shinomori
(@shinomori)
Member

Awesome. Thanks Luke.

I am building a new profile for a print I need to do with a 0.4 nozzle for some extra details.  I'll start from your profile for that and see how it goes.

 

Most of our prints are functional in use parts or prototypes that will be used for testing the final part.  That's why we have extra shells-gyroid and top and bottom layers (though tops can be ugly without enough layers to clean up).   Some of our parts have needed to be water tight.  And the most recent ones needed even more shells so they could be sanded and painted afterwards.     I will definitely keep these tips in mind for when we need to save material and the parts we're doing can get away with it.

 

Detect thin walls is definitely a per print setting.   It is the only way to get the printer to attempt single extrusion walls for detailed parts, but it can also do weird stuff on perimeters.

 

This machine has ALOT of cooling.  Even PLA doesn't work well with 100%.   I'm actually hoping to be able to go faster! 😎   But I think I might have to turn speeds or accelerations down as I try to enable crash detection and filament jam detection.

 

I'll get some pictures of the test prints for this next project with the 0.4 nozzle.

ReplyQuote
Topic starter Posted : 18/06/2021 11:18 pm
Page 1 / 2
Share:
Share

About us

We’re a bunch of 3D printing and design nuts. We just want to make great 3D design available to everyone. 

Our mission

Good design takes time, great design takes a process. You can learn this with us, to build your skill set in this rapidly expanding market. 

Learn with us:

© Copyright: 2019 io3dp.com
Close Menu