3

What could cause that?

When linear_advance (v1.5) is "off" (M900 K0) - everything is uniform and smooth, only corners are protruding as expected.

When it is "on", I get these horizontal defects. My calibrated K-value is 2, I am getting same defects with K=1 as well as K=2. Position of defects is repeatable, so it is not random mechanics-caused underextrusion.

My setup uses a BMG Bowden extruder, A4988 drivers for Z and extruder with diode smoothers (X, Y are on TMC2209). I print the PETG with 245 °C, Perimeters at 50 mm/s, Gyroid Infill at 120 mm/s.

After the original posting of the question I tried the following:

  • Removed diode smoothers on extruder and Z - no improvement.

  • It seems defect is somewhat connected to infill. Without infill print is nearly perfect. 50x50 box without infill prints perfectly, so it does not look like outer-wall speed-related. Reduced outer wall to 33 mm/s, infill to 100 mm/s (accelerations are also down 750->500, 1000->750), no improvement. Returned to A4988 on all axis - no improvement. Switched to 1/8 step on extruder (maybe it is step speed-related) - no improvement. Disabled infill before walls - no improvement.

  • Starting shooting time-lapses, and outer perimeter is extremely weird on some layers (see below)

  • Migrated configuration to latest Marlin 2.0. Enabled SQUARE_WAVE_STEPPING - no improvement. Gcode is uploaded here: https://s.14.by/3d_issue.gcode

  • After inspecting larger parts printed before linear_advance (on large 100 mm circular parts) - I also noticed this defect, more rare though.

  • Connected to oversized lab power supply, to ensure there is no power sag during high load. 24V 16A max power consumption. Print defects remain exactly the same though.

enter image description here

3D print showing horizontal defects

BarsMonster
  • 71
  • 1
  • 8
  • 1
    Which version of Linear Advance (1.0 or 1.5 - that is, pre-1.1.9 Marlin or >=1.1.9) are you using? The K values look like 1.5, but 2 is very high for normal PLA and even 1 is somewhat high. Did you use the calibration pattern to select a K or just guess? – R.. GitHub STOP HELPING ICE Sep 15 '19 at 00:30
  • 1
    It looks to me like you have severe underextrusion in the areas where the print head has accelerated to significant speeds, but just on some layers. I suspect the mechanism of the problem is that your extruder gearing or idler is not up to producing the pressure that linear advance with K=1 or K=2 is attempting to get, and the filament slips backward. Can you watch closely and see if that appears to be the case? – R.. GitHub STOP HELPING ICE Sep 15 '19 at 00:54
  • @R.. It's 1.5. This is PETG, I used calibration pattern. – BarsMonster Sep 15 '19 at 00:56
  • @R..It's hard to catch this in real time, but it is fairly consistent. I printed like 6 of these, and defects are in the same positions. I will try to tighten BMG gear and see if it helps. Also, in previous attempts I was reducing Ejerk (20->10->5) and Eacceleration (20k->10k->6k) - without any changes. – BarsMonster Sep 15 '19 at 00:58
  • 2
    @BarsMonster: What temperature? PETG is hard to move through a nozzle fast even at high temps, and at lower temps I would not be surprised at all if it slipped from too much pressure. K between 1 and 2 does make sense for PETG with a bowden though. – R.. GitHub STOP HELPING ICE Sep 15 '19 at 01:01
  • @R.. 245°C, at the upper recommended temperature range – BarsMonster Sep 15 '19 at 01:10
  • 1
    @BarsMonster: OK, temp shouldn't be the problem then. Are you printing it too fast? Otherwise my guess is just that the tension holding the filament is not sufficient. – R.. GitHub STOP HELPING ICE Sep 15 '19 at 01:13
  • @R.. Perimeters where problem is visible are at 50mm/s. Infil at 120. Accelerations are 500 for perimeters and 1500 for infills. And puzzling part is that without LA these speeds are not the issue, before LA I was even pushing it to 80... – BarsMonster Sep 15 '19 at 01:18
  • @R.. What makes it unlikely a random slip of gear is that it is very consistent. Every print defect is in the same place... – BarsMonster Sep 15 '19 at 01:29
  • 1
    @BarsMonster: I don't think you can actually extrude PETG at 120 mm/s. At 0.2 mm layer height (guess) and 0.4 mm nozzle (guess) that's nearly 10 mm³/s volumetric extrusion. There are likely lots of issues going on that just show up more visibly with LA enabled. – R.. GitHub STOP HELPING ICE Sep 15 '19 at 01:31
  • @R.. You diameter guess is perfect in both cases. 0.2 / 0.4... Cracked the part open - infill (gyroid) looks ok, I guess it is unable to accelerate to 120mm/s anyway's. – BarsMonster Sep 15 '19 at 01:35
  • @R..Tightened BMG's gear quite a bit - but defect is still at exactly the same place. – BarsMonster Sep 15 '19 at 01:46
  • @BarsMonster: Yeah, I think you'd need some *extreme* acceleration limits to actually hit 120 mm/s while doing gyroid. – R.. GitHub STOP HELPING ICE Sep 15 '19 at 01:58
  • 1
    Can you try to print the tower with 40 mm/s for the outer shell? It seems to only have an effect on long lines... – Trish Sep 15 '19 at 09:28
  • @Trish I will try, but I am not sure why quality is different for different identical layers, where speed profile on the perimeters should be identical. – BarsMonster Sep 16 '19 at 16:41
  • hm, is the extruder bar plastic or metal? – Trish Sep 16 '19 at 16:45
  • @Trish Hot part of extruder is mounted on extruded-aluminum plate with 2 press-fit steel linear bearings. Cold part of extruder (BMG) has injection molded plastic case, the rest of BMG is metal. – BarsMonster Sep 17 '19 at 06:46
  • 1
    I had uneven extrusion when the lever pressing the filament to the hobbed gear had secretly broken. Checking the lever and/or spring for even pressure could help – Trish Sep 17 '19 at 08:57
  • @Trish Extruder is solid, no slipped steps. Also, defect position is consistent - on every print it is on the same places. – BarsMonster Sep 19 '19 at 00:28
  • Does disabling combing in the slicer make the problem go away? (Cura uses the name "combing"; other slicers may call it something different. It's the feature to allow moves without retraction over already-printed regions or infill regions, and combing over infill region can lead to loss-of-material problems as described here: https://github.com/Ultimaker/CuraEngine/issues/1084) – R.. GitHub STOP HELPING ICE Sep 19 '19 at 03:17
  • that the error is Height dependant might indicate that the error is in the Z axis... What kind of printer is it? – Trish Sep 19 '19 at 07:33
  • @R.. I see, interesting. In my case combing was enabled only in infill (Cura) – BarsMonster Sep 19 '19 at 09:10
  • @Trish It's Flyingbear Ghost4. If it's Z-axis-related, then it is not clear why it disappears when linear advance = 0.... – BarsMonster Sep 19 '19 at 09:11
  • @BarsMonster Have you disabled "pressure compensation" in the slicer? when you use linear (or "pressure") advance in the firmware the similar feature in the slicer must be turned off. – FarO Sep 19 '19 at 10:24
  • @BarsMonster: In infill is exactly where combing has this problem. – R.. GitHub STOP HELPING ICE Sep 19 '19 at 10:32
  • @R..Disabled combing everywhere - no changes :-/ – BarsMonster Sep 19 '19 at 11:00
  • @FarO: I think the only "pressure compensation" Cura has is "coasting". That should definitely be off. `speed_equalize_flow_enabled` ("Equalize filament flow") should also be off. It's a feature to compensate for lack of linear advance, and produces problematically-fast moves while extruding. – R.. GitHub STOP HELPING ICE Sep 19 '19 at 13:28
  • @R.. "Equalize filament flow" is disabled for all prints. Coasting is also disabled. – BarsMonster Sep 19 '19 at 16:15
  • @BarsMonster: Can you observe the affected layers while they're being printed (maybe even take some video so you can go back and review) to figure out what was printed *just before* the problematic parts? If you know the exact height you could do this by reading the gcode too. – R.. GitHub STOP HELPING ICE Sep 19 '19 at 17:40
  • By chance are you doing the outer perimeter before or after the inner wall? That might be somewhat related. – craftxbox Sep 20 '19 at 00:57
  • @craftxbox Tried both options. Currently outer perimeter first. – BarsMonster Sep 20 '19 at 01:21
  • @R.. craftxbox FarO Please see photo of the layer in the middle of the print... – BarsMonster Sep 21 '19 at 21:40
  • @BarsMonster: The infill looks *really* weird in that new image. Could you perhaps share the gcode file somewhere? – R.. GitHub STOP HELPING ICE Sep 22 '19 at 01:32
  • @BarsMonster Try to ask Marlin developers, maybe they know more. Interesting case. Does it happen with other STLs? From the middle layer it looks like that underextrusion alternates with overextrusion. – FarO Sep 23 '19 at 16:00
  • @R..gcode uploaded, https://s.14.by/3d_issue.gcode – BarsMonster Sep 24 '19 at 00:14
  • @FarO No issues for STLs which do not require infill... Will try to ask at reprap.org firmware section... – BarsMonster Sep 24 '19 at 00:16
  • @BarsMonster: Have you tested without whatever extension you're using to do the timelapse camera shots? It looks like it's corrupting your gcode; at least in the analyzers I'm using, the motion back from the head-partking position seems to have *replaced* the first line of each layer, and your photo seems to show material wrongly deposited along that path. – R.. GitHub STOP HELPING ICE Sep 24 '19 at 03:40
  • @BarsMonster: It also really looks like you have coasting enabled, despite having previously said you don't. Can you make sure it's off? Coasting would definitely cause the type of problem you're seeing, and make it worse with LA. – R.. GitHub STOP HELPING ICE Sep 24 '19 at 03:41
  • 1
    @BarsMonster: I tried editing out the timelapse-inserted stuff and the missing first lines are no longer missing, so it's possible my analyzers are just barfing on it but it's not the problem. It still looks like the infill ends with coasting on each layer though, leaving insufficient material when starting the next layer's outer wall. – R.. GitHub STOP HELPING ICE Sep 24 '19 at 03:53
  • Does changing slicer affect the result? – FarO Sep 24 '19 at 11:53
  • On second review, the infill doesn't necessarily look coasted; the slicer just stops the last line in weird ways (weird with respect to what I'd expect from zig-zaggify-infill) when no more gyroid oscillations will fit. – R.. GitHub STOP HELPING ICE Sep 24 '19 at 23:42
  • Could your problem just be fan? I see from the gcode your fan speed is being dynamically adjusted but is generally 75-100%. In my experience fan messes up PETG badly, and should be either off or at most something like 20-40% depending on your printer. – R.. GitHub STOP HELPING ICE Sep 24 '19 at 23:44
  • Is it only me, or the infill on the left, first photo, looks way too thick? when I print it's usually as wide as the outer layers, here it's very fat and not constant. I think there may be more issues with the printer than just some settings for perimeters or linear advance. – FarO Sep 25 '19 at 09:59

0 Answers0