Ooze/Stringing issues that can destroy prints

Warning - wall of text ahead

tl;dr I would really appreciate suggestions on troubleshooting an aggravating stringing/oozing issue on a delta printer I've recently (re)built. I suspect the E3D Lite6 or something else in the bowden configuration is set up wrongly.

Hi all,

I've recently come back to 3D printing after a few months off to build a dedicated workshop (to keep on my wife's good side). At the same time, I've rebuilt the printer, which was a stock Griffin OS, with a number of upgrades.

I'm really struggling with stringing/oozing issues for some reason. Travel moves consistently leave behind 'slug trails' that, best case, mar the surface and, worst case, foul the effector and lead to lost steps and ruined prints.

I just can't dial the problem out and it's driving me nuts. If it was just cosmetic I'd get over it but my failed print count is way too high. The failed prints happen when, layer after layer, travel moves between perimeters leave a build up of slug trails that leads to skipped steps. I'm going to randomise the layer start point in the slicer to reduce this effect but I hate doing that when the printer worked so well before I started messing with it. I'd really appreciate any ideas you guys have.

Steps I've taken to isolate cause include:

• New filament (All PLA, by the way, I have no heated bed)
• Reseated PTFE tube and installed shims under the black collet to prevent any vertical movement
• Temperature ranges from 175C to 220C
• Speeds from 75 - 150mm/s for travel, 25 - 60mm/s for printing
• First layer height from 'barely adhering' to 'mashed so completely flat the nozzle is scraping through previous plastic and causing the extruder to skip'.
• Nozzle diameter 0.4mm and 0.6mm (actually changed the nozzle, not a slicer setting)
• Pressure advance from 0 to 0.35 with lots of steps in-between
• Retraction from 2mm to 8mm at speeds of 25mm/s to 90mm/s
• Extrusion multiplier from 90% to 105%, bearing in mind the reports of strange issues with Titan and over extrusion.
• Motor currents set to 80% of rating.

I've got the following on my list of things to check - additions welcome:

• E3D Lite6 is either assembled incorrectly or defective somehow. I've cleaned it and reseated the bowden tube as best as I can, see above - Could test by reverting to J-Head, would be a pain now I've moved to magnetics.
• The bowden itself is damaged increasing back-pressure. - Could test by replacing tube, cheap.
• Some minor steps/mm error or backlash could conceivably cause the effector to not lift enough, leading to scraping through the printed layer? I've measured 100mm carriage travel and it looks ok but testing a larger travel would reveal smaller inaccuracies. Not sure what tolerance is acceptable here. At the moment, I'm using the 'book' value for my stepper/belt combo.

Calibration shows excellent consistency, I don't have the figure to hand but I think it's 0.06(could be 0.006, I'll confirm later), which I believe is acceptable. The machine is a mashup of Griffin OS and Kossel, briefly comprising:

Robotdigg aluminium corners (not the ones you sold me dc42, but the same model)
Vslot extrusions (~280mm base, ~750mm uprights)
Duet WiFi + PanelDue + IR Probe
Hayden Huntley's Magnetic Arms and Joints. (Currently on custom, Griffin-based carriages).
1.8mm steppers
E3D Lite6 + Titan extruder in ~450mm bowden configuration

Thanks in advance to anyone who made it this far!

adavidm

I find that blobs on the print at the start of travel moves are caused by too little retraction, and stringing is caused by too high an extrusion temperature. Perhaps your temperature calibration is out. What is the lowest temperature at which you can extrude 5mm/sec?

My Kossel uses 7mm retraction, 0.25mm Z lift, 3600mm/min retraction speed, and the retraction acceleration is either 1000 or 3000mm/sec^2 (sorry, I can't remember which). The Bowden tube is 600mm long.

HTH David

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Thanks David, useful insight, as always. Updates below.

Motion Configuration
With the exception of the linear rails my printer is pretty similar to yours so I'd expect similar accel/jerk to be acceptable, especially as my verticals are somewhat shorter than yours. I'll double check my settings and use your settings as a base

Temperature
I'll run that 5mm/s test when I get home.

I've got a thermocouple temp probe on my multimeter. If I trap that under the e3d sock against the heater block would that give a useful comparison to the thermistor/duet reading? The block is a genuine E3d with cartridge thermistor and I used the updated beta values you provided to E3d so I can't see temp accuracy being too far out, to be honest. I've got the PT100 sensor from the hotend you sold me as a potential replacement but I'd need to buy a daughterboard for the Duet.

Retraction
I have made progress here. Although my reseat/rebuild of the hotend has not affected the oozing during travel that much, it has helped make retraction more reliable. Following advice on this forum I used these clips to positively lock the collets on the Titan and the Lite6. By disabling Combing on Cura, which in-turn forces retraction for travel moves, I can now get smaller prints to work without too much trouble at the expense of longer print times. I've not tried Slic3r or S3D since fixing the bowden issue so that's another avenue to look at now. I'd previously done most testing with Cura 2.4 because it was giving me the best results but it can't do firmware retraction without volumetric extrusion and other small oddities so if retraction is fixed maybe it's time to move on.

Out of interest, have you tried differential retract/unretract speeds? I know you built it into RRF on request but I guess you'd use pressure advance to solve the same problem more 'correctly'? I used to have the extruder at the top of the frame (as per the griffin design) but the long, corkscrew shape, bowden that resulted looked pretty bad from a back-pressure point of view. I've since remounted the extruder at the mid-point of a vertical which I believe is a more traditional Kossel solution and there is now a single continuous curve into the hot-end. Short of a flying extruder or flexdrive solution I don't think there is anything else to do mechanically.

I've just tried putting a thermocouple probe inside the silicone sock on my Kossel. With the PT100 reading stable at 205C the thermocouple reading settles at 193C, with the probe inserted under the heater and the sock on as much as possible. I had to experiment a lot to get this configuration, because if the sock isn't holding the probe tightly in place then the reading is much lower. I think the problem is that the thermocouple leads conduct heat away from the tip too much, which is a problem if the tip isn't in very good thermal contact with the heater block. If I push the thermocouple tip against the heater block instead of putting it inside the sock, the reading settles at about 170C.

I haven't tried different retract and un-retract speeds.

My extruder acceleration is 1000. Higher might be better.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
I've just tried putting a thermocouple probe inside the silicone sock on my Kossel. With the PT100 reading stable at 205C the thermocouple reading settles at 193C, with the probe inserted under the heater and the sock on as much as possible. I had to experiment a lot to get this configuration, because if the sock isn't holding the probe tightly in place then the reading is much lower. I think the problem is that the thermocouple leads conduct heat away from the tip too much, which is a problem if the tip isn't in very good thermal contact with the heater block. If I push the thermocouple tip against the heater block instead of putting it inside the sock, the reading settles at about 170C.

Thanks David, useful to know.

What's your Z hop set to? I find that 0.2mm z hop, a wipe, and fast travel speeds combat this nicely.

Z-hop is 0.25

I think I've been fighting a couple of issues simultaneously which was making troubleshooting difficult. My current theory is that the bowden seating issue and some as yet undefined misconfiguration of Cura 2.4 were combining to lead me down the garden path.

Now the bowden issue is out of the way I'm going to focus on the slicer side. For reference, this is what I'm getting now:

[goo.gl]

As you can see, aside from the horrendous 'zits', the print quality is pretty good. I don't have the file to hand but the G-code preview in Cura has odd little jagged paths that seem to line up with the zits, suggesting something is badly broken in the config. I'm suspecting a wipe setting or similar forcing the nozzle 'though' the outer perimeter.

I'm going to repeat the print with a conservative Slic3r profile tonight and see if the problem disappears. It it does then I can be reasonably confident that my Cura config is to blame.

I'll update here but in the mean time, thank you both for your help.

Regards

David

I'll admit I have not read every word here, but those zits, I only get them when my seam alignment is random, rather than nearest or aligned (these are slic3r settings but should have cura equivalents). Prusa Slic3r even has a new option for rear for seams. Worth a shot.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Ha,

That's what I thought initially. I had deliberately swapped to random to let me actually print parts (the ooze was that bad near the layer starts prints were failing. I've since discovered that the bowden tube was seated badly and moving during retracts. Now that's fixed I'm back looking at slicer configs.

I've since printed parts with 'shortest' or whatever the Cura setting is called and I'm still getting the same zits. Going to double-check the profile settings in the gcode itself just in case it's a glitch in the GUI. I'll try and get a screenshot when I get home but it looked to me like there is a strange wipe motion in the gcode preview. I'm wondering if I've accidentally set a wipe distance that exceeds the thickness of the walls or something.

I'll see what the Slic3r result is tonight.

Thanks

Two common causes of zits:

1. Insufficient retraction. I find I need about 1% of the length of the Bowden tube plus 1mm. Too little results in zits at the start of travel moves.

2. Pause zits caused by lookahead pipeline stalls because gcodes are not being read and processed as fast as the machine is executing them. A common problem when printing curves using 8-bit firmware and sending the gcodes over USB. Less common when printing from SD card.

Reducing printing speed for outlines will help with #2 but not #1.

I have also known zits produced by Cura inserting tiny non-printing moves between printing moves, but that was in old Cura, version 15.04 or something like that.

Edited 1 time(s). Last edit at 04/20/2017 03:16PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Here is the latest:

[goo.gl]

Cura 2.4.0 on the left. Slic3r Prusa Edition on the right.

I forgot to get a screenshot once again, but Cura 2.4.0 definitely seems to insert those tiny, non-printing moves. I've assumed they are part of the nozzle wipe strategy. For now I'm going to focus on Slic3r because it supports firmware retract, which is much more convenient to tune as it can be done during a print.

I've previously tried retraction from 2 - 8mm at temps from 175 - 210C and speeds of between 20 and 100mm/s. Testing was a bit scattergun so I'm now going to run through some more methodical tests and I'll use your rule of thumb as a base. That was one reason for the question about zlift speed on the duet forums last night, I was going to add that into the mix. It seemed to me that the zlift was happening relatively slowly, allowing time for a vertical blob to form between the print and the nozzle before the nozzle starts it's travel move. You can just make out one of those blobs in the photo below, under the corner of the E3D Sock:

[goo.gl]

Edited 1 time(s). Last edit at 04/21/2017 07:08AM by adavidm.

@DC42

Sorry to be a pain. I've been looking at the way Smoothie handles zlift vs RRF. I've not come to any conclusions on which is better, if indeed there is such a thing, but the difference is big enough that I think it warrants a second look.

I've made a quick spreadsheet that I believe recreates how RRF constructs it's zlift parameters in GCodes.cpp. You can see it/play with the variables here:

[docs.google.com]

I've plugged in some fairly sensible values and arrive at a lift feedrate of 144mm/min which, I believe, is then added to the queue as a coordinated move with the extruder.


Smoothieware, on the other hand, has a "extruder.hotend.retract_zlift_feedrate" parameter, also settable via M207 Q. This seems to default to 6000mm/min, which is considerably faster. I can't pretend to understand how the motion queue in either firmware works but it does look like that value is plugged directly into the move in extruder.cpp, after the retraction has already taken place:

THEROBOT->delta_move(delta, retract_zlift_feedrate, 3);

In summary, if I'm reading the code correctly:

Smoothie: G10 -> Perform retract -> Perform zhop (at 6000mm/min) -> Travel -> G11

RRF: G10 -> Calculate feedrate based on retract -> Perform retract and zhop together (the latter much more slowly) -> Travel ->G11

If that is correct then I'd suggest that the behaviour of the same Gcodes on both firmwares is sufficiently different that a note on the Wiki might not be a bad idea. Ideally, adding options to RRF's M207/G10 to enable more flexibility would be most welcome.

If that is incorrect then I'll retract(sorry) my statement and go back to the day job..

Once again, thanks for your continued support and the wonderful work you do on Duet, RRF, IR Probes, PanelDue, PCB effectors/carriages, etc.

David

I can see 4 ways of doing a firmware retract:

1. Coordinated move, as RRF does. The Z hop will be slowed down to match the retraction time; unless you have a large Z hop and a slow Z axis, in which case the retract will be slowed down.

2. Retraction, followed by Z hop.

3. Z hop, followed by retraction. Almost certainly a bad idea.

4. Execute both the Z hop and the retraction simultaneously at maximum acceleration and speed. One will finish before the other.

It's not clear to me which of #1, #2 and #4 would be better, however with a long Bowden tube I can see that #2 might be better than the others, because retraction at the hot end won't start until a little while after retraction starts at the extruder drive. So I'll look at implemeting #2.

Should un-retraction br done the same way in reverse, i.e. undo the Z hop before undoing the retraction? Or should they stay coordinated?

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Possibly a stupid question but this refers to firmware retraction only? In which case why aren't we comparing smoothie versus RRF versus slic3r or s3d given that they also generate retractions and z hops - how do they handle it?

Edited 1 time(s). Last edit at 04/21/2017 12:30PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

You are correct. This is smoothie 'vs' RRF G10/G11.

I need to be clear that I'm not blaming RFF for the the problems I'm having. I just noticed something odd and started prodding at it.

That said, a comparison of the various approaches to retraction would be interesting, as it's such a critical part of tuning print quality.

If #2 was an option it would be like for like with smoothie, which would be useful for comparison if nothing else.

In scenario #1 If there is pressure in the Bowden then I'd guess it would act in a similar fashion to #3 as the slow zlift allows room for the hot filament to leak out.