i3 Suddenly OverExtruding: ideas?

[clough42]

Anyone have any ideas why my i3 might be suddenly over-extruding?

I've got a Prusa i3 (rod version) that I've been running for about a year with the Itty Bitty Belted (single) extruder. It's been running well, but it's been sitting for a couple of months while I worked on other projects. Now that I've started it back up, it's overextruding significantly. Plastic is piling up and the nozzle is plowing through and re-melting the excess plastic. It's oozing out over the sides of the parts and making a real mess. It's probably over by 10-20%, so after a couple of layers, the parts start to look like Dr. Seuss.

I've been through everything I can think of, and I haven't found the cause. Any idea what might have happened? Where else should I look?

Here's what I've tried:

• Sliced with Slic3r 1.1.7 stable and 1.2.9 stable, with the same results.
• Installed the printer on a flat Travertine tile and re-leveled the X axis.
• I'm using ABL, and the loops measure with a micrometer between .21 and .23mm all the way around the bed. First layer should be .2mm.
• The Z axis is moving correctly. 50mm vertical movement measures 50.01mm with digital calipers.
• X and Y steps are correct and parts are the right size.
• E-steps is still 644.2, the same as it was set up originally.
• Extruder is moving filament 98mm for 100mm commanded extrude, so it's under-extruding slightly.
• Filament measures between 2.93 and 2.98mm, measured at various angles.
• I can't find any compensation settings in the slicer or in the printer that are set to anything but unity.
• Motor is running smoothly. No step or bolt skipping.
• Cleaned out the bolt with a dental pick.
• Sitting next to the printer and watching it print, I'm noticing nothing unusual, except the overextrusion and the eventual clatter of the nozzle running into the cooled, overextruded plastic.

I tried to calibrate the extruder, with odd results. At 644.2 e-steps, it's extruding 98 instead of 100mm, so I calculated the new value of 657.3. At 657.3, it extrudes 103mm. Re-doing the math with 657.3 e-steps, extruding 103/100, it says I should be at 638.2, which is less than where I started. Bumping it up and down slightly seems to be resulting in non-linear response, with no change for small differences followed by a big step in the response. I can repeat the same test over and over and get exactly the same result, so it isn't intermittent.

Please check my math. I must be overlooking something obvious.

Running through Triffid Hunter's guide, with rectilinear fill at 95%, I had to pull all the way back to 550 e-steps before I started getting gaps. That's 15% down from a setting that was only 2% off by measured filament movement and working previously to print all manner of parts.

I can keep tuning by printing, and I'll zero in on something that works (probably around 575), but why is it so far off from the measurements at the extruder?

Is there some global compensation value somewhere that I bumped and I'm forgetting?

Tool chain: STL -> Slic3r 1.1.7/1.2.9 -> Octopi -> Marlin (8 months old, not re-flashed) -> RAMPS 1.4

[Roxy]

Do you need your EEPROM turned on? If you rebuild your firmware with it turned off, there is a fair chance the problem goes away. This might be like people trying to get Auto Bed Leveling going and their firmware is picking up a bad value from the EEPROM.

[clough42]

Yes, I am using the EEPROM so I can adjust values on the fl.y. I'll try resetting it

[clough42]

Hmm...this is getting bizarre. I think I may be dealing with a dimensional control issue as heat soaks into the printer mechanics.

Here's what I get if I print a single part in the middle of the bed (10x magnified):

2015-06-30 07.10.14.jpg2015-06-30 07.14.30.jpg

And here's what I get if I print a whole bed of 21 identical parts:

2015-06-30 07.08.06.jpg2015-06-30 07.15.11.jpg

The only difference between these parts is that the first one was printed by itself (about 12 minutes) and the second one was printed as a part of a batch that took 3.5 hours. I used Slic3r 1.2.9 stable with the same settings for both (3 solid layers, 2 perimeters, 40% infill).

My working theory is that as the print continues, heat is soaking into some part of the printer and closing up the Z height of the nozzle (warping the bed, expanding the hot end, something else?). The part should be exactly 7mm high. When I print one by itself, it comes out clean with a measured height of 7.11mm. When I print the entire plate, the bottom layers look squished and the overall height is only 6.55mm. The upper layers of the part look fine and the printer LCD says it's at 7.00mm after the print.

So I'm hypothesizing that my Z distance is closing over the first 30 minutes to an hour of the print, squishing the first few layers. The single part finishes printing before the effect is noticeable.

Has anyone else seen anything like this?

[Renfro]

DISCLAIMER: I have no idea what am doing helping out here (I just got into this), but sometimes an uninformed outsiders viewpoint can shed light on obvious oversights by long time experts. Programmers experience...

I was observing a similar variations when calibrating my extruder. It turned out to be a jerky bearing against the filament. I've adjusted the spring tension and spun the bearing on a high speed drill while lubricating it. It seemed to help and now I get a more consistent extrusion calibration.

This has nothing t o do with Z height deficit on your new findings, so I might be (and probably am) completely off the ball park here.

[clough42]

DISCLAIMER: I have no idea what am doing helping out here (I just got into this), but sometimes an uninformed outsiders viewpoint can shed light on obvious oversights by long time experts. Programmers experience...

I was observing a similar variations when calibrating my extruder. It turned out to be a jerky bearing against the filament. I've adjusted the spring tension and spun the bearing on a high speed drill while lubricating it. It seemed to help and now I get a more consistent extrusion calibration.

This has nothing t o do with Z height deficit on your new findings, so I might be (and probably am) completely off the ball park here.

No problem. Jump right in. There's no better way to learn, and the outside viewpoint is often quite helpful.

In this case, I think the bearing is okay. I'm getting consistent (ly bad) results, and I think if it were dragging, it would cause underextrusion.

I'm thinking at this point I need to let 'er run and take readings with a caliper between the shelf and the bed at various points to try to confirm the Z height issue. Still not sure that's happening, or what might be causing it.

[printbus]

Clough42 - it may not be related, but there have been a few people who determined their heat bed would flex as the heater cycles on and off. That's the closest I remember anyone saying they see the Z axis changing with temperature. In those cases the varying Z-axis height led to print quality issues throughout the entire print, not just the beginning. Maybe in your case the Y-bed warms up with run time and keeps your heater flexed upwards.

Are you running with the same glass as before? Same clips and clips installed in the same place? Ambient room temperature about the same as before?

[clough42]

Oh, man...you may be on to something there!

That's a definite possibility. First layer is at 100C and subsequent layers are at 110C. I'm testing my Z height by measuring the thickness of the loops with calipers, but those are printed at 100C before ramping the bed up to 110C for subsequent layers. I should be able to test that by printing the same model with the same settings in PLA on a cold bed. It doesn't totally explain everything, but it suggests a great test to learn more.

I also noticed what looks like Z banding on my prints, but under close inspection, the layers aren't offset--some of them are just wider. The bulging layers bulge all the way around the part. That could be long-term oscillation of the bed shape due to heater cycles. I'm running bang-bang control of the bed, so it could definitely be related.

I could probably put a dial indicator on the center of the glass and just let the bed run...

And yes, the printer is in a new location, near an AC vent in an exterior corner with outside temps hitting 111F. I thought I had the air off the bed, but who knows. Worth checking.

Clough42 - it may not be related, but there have been a few people who determined their heat bed would flex as the heater cycles on and off. That's the closest I remember anyone saying they see the Z axis changing with temperature. In those cases the varying Z-axis height led to print quality issues throughout the entire print, not just the beginning. Maybe in your case the Y-bed warms up with run time and keeps your heater flexed upwards.

Are you running with the same glass as before? Same clips and clips installed in the same place? Ambient room temperature about the same as before?

[clough42]

Okay, looks like Printbus was right. The bed is warping and closing up the gap. I have my heat bed hard-mounted to the wood with standoffs, since I'm using auto-bed leveling. I put a machinist's dial indicator on one corner of the bed and turned on the heat. The relative expansion of the bed and the wood mount are causing the issues. Here's what's happening:

1. The bed starts heating.
2. The heat bed expands horizontally, pushing on the standoffs, causing the wood to bow downward, lowering the bed .012"
3. The printer probes the bed and establishes the zero point.
4. As the print runs, the heat soaks through into the wood, causing it to expand over the next hour. This flattens out the warp and pushes it the other way, raising the bed by an additional .021".

The rising bed compresses the print in progress by a total of .033", or .84mm over the course of an hour or two, causing the parts to appear massively overextruded.

I think the solution is going to involve relieving the stress by soft-mounting the heat bed in a way that it can expand. I'll try the springs. That may be enough. If the bed is warping because of the heat gradient across it, another material might be called for. Aluminum wouldn't build up a gradient because of its conductivity.

I'll post as I learn more.

[clough42]

Some photos of the process...the setup:

IMG_2224.jpgIMG_2229.jpg

After the preheat to 100C:
IMG_2228.jpg

And an hour and a half later, after soaking at 110C:
IMG_2243.jpg

Visible warping of the bed mount:
IMG_2231.jpg


I have another test running with three of the bed screws removed.