negative scar on layer start

Hi,

I have a prusa i3.
Lately i'm having the issue you can see on the photos below.
I'm using slic3r

Settings are: Pla 190°, extruder 0.5mm, layer height 0.3, fan on, retraction disabled both on slic3r and marlin fw

These are printed at 190°:
[imgur.com]

These are printed at 215° :
[imgur.com]

The motor is not issuing a retraction (I checked looking at it while printing). The point where the negative scar occurs should be at new layer beginning

Any clue ?

Thank you.

Are coasting at the end of a line?
Would be the only reason not to have filament there with no retraction active.

Not as I know of.

Here are the slic3r extruder settings.

Note: I tried with lenght: 0 (ie disabled) with the same result.

Edited 1 time(s). Last edit at 01/06/2016 06:00PM by robse.

Since higher temps look better (from more extrusion), it might not be related to retraction, and instead it might just be under-extrusion.

Well in the pic you have set a retraction of 2mm and 40mm/s should work fine for that too.
Did you try a highe retraction on a testprint to really confirm the motor does not just skip the steps for the retraction?

what is your extrusion multiplier and the perimeter speed?

Downunder35m:

it doesn't retract. It's a direct drive extruder with the pulley really visible

I may do another test anyway

Edited 1 time(s). Last edit at 01/07/2016 04:17AM by robse.

ignacio_3D, RRuser:

0.8 extrusion multiplier and 20mm/s perimeter speed

Would it be underextrusion I would have issues in other parts of the print too

Steps/mm are well calibrated, slic3r just over extrude by default for me (and many others)

Edited 1 time(s). Last edit at 01/07/2016 04:38AM by robse.

Quote

slic3r just over extrude by default for me (and many others)

Yes, but it doesn't do it evenly. If you try and fight the slic3r bug with an extrusion multiplier you'll drive yourself nuts. If you're not confident in what slic3r is doing, then I'd recommend using different software.

Quote
JamesK

Quote

slic3r just over extrude by default for me (and many others)

Yes, but it doesn't do it evenly. If you try and fight the slic3r bug with an extrusion multiplier you'll drive yourself nuts. If you're not confident in what slic3r is doing, then I'd recommend using different software.

Not so sure about this but IIRC I had a similar problem trying out simplify3d

What about the hot-end itself ? Could uneven heating lead to that problem ? I have the impression that the negative scar appears right after the nozzle is idle for a few seconds during long moves

Edited 1 time(s). Last edit at 01/07/2016 04:57PM by robse.

Quote
robse
I have the impression that the negative scar appears right after the nozzle is idle for a few seconds during long moves

I've revised what I was going to ask, not enough to go on, so:

Can you post a link to a video of this happening?

Can you include a link to the entire gcode file? Use Dropbox etc.
If you do so, please turn on Verbose gcode, under Print settings, Output options. Believe me, this will tell us far more than you can describe, and I can run it through my gcode debugger/visualizer...

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

Also, printing it with external perimeters first, might tell you something. And, rotating 90 degrees.

I try to do it tomorrow

Hi There,

a quick update (sorry for the delay).

Printing PLA I had success in repairing the issue showed here imgur, setting "extra length on restart" on slic3r at 0.2mm

Now I'm trying PETG, great material so far, but this problem is much more visible.
Having some time to do tests, while I did see and improvement with the extra lenght on restart at 0.4mm, I tried disabling retraction altogheter.
This is the result: [imgur.com]

I'm attaching a video of the extruder printing. As you can see there is no retraction, yet the scar is there.
video 1

The layer start is at the top-right corner of the printing area (sorry I was unable to rotate the whole print with the layer start in front of the camera).


I'm clueless.

bump

Robse,
Sorry didn't see you had posted the update.
Now are you talking about the bulgy stuff showing on the top photo of the black rectangle, on the left side?

If that is the case then I think what is happening is due to the motion of the heavy Y axis.
There are a lot of things actually that can cause it.
Take a look at each of items 3-11 on the Shifted Layers page I compiled: [reprap.org]

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

Hi,

black is difficult to see, but anyway:

- The problem arise only when the extruder motor stops - pause - starts.
- When it starts, it looks like it doesn't start as fast as it should, leaving 1-2mm of empty perimeter
- It improves putting "extra restart" when retract is enabled (but I want to fix that without retraction)
- The less current the motor has, the more the problem arise. I did set the current to the maximum the hardware allowed before malfunctioning (clicks and so on)

Regarding speeds:
Higher speed is better than lower speed
Setting acceleration to the extruder didn't seem to sort any effect (from 100 to 10000)
Low accelerations make the problem much worse (?)

I have to try:
- avoiding "external perimeters first" and putting at least 3 perimeters
- changing the direct drive motor spring, maybe is too strong
- changing the nozzle to another brand/model (seems unrelated but I don't remember having this issue time ago with the standard nozzles, now using e3d ones)

I'll keep the post updated.
Bye,
robse

Quote

The less current the motor has, the more the problem arise. I did set the current to the maximum the hardware allowed before malfunctioning (clicks and so on)

Could you expand on that a bit? 'clicks' is normally a symptom of not enough current to the motors, not too much. The primary symptom of too much current is overheating, either of the stepper motor or the stepper driver. A typical direct drive extruder like you are using is capable of grinding away the filament before it stalls when supplied with sufficient current, so my thought is that you have the driver set for too little current.

I took a portion of that image, cropped and enhanced it:



So here are my thoughts:

1. You could try rotating the part by 45 degrees and reprint. This will help determine if it is related to bed movement or extrusion. If it is due to bed movement, then with the results should be different. If it is due to extrusion then there should be no change.
2. Without the Gcode it is not possible to determine much more about the motions. Please turn on verbose gcode (output options) and reslice, then post a link to the gcode. I can run it through my gcode debugger/visualizer and see line by line what is happening.
3. I'm still guessing towards a mechanical issue (belt not tight enough, Y axis acceleration, etc.)
4. In Slic3r, with expert settings on, on the printing tab, speed, Acceleration control: you can put in your default acceleration, and the a lower acceleration for perimeters. This helps to control ringing.

Quote
JamesK

Quote

The less current the motor has, the more the problem arise. I did set the current to the maximum the hardware allowed before malfunctioning (clicks and so on)

Could you expand on that a bit? 'clicks' is normally a symptom of not enough current to the motors, not too much. The primary symptom of too much current is overheating, either of the stepper motor or the stepper driver. A typical direct drive extruder like you are using is capable of grinding away the filament before it stalls when supplied with sufficient current, so my thought is that you have the driver set for too little current.

letting the motor run freely without filament (extrude 200mm from the panel) and playing with the current regulator:
1. Low current: the motor skip steps (along with the common "clicks")
2. Normal current: runs fine
3. With too much current (say high) the motor would randomly stop and start .. i've read this is due to overheating issues
4. With too much current (very high) the motor emits abnormal sounds.

I've found this behaviour (sounds on too high current) common in steppers, but if you want I can make a video with sound

I set up the current between 2 and 3, ie the maximum allowed

For the overheating part, the board is a melzi, the drivers are soldered into the board and a little heatsink is placed on them. It's possible that putting a larger heatsink with a fan would allow more current to the motor without stopping (but again, not too much current as motors do behave incorrectly in that case).
The motor (along with the whole cold-end) has a 40mm fan directed to them, I removed it to make the video. Anyway the motor is very cold to the touch too

Quote
Paul Wanamaker
I took a portion of that image, cropped and enhanced it:

Robse_Blowup_enhanced.JPG

Black is not a colour to print with

Quote
Paul Wanamaker
So here are my thoughts:

1. You could try rotating the part by 45 degrees and reprint. This will help determine if it is related to bed movement or extrusion. If it is due to bed movement, then with the results should be different. If it is due to extrusion then there should be no change.
2. Without the Gcode it is not possible to determine much more about the motions. Please turn on verbose gcode (output options) and reslice, then post a link to the gcode. I can run it through my gcode debugger/visualizer and see line by line what is happening.
3. I'm still guessing towards a mechanical issue (belt not tight enough, Y axis acceleration, etc.)
4. In Slic3r, with expert settings on, on the printing tab, speed, Acceleration control: you can put in your default acceleration, and the a lower acceleration for perimeters. This helps to control ringing.

I've attached the gcode with the problematic settings.

I "solved" the visual problem (retraction disabled so no extra-restart) by unflagging external perimeters first and doing 3 perimeters instead of 2.
This way there still is a lack of plastic but it's not visible since the layer start is inside the object and not outside (in this simple model at least).

So the issue is still open.

It would be interesting to see a single perimeter hollow cube. Presumably the walls wouldn't connect up at the seam? That's something we've all printed and know that it should work If you haven't printed one before, just make a solid cube and set up the slicer for one perimeter and no infill - much easier than trying to model a wall of the right thickness.

Paul's got way more experience than me, so go with his suggestions. I'm still interested in the extruder setup - current vs acceleration sound like good areas to dig into. As a point of reference my extruder stepper motor runs noticeably warm and provides enough torque to force abs out of the nozzle even when contacting the glass. Do you have a voltmeter to check what the stepper driver is set to?

This is what I found in the gcode:



You can see from the graph on the bottom, there are large pressure changes in the nozzle. Here is the annotated gcode for that area:

G1 X97.969 Y114.225 E263.16113 F3000.000  ; 0.294mm infill  8.848 mm3/s
G1 X96.950 Y114.225 E263.23865            ; 1.019mm infill  8.838 mm3/s
G1 X96.950 Y113.688 E263.27949            ; 0.537mm infill  8.835 mm3/s
G1 X97.893 Y113.941 E263.35374            ; 0.976mm infill  8.835 mm3/s
G1 Z3.100 F7800.000                       ; 0.300mm Z move to next layer (10)
G1 X122.990 Y115.165 F7800.000            ; 25.127mm move to first perimeter point
G1 X96.010 Y115.165 E264.78218 F1200.000  ; 26.980mm perimeter  2.460 mm3/s
G1 X96.010 Y102.835 E265.43498            ; 12.330mm perimeter  2.460 mm3/s
G1 X122.990 Y102.835 E266.86342           ; 26.980mm perimeter  2.460 mm3/s

The extruded volume per second at the end of one layer is 8.8mm3/second (for infill), and after a fairly short move the perimeter is printed at 2.4 mm3/second.
I'm not saying that is definitely the cause, but that's a pretty wide swing. If your printer had a bowden extruder then I would be more sure... but I'm not.

In general you will get more predictable performance if the extrusion rates are more constant.

I extracted these from the file: (*)
.3mm layers
infill extrusion width = 0.72mm
infill_speed = 50
external perimeters extrusion width = 0.52mm
external_perimeter_speed = 20

So you can see where the big change in extrusion is coming from - bigger width and speed for the infill.
So the only thing I can recommend testing based on the gcode analysis is to lower the infill speed a bit to bring the extrusion rate more in line with the perimeters, or just continue do what you are doing - print the inner perimeters first.

I still think it would be good to print the original way, and just rotate the part 45 degrees to determine if another mechanical issue is involved.

(*) FYI: Using these dimensions, the mm3/second calculates higher than the figures I have for mm3/second due to the extrusion multiplier. I calculate based on the actual filament extruded for each segment.

I think James' idea of printing a hollow rectangle is also a good idea, to test mechanics (if there is anything loose it would show) and to check your extrusion multiplier. See the new Slic3r Tutorials page I just put up for how to do this test, and a video demonstration. I put up a section on tuning retraction there also.

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

That ability to graph rate of flow is brilliant, I wish that was more common in g-code viewers.

It's worth mentioning that recent versions of slic3r provide autospeed mode which is intended to minimize pressure changes during the print. You lose a lot of fine grained control when you use it, so it's a bit of a leap of faith, but I've been having good luck with it and am leaning towards using it for everything now. (I originally was just using it for nylon where pressure changes are a big problem.) There's not a lot of documentation for the feature yet, you enable it by setting all of the individual speed settings to 0 and placing a reasonable speed in the max print speed field of the autospeed section, and optionally setting a max volumetric speed.

I agree that the Autospeed/max voumentric speed can help keep the extrusion rate from going over the max you set for your nozzle.
It does do that by slowing down the wide or fast extrusions.

Note that there are a few issues you can run into with autospeed:
- Not all of the speeds can be set to 0, the lower set of items (Support material and below) do not indicate that you can set them to 0, and I did have problems with bridging set to 0 - you need a speed set there at least. Support material is very similar to bridging in the way that is is extruded - unsupported - so it may need a wider extrusion, and similar specific speed to bridging depending on your extruder.

With auto-extrusion widths you can get some wide swings in extrusion width - Slic3r will try to do what it considers the maximum extrusion width your extruder can handle for things like infill (1.7x nozzle diameter) - and 1.05 x nozzle diameter for outer perimeters. This is guaranteed to produce very wide swings in pressure. However pressure control is not implemented in Slic3r (or any other slicer, yet). The autospeed feature only caps max mm3/second, this can help eliminate extrusion stalls, so that's a good thing.

So let me relate what I believe happens per my testing (particularly with a bowden extruder) with pressure swings:
- When changing from a high volume extrusion rate for infill - to perimeters (or "small perimeters" around circles) the excess pressure in the nozzle causes excess extrusion for some distance.
- When changing from low volume to high volume, the higher volume requires a higher pressure, so the beginning of the extrude is starved. You can see this some times when wide infill or bridging starts.

There are ways to handle all this. I will publish my findings after my testing is complete.

So I rather set my own extrusion widths and speeds, keeping in mind what I know is my extruder's max safe mm3/second -12, or for bridging - 20. But every extruder/hot end/plastic combination is different, and I'm using a nema23 bowden extruder and custom hot end.

You can calculate approximate mm3/second with the following formula: extrusion width x layer height x feedrate.
I think I should publish some graphs for this when I have time.

(I'm rambling again...)

Paul: wonderful software (is it released ?) and brilliant analysis.
JamesK: I did the hollow cube print and it's perfectly fine (image below). The reason is that in that case the nozzle prints continuously without pauses
Here you can see the images: hollow cube test

I did some other tests, the problem did improve putting the same width and speed for infill and perimeters (they should have the same extrusion pressure now), but it didn't solve. I tried a very different nozzle (different brand and model) too, without changes, so I rule out a nozzle problem.

I attach some pictures, I changed the nozzle to 0.4mm and material to verbatim yellow PLA. The issue is less severe with PLA in general (I have only black PETG).
It somehow improves at higher speeds (not a big difference but it's visible). I attach the gcode of the 15mm/s print


unbalanced pressure (60mm/s infill 20mm/s perimeters IIRC)


balanced pressure prints at 15mm/s


balanced pressure prints at 30mm/s


balanced pressure prints at 45mm/s

Awaiting comments. Thank you

Edited 3 time(s). Last edit at 02/03/2016 03:20PM by robse.

That's interesting. Why did the hollow cubes not pause between layers? Perhaps you used the spiral vase setting? If you just used 1 perimeter and no-infill you should have a test where the nozzle still comes to a stop after each layer and rises to the next Z.

Sorry for not going back through previous posts - did we check that the extruder grub screw(s) are tight and the gear not slipping? That can produce some very strange artefacts.

Quote
JamesK
That's interesting. Why did the hollow cubes not pause between layers? Perhaps you used the spiral vase setting? If you just used 1 perimeter and no-infill you should have a test where the nozzle still comes to a stop after each layer and rises to the next Z.

Sorry for not going back through previous posts - did we check that the extruder grub screw(s) are tight and the gear not slipping? That can produce some very strange artefacts.

I confirm no spiral vase, 1 perimeter no-infill.
The extruder does not pause enough for the artifact to happen. It looks like that the mess happens when there is a travel (>2cm), probably not only on perimeters, but elsewhere you would not notice it except on top layer (and I have some prints with top scars-after-travel that I was unable to address)

Extruder screws are tight, I added pressure to the spring (and then restored to normal) without any change in print issue. Filament doesn't seem to slip. I made an high quality video and frame-analyzed it (!) to see if there is back-stepping on layer start. The answer is no

You may want to take a look, the layer change starts around frame 50 link to dropbox


What about an hardware problem ? The melzi board perhaps or the china hot-end ? Or the motor ..
Marlin is 1.0.2

I haven't released the software yet.

Part of that software is for gcode post processing for things like pressure compensation, first segment acceleration, coast while retracting, angular hops - things that I want to prove, so they can be part of Slic3r. So as part of that I'm also working on a great many test generators for guided calibration - since there are some complex and interacting elements. I have a lot more testing to do with it.

There are multiple effects of retraction (good and bad):
- since the filament is retracted - while moving there is less ooze, and so less pressure loss when it is un-retracted.
- too much retraction and you get a cold start on the next segment - filament must reheat.
- no retraction - the beginning of the next segment after the move will be starved due to any ooze
- retract/unretract deforms the filament - so a small amount of extra length is needed.
- retract must be as fast as possible to keep the blob at the end as small as possible (and preserve pressure)

So you need some retraction, and therefore also some small amount of extra length.

However there are some things not handled - the pressure takes time to build when starting a new loop.

With all slicers there not yet a perfect way to compensate for this, but there is a very partial way:
- Under the Speed setting in Slic3r you can set the acceleration settings for each kind of extrusion. You could set the default acceleration to your Gcode's default acceleration. For instance for my big delta I use 2200. Then for perimeters I use 1500 (based on a test tower with increasing acceleration where I found where ringing was occurring). There will be a slightly shower start, and less ringing. So you could play with different perimeter acceleration settings and see if that helps.

The solution is not yet proven - but I think will be this - set an acceleration for the first mm or so of the first segment - that is inversely proportional to the pressure change, big pressure change, slow acceleration. This will give the pressure time to build. (I have already tested this, dwell is not needed). The other part of this is pressure compensation - adding or removing filament when un-retracting to raise or lower the pressure of the nozzle.

Hope that helps...

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

Grand stuff Paul, but the effect Rob is seeing seems to be much greater than normal - I wonder why. Robse, what travel speed do you use? If it's unusually slow, perhaps you are losing more to ooze than normal?

Edit - sorry that was a bit brief, had to dash off for dinner. I'm using 250 mm/s travel (but only 800mm/s^2 acceleration) for travel moves. That seems to mostly control ooze for pla and abs, but I'm still not there with nylon. This in conjunction with 4mm of retraction with a mk8 style direct extruder running at 120mm/s - I forget the acceleration.

Edited 1 time(s). Last edit at 02/03/2016 06:22PM by JamesK.