Weird outline that is not existing in GCode – Hardware problem?

Ted

@phaedrux Thx for the clearification! I will try to find more information about SC and typical SC settings. Right now it is hard to find beginner friendly information.
I switched to an 24V/320W PSU for my Cetus.

@Veti Thanks for your help! I thought about an error increasing with temperature as well but had no clue about thermistors. Since you are supporting that conclusion I'm trying to find out what the correct values are. I'm not using the Cetus thermistors any more. I switched to those to have the same everywhere.

Right now I still can only find sources (only forum entries out in the internet) that point me to those values:
@ted said in Weird outline that is not existing in GCode – Hardware problem?:

M305 P0 B4725 C7.06e-8 H30 L0
M305 P1 B4725 C7.06e-8 H30 L0

I think the error when setting the heater to 170°C is only about -10°C as when setting the temperature to 167°C it already starts to block the nozzle.

For the main problem I tried to figure out how max speed, acc and jerk will influence the print quality. I tried out very sluggish settings with XYZ-Cube that turned out much better for corner sharpness. It also introduces the same bad extrusion (suppose the problem is located there!) to the cube where it wasn't that much present before. I could think of just bad jerk, acc, max-speed settings for extruder that introduced that initial problem. At least I now know why my corners were that round.

M201 X200.00 Y200.00 Z100.00 E250.00          ; Set accelerations (mm/s^2)
M566 X230.00 Y230.00 Z90.00 E120.00           ; Set maximum instantaneous speed changes (mm/min)
M203 X28000.00 Y28000.00 Z15000.00 E10000.00  ; Set maximum speeds (mm/min)

I will need to find better settings – The printer now accelerates so slow that I always have to laugh watching it.
Can someone of you recommend any process to find those settings?

Veti

@ted said in Weird outline that is not existing in GCode – Hardware problem?:

I'm not using the Cetus thermistors any more. I switched to those to have the same everywhere

that thermistor has a beta value of 3950. Your configuration is for a semitec 120-gt.

the problem with the 3950 thermistors is that the beta value is only accurate for the range of 25-50C. i have not seen good accurate values that can be used for the duet yet.

dc42

@ted said in Weird outline that is not existing in GCode – Hardware problem?:

@dc42 Thanks for your help! I thought about a wrong thermistor value as well, but for the bed where the same thermistor is used it is working perfectly. I could check with an optical thermometer, but only up to 60°C as my heatbed doesn't want to get higher. Are my thermistors defined right for that type?
M305 P0 B4725 C7.06e-8 H30 L0
M305 P1 B4725 C7.06e-8 H30 L0

Delete the H30 L0. Those two parameters are normally needed only on older Duets (06 and 085). In other respects, the parameters are correct if you are using the Semitec 100K thermistor as used on E3D hot ends.

Anyways, also wrong temperature shouldn't cause that movement failure. The most curious thing is that the print seems to recover on upper layers. That means that missing steps (hardware) cannot be the problem as the absolute positioning origin is the same for those layers.

I didn't see anything to indicate missing steps. Can you point out where you think they are?

Ted

@veti @dc42 I actually thought I bought semitecs. Oops! I will than try to replace the thermistors with semitecs (or others?) then and for the moment try to find somewhat suitable settings for my NTC 3950 100K until switched. Thanks for pointing out!

@dc42 You misread my post: I'm excluding missed steps as a cause for my problem because the print recovers on higher layers. I now rather think that wrong Acc/Jerk/Speed settings are the problem. And hope someone can recommend a good way/process to figure out what the best settings are.

Veti

@ted said in Weird outline that is not existing in GCode – Hardware problem?:

I will than try to replace the thermistors with semitecs (or others?)

semitecs are the most used due to them beeing in the e3d hotends.
therefor a good choice

Veti

since there is no real standard for B3950 they can vary a lot at higher temperatures.
using this https://www.makeralot.com/download/Reprap-Hotend-Thermistor-NTC-3950-100K.pdf
i get
β: 4666 K
C : 1.072747e-7

Veti

comparing that with https://atcsemitec.co.uk/wp-content/uploads/2019/01/Semitec-NT-4-Glass-NTC-Thermistor.pdf

170C with semitec configuration on a B3950 from the table above would correspond to 153C

Phaedrux

@ted said in Weird outline that is not existing in GCode – Hardware problem?:

And hope someone can recommend a good way/process to figure out what the best settings are.

I've got a detailed tuning guide in the works, but in the meantime, perhaps you can try some settings similar to the default Cetus values?

It looks like the values you posted in your last post match the values used by this user? https://forum.duet3d.com/topic/7709/controlling-a-cetus3d-with-duet3d-0-8-5/4

The Cetus isn't a fast printer, but I would think that acceleration values in the 400-600 range should be doable. And for jerk I would think the same 400-600 range would be alright. Print speed in the 40-60mm/s range. That's probably where I'd start tuning anyway.

Ted

@phaedrux That's the post I found them in. Will try out your values as a starting point. Thanks! At the moment the printer is much slower than usual with the settings from the post. It looks so ridiculous! It looks drunk.
I am looking forward to your tuning guide – Great that you are doing this!

Ted

I've found another problem that I want to solve first as it more pronounced, easier reproducible and might even have the same cause:

Looking at the following image (A standard XYZ test cube) the printer "lays its sausage" for the three perimeter walls counter-clock-wise around the already printed infill, starting at the most inner perimeter line and stopping at the most outer.

0_1559944666179_signal-attachment-2019-06-07-232923 (marked).jpeg

On the image I have drawn red and blue arrows for the movement. The red arrows indicate the problem area: Instead of doing a 90° corner the printer cuts the corner starting about 1 mm bevor it in a much smaller angle.

This was printed with PLA (same settings as above) at 20mm/s using Cura 4.1.0.

While it does look very small on the image, it is very noticeable and ugly holding the print in the hand. Every 90°-to-the-left corner does have a phase of 1 mm. Text printed on top of things is hardly readable.
Curious about it is also that when looking at previous prints made before switching to Duet Maestro and to Cura with the same setup no such round corners are visible. They all are crisp!

At the moment I think this is some hardware configuration problem as the problem is invisible looking at the gcode:

0_1559945302010_daaf6026-39a5-4d17-8f4d-4c197faa2a2b-image.png

I don't know what causes this. I have no clue. Hopefully you have one!

Phaedrux

Can you save that sliced file in Cura as a 3MF and post it for download? I'd like to see your Cura profile. What's the extrusion width?

I actually think it might be from printing too slow. When it's making corners so slowly, it's keeping the filament warm and soft and the friction of the nozzle has a tendency to pull it along. This is also a very small part so it doesn't get much time to cool down and solidify.

I also dislike that cura defaults to printing infill first. It has too much of an effect on the surface. I also rarely use more than 2 perimeters. All the little errors from each perimeter starts to build up. So in your case you have infill acting as a first perimeter, and then 3 more.

dc42

My suggestions:

Try setting the slicer to "external perimeters first". This usually produces more accurate perimeters.
Use a lower printing temperature.
Use more print cooling.

Ted

@phaedrux @dc42
I uploaded the files here (an xyz cube, a *.3MF file for it from Cura 4.1.0 and the GCode, all for 40mm/s at lowest temperature possible with 0.2 mm layer height using a 0.3 mm nozzle).

The prints all are made with line width of 0.3 mm (was recommended default setting from Cura for 0.3 mm nozzle).
The print is printed walls before infill (I was wrong at this!) and with inner walls first.

I printed the same cube again at 5 mm/s (~1.5h to that level) and at 40 mm/s (< 10 min to that level) as it can be seen in the following image (click to see details).
0_1560028183611_signal-attachment-2019-06-08-225113.jpeg
That the problem is present the same at the picture independent to the print speed indicates that the low speed is not the problem here.

@dc42

list item"External perimeters" first will cause bad overhangs (had that before), but I will definitely test it in the next run again.
I can't go lower with print temperature as the nozzle will clog even at 1mm/s extrusion without part cooling fan on.
The cooling was my initial idea. I've tried it with an air pressure gun from my painting compressor at first, blowing into a flexible tube fixed to my main part cooling fan. While it was fun testing, my wife wasn't impressed by the noise, the print wasn't either. No visible difference except increased stringing and an angry wife.

So far, thank you very much for your support!

Phaedrux

@ted said in Weird outline that is not existing in GCode – Hardware problem?:

0.2 mm layer height using a 0.3 mm nozzle

Try a lower layer height. 0.1 or 0.15.

Alternatively try a wider extrusion width, 0.4 for example.

Phaedrux

I'm also suspecting mechanical issue. Loose belts?

Have you tried printing any other models? Something a bit more complex? Benchy perhaps?

Phaedrux

It looks like Cura has acceleration control enabled and set to M204 P1500 so definitely higher than what you have in config.g. Try disabled acceleration control for now and control it via config.g. You can experiment with changing it mid print by sending different M204 P values to see the effects.