Cura - slicing causing horizontal banding? (14.03 vs. 14.12.1)

[tomax 0]

Hi folks,

 

I've been dealing with a horizontal banding issue. We have one of the early UM2's with the rear fan pin on the end of the board rather than in the middle. All belts are well tensioned and all axes are greased or oiled.

 

Initially this journey started with lots of underextrusion problems, we were printing ABS and PLA (PLA/PHA) with the same nozzle. Despite atomic cleaning, there was some material burnt to the inside and causing insulation. We constantly had to increase temp to minimise underextrusion. Robert's great extruder upgrade helped, but it it was fully solved with a new hot end and teflon spacer.

 

We had been running on CURA 14.03, and Ultimaker2Marlin 14.03 up till this point. Having put the new nozzle on, the prints were great.

 

 

• Machine UM2
  
• Material: Colorfabb PLA/PHA
  
• Nozzle: 205c
  
• Bed: 60c
  
• Layer: 0.2mm
  
• Speed: 50mm/s
  

 

Curious to know what the latest improvements to Cura, and the Ultimaker2Marlin were I installed 14.12.1 on both machines. Did a factory reset and re-started the machine.

 

The first print (a hollow test cone), with the exact same settings as above showed noticeable horizontal banding, and also the regular z-scar was wavering around. Pics attached. The banding occurs in the exact same place each time as the print proceeds.

 

Firmware:

Knowing that all settings were the same, I assumed it was some sort of firmware issue. This led me to the http://umforum.ultimaker.com/index.php?/topic/6613-firmware-surface-kwality-bug-1407-vs-1403/page-2?hl=%2Bhorizontal+%2Bbanding&do=findComment&comment=75658, and the other http://umforum.ultimaker.com/index.php?/topic/6907-horizontal-banding-on-um2/.

 

Using Winmerge, I did a roll-back on 14.12.1 Configuration.h and re-implemented the PIDTEMPBED, and tried again but with no change. I also noted that some other settings had changed between 14.03 and 14.12.1, namely PID values for the Nozzle, and the sampling rate for PID on the Nozzle. I also tried with the bed turned off.

 

I've printed 16 test cones trying various changes to firmware 14.12.1. None of these made any difference. Including Marlin Autotune for the PID bed. The banding is still evident. All slicing done with Cura 14.12.1.

 

Yet, if I return to 14.03 Cura, and 14.03 Ultimaker2Marlin all is resolved. I've checked that all editable settings are the same between Cura 14.12.1 and 14.03.

 

Cura:

As a final step I turned to look at Cura, so I uploaded the 14.12.1 firmware to the printer (using Arduino 1.06 IDE), and then sliced the file using Cura 14.03.

 

And the result is no-banding.

 

So did something fundamentally change in the slicing between Cura 14.03 and later versions, namely 14.12.1? (The movement of the z-scar position indicates some sort of adjustment) (I note that the outputted gcode files sizes are quite different)

 

Any help is very much appreciated.

 

Many Thanks,

 

Andrew.

 

Gcodes:

https://drive.google.com/folderview?id=0B2mtqlFMA-suM05wMGx1UGxwdU0&usp=sharing

TestCone_14.03.gcode (3,695 KB)

TestCone_14.12.1.gcode (2,470 KB)

STL:

https://drive.google.com/file/d/0B2mtqlFMA-suWUoyUV92S0x5aFE/view?usp=sharing

TestCone.stl (944KB)

Images now included - thanks gr5 (They had uploaded but I didn't see the publish button)

 

[gr5 2,271]

Please post a picture - to do that go to "gallery" link on top left corner of this page, then click "upload" then edit your above post or add a new one and then click "my media" next to smile face icon.

 

[pm_dude 27]

The fact of using Cura 14.03 to slice and get resolved banding issue even on latest firmware is interesting. I wonder if something in the rounding numbers or some other simplification could lead to a change of volume.

Daid could probably have some clue about that...

 

[tomax 0]

That's what I was thinking pm_dude.

I've just uploaded the original STL also, if anyone wants to test the slicing.

Many Thanks,

 

[Anders Olsson 136]

The first thing I notice is that Cura 14.03 generates two decimals while 14.12.1 generates three.

Two decimals should more or less be enough, since the UM2 has an X-Y positioning accuracy of 12 micrometers and the second digit represents 10 micrometers.

I am not sure how the UM2 interprets the third decimal, but if simply ignores it, positioning accuracy will suffer since the third decimal was then not properly rounded off to the second decimal.

So if Cura rounds of the numbers properly in both 14.03 and 14.12.1 but the UM2 ignores the last digit in the 14.12.1, I could see potential for decreased surface quality.

 

[Daid 306]

The UM2 uses the 3th decimal. Also, if you have an accuracy of 12um and you're only asking for positions with a 10um steps then you are losing accuracy, as 12 is not dividable by 10.

However, I think it's more likely that changes in the polygonOptimizer code are causing this. This code is removing access movements and has seen some changes between 14.03 and 14.12.1 (to fix issues with super high polygon models)