New Linear Advance Extrusion Algorithm - significant print quality increases.

Discussion in 'BigBox General Chat' started by moshen, Aug 5, 2016.

  1. moshen

    moshen Well-Known Member

    Joined:
    Dec 20, 2015
    Messages:
    90
    Likes Received:
    12
    I compiled a custom Marlin firmware enabling the new LIN_ADVANCE feature for my BigBox Pro Hybrid. This results in significant quality increases when printing at high speeds.

    I printed two 25mm test cubes at 70mm/s, one with the feature disabled and the other with a 'K value' of 50 which seems to work well with the titan extruder. You can see the left print as filament bulges every time there is a direction change while the right uses an algorithm to better control filament pressure, anticipating speed changes. The part is more dimensionally accurate without the bulges.



    More info here: http://www.marlinfw.org/docs/features/lin_advance.html

    Something to look forward to in future firmware, I think it should definitely be enabled in future stock firmware. If you want try this yourself with a custom firmware build, be sure to also set the communication speed down to 115kbps to give room for the extra processing overhead required by the feature.
     
  2. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    Nice. So did you try with RC7 or RC6?
     
  3. TimV

    TimV Well-Known Member

    Joined:
    Mar 13, 2016
    Messages:
    151
    Likes Received:
    26
    I think he did it in rc7 as he has been saying that he absolutely needs this feature :)
     
  4. moshen

    moshen Well-Known Member

    Joined:
    Dec 20, 2015
    Messages:
    90
    Likes Received:
    12
    Yes, RC7 although @Alex9779 says it's in RC6 as well. I have a 200+ hr print job ahead of me and want to run fast print speeds so have been very eager to test this feature out :) So far I feel like this is one of the largest print quality increases due to firmware I've seen yet..
     
  5. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    Ok it is not in RC6. RC6 has an ADVANCE feature but not LIN_ADVANCE... not sure about the difference.
    To activate LIN_ADVANCE you don't have to activate ADVANCE too, do you?
     
  6. Stian Indal Haugseth

    Stian Indal Haugseth Well-Known Member

    Joined:
    Sep 11, 2015
    Messages:
    573
    Likes Received:
    95
    @moshen, I can not see the pics.
     
  7. moshen

    moshen Well-Known Member

    Joined:
    Dec 20, 2015
    Messages:
    90
    Likes Received:
    12
    The ADVANCE feature was an older algorithm and LIN_ADVANCE is the newer more sophisticated one. You don't activate ADVANCE for LIN_ADVANCE.
     
  8. moshen

    moshen Well-Known Member

    Joined:
    Dec 20, 2015
    Messages:
    90
    Likes Received:
    12
  9. Stian Indal Haugseth

    Stian Indal Haugseth Well-Known Member

    Joined:
    Sep 11, 2015
    Messages:
    573
    Likes Received:
    95
  10. Dr Jeep

    Dr Jeep Well-Known Member

    Joined:
    Sep 12, 2015
    Messages:
    530
    Likes Received:
    169
    I have to agree, Taken Moshen's RC7 branch, tweaked a few things for the original KS Bigbox Pro.

    Printing the same gcode as before on a part in Edge filament and before I have even played with that K factor setting (left at Moshen's default of 50) there is a very significant improvement on a part that has always given me problems with ringing on the corners.

    I'll try and attach some pics later.
     
  11. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    Ok I tried this today...
    Do I understand this correct that if you print with K=0 you have it deactivated and then you increase the value and it gets better up until some material specific point where it gets worse?
    So if K=50 is a good starting point for PLA you could also print XT or nGen and get better results even if the real K for XT would be 80?
    I mean is a curve you draw like quality above K value continuously rising until you reach the optimal value and then drops again or could it also be that lower values are worse than K=0?
     
  12. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    @moshen the part you printed is that available somewhere or a self designed test part?
     
  13. moshen

    moshen Well-Known Member

    Joined:
    Dec 20, 2015
    Messages:
    90
    Likes Received:
    12
    Alex9779 likes this.
  14. Dr Jeep

    Dr Jeep Well-Known Member

    Joined:
    Sep 12, 2015
    Messages:
    530
    Likes Received:
    169
    Duh I've only just realised that you can change K on the fly in gcode. When I get time I think I'll print a calibration piece. Do you think it can be changed mid print ? (I'm thinking here of starting at K=50 and then adding 10 to it at every 10 layers and then print a 100 layer square section.
     
    eca likes this.
  15. moshen

    moshen Well-Known Member

    Joined:
    Dec 20, 2015
    Messages:
    90
    Likes Received:
    12
    That'll be a good test - a calibration that changes in mid print. I think it'll work..
     
    eca likes this.
  16. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    Ok some impressions on LIN_ADVANCE...
    I tried some Benchys and the effect is nice though the reduced BAUD rate is a problem when printing over USB with OctoPrint at least if there are a lot of moves. On a Benchy the print head stalls at seem points, the text on the bottom is a problem.
    But also the stalls occur somewhere mid print. The problem here is that while the print head is very very slow filament oozes and then the pressure is too low. I have several outer perimeter under extrusions in a Benchy compared to a print with "high speed" firmware...

    So the corners and line endings are nicer and can still be improved by tweaking the K factor but I think it is a must to print from SD with LIN_ADVANCE at the moment... I will try this too and compare it to my current results, though this does not support my workflow I run at the moment...
     
  17. W1EBR.Gene

    W1EBR.Gene Well-Known Member

    Joined:
    Apr 10, 2016
    Messages:
    188
    Likes Received:
    26
    Alex, I would like to start playing with the firmware - which of your branches do you recommend for a hybrid dual as a starting point?
    Thanks!
     
  18. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    To play with LIN_ADVANCE you need RC7 and I have no ready built RC7...
     
  19. W1EBR.Gene

    W1EBR.Gene Well-Known Member

    Joined:
    Apr 10, 2016
    Messages:
    188
    Likes Received:
    26
    I wasn't clear - sorry. I want to start building and possibly modifiying the firmware - which of your branchs do you recommend as a starting point?
     
  20. Alex9779

    Alex9779 Moderator
    Staff Member

    Joined:
    Sep 4, 2015
    Messages:
    2,405
    Likes Received:
    731
    Ummm hard to say... the precompiled are literally out of date because all RC6... they work though...
    Just pull one for your setup compare with base Marlin and then get the latest Marlin any try to get it working...
     

Share This Page