Linear advance calibration technique

3

Marlin has a Linear Advance calibration pattern generator, but I find it's hard to use because:

  • It only prints the initial layer on the bed, where bed irregularities interfere with accurate reading of it.
  • It doesn't do proper retraction and priming, so a mess of strings and underextruded initial segments/non-adhesion mess up the results.
  • It's hard to visually evaluate.
  • Getting a very wide range of K values involves multiple runs.
  • It's hard to clean up.

Is there a better procedure for calibration of K value for linear advance?

R.. GitHub STOP HELPING ICE

Posted 2020-01-28T22:00:17.627

Reputation: 5 311

Note that while this is a self-answer, I'm open to more/better answers too. – R.. GitHub STOP HELPING ICE – 2020-01-28T22:12:29.087

Answers

1

The same effect as Marlin's test pattern, with alternation between fast and slow segments, can be achieved with a skewed tower and Cura's overhang settings, providing the opportunity to interactively adjust K value during the print, or insert changes at particular layers in the gcode for automated test towers.

I'm using the following OpenSCAD to make a skewed cylinder:

multmatrix(
   [[1,0,1/8,0],
    [0,1,0,0],
    [0,0,1,0],
    [0,0,0,1]])
cylinder(d=40,h=40);

with Cura set to:

  • no top/bottom layers or infill
  • single walls
  • treat walls with 1° overhang as overhangs
  • print overhangs at 25% speed

These values can of course be tuned, but the extreme low overhang angle and 8/1 slope is to prevent actual physical overhang-printing considerations from messing up the test.

Vase mode might work even better for this if it supports overhang settings, since it would avoid the layer seam.

R.. GitHub STOP HELPING ICE

Posted 2020-01-28T22:00:17.627

Reputation: 5 311

Nice. I've been using the Marlin pattern in the past for Klipper's Pressure Advance, but there are some really disgusting features hidden in the g-code that's generated. – towe – 2020-01-29T07:48:55.143

1

You can use Klipper tuning tower and change the K value at each layer using Gcodes. You can find the model here: https://www.thingiverse.com/thing:3957406

I find easier to see the best results this way, also you can test way more settings in a single run, while the Marlin pattern generator lets you test only few values at time.

Remember to print without infill.

enter image description here

FarO

Posted 2020-01-28T22:00:17.627

Reputation: 2 532