I though this question had moved beyond what I had to contribute, but since it is still not solved, I decided to follow through with running g-code on my machine. What I found was interesting. I have a Rambo 1.4 board and am running Marlin 427D headless.
-
Running Kinan’s Latency delay troubelshooting test Tolenarnce 0.002mm.gcode - It jittered on my machine exactly like his video.
-
Running Kinan’s Latency delay troubelshooting test Tolenarnce 0.055mm.gcode - It jittered also. The jittering was regular, almost like it was pausing at the end of each line segment used to emulate the curve.
-
Running the crown test g-code Ryan authored - I believe that it happened occasionally. Given the much smaller movements and the legitimate reasons for deceleration, it was difficult to tell.
-
Estlcam authored enlarged crown - I created a crown much larger than Ryans. I saw the occasional pause on long arcs where I would not expect a pause. It wasn’t a jitter, just occasional places on a continuous arc where I could hear the steppers wind down and it could see the tip of the bit pause just a bit.
-
Fusion 360 authored curve - I authored a large closed spline in Fusion 360 - I noticed a pause in just a few places…fewer than the large crown test.
Some observations:
-
Given that it happens on my machine running headless without any USB cable, it point away from any kind of electronics issue specific to Kinan’s machine.
-
Kinan’s g-code has 6 digits after the decimal point, Ryan’s crown and the enlarged crown had 4 digits after the decimal point, and the fusion 360 generated g-code had 3 digits after the decimal.
-
I cannot see these pauses from EstlCam or Fusion generated g-code being noticed when using a router, but they might leave an artifact when using a laser.
Is it possible that newer versions of Marlin just have more overhead and therefore less processing time for floating point calculations? I love to see this test run with an old version of Marlin.