Why mixing up axes?

[Dim3nsioneer 558]

Due to some tests I did today for checking a faster way of the retraction and priming procedure (cf. http://umforum.ultimaker.com/index.php?/topic/4094-raised-edges/page-2&do=findComment&comment=33433) I had a look into the Marlin planner (planner.cpp).

Maybe I got it wrong... is it true that there is only one velocity (I'm aware of nominal speed vs. initial speed vs. final speed of the block) calculated per block for all four axes (X,Y,Z,E)?

What kind of strange reason is there to prevent parallel independent actions of e.g. the x/y axes and the extruder?

If you use a G0 or G1 with both X/Y and E specified, the lower jerk value (which is the one of the extruder) slows down the x/y movement!

If I look at this part of the Marlin code I get the impression it is somehow a bit crappy difficult to understand (edited due to inadequate wording). It definitively has not the same level (also in terms of documentation) as other parts of Marlin which are well commented und well structured.

I.m.h.o. blocks (the trapezoids respectively) should be calculated individually for each of the four dimensions X,Y,Z,E with a coupling between X and Y. But to implement this properly, it would need quite some effort I fear...

 

[gr5 2,267]

It is VERY IMPORTANT for the X,Y,Z,E axis to all 4 move linearly. In a straight line. If the extruder ran max speed for a block it would finish early and you would get overextrusion for the begining and zero extrusion for the end of every line segment.

One way to think of this is that all moves are in a straight line in all 4 dimensions.

Even the acceleration is matched in all 4 axis so that it is a straight line even during acceleration and deceleration.

 

[gr5 2,267]

By the way there are many contributors to Marlin including some very smart people. That code you looked at has been looked over by me and probably 100 other people. It seems strange at first but it is pretty well done. It even looks 12 segments/blocks ahead of time so that it doesn't have to stop. For example if printing a circular path the old firmware (before marlin, before sprinter) would come to a complete stop at every vertex. The new software hardly slows down at all due to the "jerk" setting which isn't jerk but more maximum instantaneous velocity change at a vertex.

 

[Dim3nsioneer 558]

I appologize if I offended anyone... :oops: I didn't mean to...

I see the way it is written has some serious background. But some more comments in the last third of planner.cpp (they are really great in the first two thirds) might make it a bit easier for a stubborn Marlin-greenhorn like me to understand...

So, maybe I can get along with tweaking the jerk settings. Otherwise retraction and priming along travel movements look not very nice. Thanks and sorry again!

 

[Daid 306]

One of the reasons the code is as it is, is because it was derived from grbl. Which had only 1 acceleration setting for all axis.

Also, you need a coupling between X/Y and E for printing. As you want a certain amount of material evenly spread over a line.

What you want, is retraction/prime independent of X/Y. A way to properly implement that would be to make a difference between G1 and G0, where G1 has the axis coupled, and G0 not.

 

[Dim3nsioneer 558]

One of the reasons the code is as it is, is because it was derived from grbl. Which had only 1 acceleration setting for all axis.

Also, you need a coupling between X/Y and E for printing. As you want a certain amount of material evenly spread over a line.

What you want, is retraction/prime independent of X/Y. A way to properly implement that would be to make a difference between G1 and G0, where G1 has the axis coupled, and G0 not.

 

I fully agree! Thank you for this nice formulation.

I'll have to do some homework checking how Marlin comes from the G0/G1 command to the planner routine. This will take some time, I guess.

 

[gr5 2,267]

These gcodes are from long before 3d printers existed. G1 is supposed to be linear I believe and G0 is just supposed to move the axes in any order (uncoupled). But for Marlin it just does them both the same way. But Cura seems to sometimes use G0 and sometimes G1 as needed.

 

[Dim3nsioneer 558]

These gcodes are from long before 3d printers existed. G1 is supposed to be linear I believe and G0 is just supposed to move the axes in any order (uncoupled). But for Marlin it just does them both the same way. But Cura seems to sometimes use G0 and sometimes G1 as needed.

 

At least for the last few versions I had the impression Cura uses quite strictly G1 for printing and G0 for travelling. Daid might prove me wrong.

So the task would be to write a routine for G0 separately... quite a big task.

 

[Daid 306]

At least for the last few versions I had the impression Cura uses quite strictly G1 for printing and G0 for travelling. Daid might prove me wrong.

 

Yes, but that's mostly to increase readability of the GCode. There is no technical reason to do so right now.