Bug in Marlin; wonder if anyone has any familiarity with it.

ThantiK · December 1, 2015, 5:19pm

Bug in Marlin; wonder if anyone has any familiarity with it. My boss was going through the Marlin code and noticed what he deemed was a “glaring” bug. From what I’ve been told, when the Delta movement on a large straight line is being moved through, Marlin completely ignores PID hot end heating. We were wondering why on large, flat prints we were getting huge temperature spikes, and we couldn’t ever seem to get our temperature under control.

Has anyone got any familiarity with this particular bug? We have it fixed now, internally and I’ve convinced my boss to do a fresh pull of Marlin, and upstream our bug fix, but if it’s already been fixed we can ignore it. Anyone heard of this issue before, or familiar with any commits which may have fixed? Admittedly we are using a pretty old branch. @Daid_Braam , @Whosa_whatsis ?

If it’s something new, we’ll do a proper pull request and upstream our fix.

ThantiK · December 1, 2015, 5:22pm

Also @Shai_Schechter , dunno if you’ve run across this or if you guys use Repetier. Ever notice this?

Daid_Braam · December 1, 2015, 5:52pm

@ThantiK I’m not familiar with the delta additions. But any loop should call “handleHeater” (or something like it)

Shai_Schechter · December 1, 2015, 6:56pm

@ThantiK Haven’t noticed it, but we’re moving away from Marlin. It’s just a hassle to work with and it’s becoming noticeably outdated. Jim Morris is doing a fantastic job with Smoothieware, have you tried it? It’s such a straight forward configuration unlike marlin that makes you pull your hair out sometimes.

Arthur_Wolf · December 1, 2015, 7:12pm

+Peter van der Walt @ThantiK Marlin has been around for a long time, and is used by many people. It’s surprising that a bug like that slipped through during all that time …

Wolfmanjm · December 1, 2015, 7:30pm

That is surprising (unless they changed it recently) all delta moves are segmented into tiny segments so there should be no long segments on a delta in Marlin.

Wolfmanjm · December 1, 2015, 7:32pm

FWIW Marlin also does delta incorrectly IMO. They calculate speed and acceleration in delta actuator space not in cartesian space, this causes inconsistent heads speeds and produces odd artifacts. (again based on the old Marlin, I have not looked at the current code).

Daid_Braam · December 1, 2015, 7:51pm

@ThantiK I cannot find proof of what you are claiming. (even tough Marlins delta support is a hack and has the problems @Wolfmanjm says. And the general code is a mess)

The delta calculations block here:

github.com

MarlinFirmware/Marlin/blob/1f20516dad3343b9589b56e59bf5fb1871c0144b/Marlin/Marlin_main.cpp#L4072



	
	calculate_delta(destination);
         //SERIAL_ECHOPGM("destination[X_AXIS]="); SERIAL_ECHOLN(destination[X_AXIS]);
         //SERIAL_ECHOPGM("destination[Y_AXIS]="); SERIAL_ECHOLN(destination[Y_AXIS]);
         //SERIAL_ECHOPGM("destination[Z_AXIS]="); SERIAL_ECHOLN(destination[Z_AXIS]);
         //SERIAL_ECHOPGM("delta[X_AXIS]="); SERIAL_ECHOLN(delta[X_AXIS]);
         //SERIAL_ECHOPGM("delta[Y_AXIS]="); SERIAL_ECHOLN(delta[Y_AXIS]);
         //SERIAL_ECHOPGM("delta[Z_AXIS]="); SERIAL_ECHOLN(delta[Z_AXIS]);
         
	plan_buffer_line(delta[X_AXIS], delta[Y_AXIS], delta[Z_AXIS],
	destination[E_AXIS], feedrate*feedmultiply/60/100.0,
	active_extruder);
}
#endif // SCARA
  
#ifdef DELTA
  float difference[NUM_AXIS];
  for (int8_t i=0; i < NUM_AXIS; i++) {
    difference[i] = destination[i] - current_position[i];
  }

In the plan_buffer_line function, as that’s the buffer that fills up and blocks other execution. But this deals with the PID controller just fine:
https://github.com/MarlinFirmware/Marlin/blob/Release/Marlin/planner.cpp#L544

ThantiK · December 1, 2015, 8:16pm

@Daid_Braam , as I said, we’re using a pretty old version. It looks like this has already been fixed in that case. We’ll double check on our end – thanks for taking a look. You’re more familiar with the code than I am. I’d love for us to move to Smoothie, but that’s a logistical problem on my end.

Daid_Braam · December 2, 2015, 6:12am

got a link to your fork? as this is very old functionality

Billy · December 2, 2015, 6:35am

During a long extrusion you are extracting a lot more energy per second from the heater block. It’s a very different scenario to when auto-tune is typically ran. Have you tried halving your speed during the fill? This should make the problem even worse if your theory is correct right? If it fixes the issue then maybe it’s PID setting.

ThantiK · December 2, 2015, 12:28pm

@Billy , I’ll make a test gcode that extrudes very little, across the entire length of the bed slowly, and record the results.