testing asymetric speeds per 3d printer axis

Since I mounted my new extruder and replaced that unreliable Makerbot MK6 junk,
my Thing-o-Matic got much more reliable.
I can actually think about larger prints again.
It cannot possibly reach the speeds of stare of the art printers like the Ultimater (lightweigth Bowden-Extruder and Y stepper not moving a heavy X stage) or the Type 1 (strong longneck stepper motors).

Testing a long pause

Today I tested doing long running prints by simply doing a pause.
Pause while bed and extruder are kept warm works well.
A pause with both cooled down mid print for 9 hours had the undesired effect I expected. The ABP belt cooled down and the object fell of.

Testing asymetric speeds

So now I'm trying to speed up the print job instead.
30mm/s worked fine.
60mm/s worked fine too and has become my new default speed. (Twice what I had for 2 years now.)
At 90mm/s and 120mm/s the Y stepper lost steps. It has to push around the X stage, so it bears the most heavy load.
Now I tightened everything down and greased the Y axis with PTFE.
I'm trying to limit the Y axis to 80mm/s via machines.xml
(Value ius "4800" because the unit in there is mm/MINUTE while everything else uses mm/SECOND.)
At the same time the X axis is allowed to move at 120mm/s extruding and 150mm/s travelling.

...let's see how this works out!
Print time estimation changed from 1h40m zu 1h1m .

Fist result: FAILED
ReplicatorG seems to generate correct g-code that stays below the maximum speed on each axis.
However when executing it, movements that specify multiple axis (as is the default in 5G = use of G1 commands with 5 arguments covering all 3 axis + extruder + speed) some sanity checking code comes in and disallows the movement. Even though the slower Y axis isn't moved in the command or it's part of the motion-vector is below it's per-axis limit while the total speed is above the limit.

That test is a bug in ReplictorG and can be disabled with the preferences setting
"[ ] Review GCode for potential toolhead problems before building"

Second result: FAILED

X at 120mm/s
Y at  80mm/s
Error message disabled.

Made a proper bug report and submitted a patch for the Line 420 in replicatorg/machine/Machine.java at http://replicat.org/reporting-bugs .
"We should have a response for you within one business day"
...let's see about that.
Answer a few hours later:
Ben McCallum, Dec 05 15:02 (EST):
Hey Marcus, Thanks for the email. I'll pass this along to the devs. Best, Ben
Update: One week later I need to "update" my ticket so it won't be automatically closed because nobody acted on it.

The second layer was nearly 5cm off due to lost steps on the X axis. It seems that X cannot go to 120mm/s either.

Third result: 

Limiting both axis to 80mm/s now.
Layer height increased from 0.27 to 0.4mm to reduce layer count and thus print time.
No larger nozzle to test with. ;)

I think I need some kind of wire card or cable drag chain for the ABP platform.
The cables are shaking loose at these speeds.
(Already tighened a lot of bolts. They'll have to get nylocs. That threatment helped with nuts shaking loose on my beloved RepMan. However tiny M3 nylocs will be expensive and hard to find compared to common M5.)


I'm not sure if this is the X axis skipping or of a cable from the ABP platform got entangled and prevented the platform from moving in X.
I secured the cables with tape and will try again.
At 80mm/s.

Fourth result:

Looks like when you use all of the tiny 100x100mm platform, the X+ end switch cable and the ABP cables do touch and can entangle.
Given that the steppers have much less force at higher speeds, my hypothesis may be correct.
If there was X skipping, I should have seen it happen much more often and not all at once.

...still printing.
This is a >1h print. Both to get the part done (an upgrade to my CNC) and to get reliable results by printing for a long time.

Lost a lot of steps in Y but none in X.

It's getting late.
Premature results:
maximum speed for X: 80mm/s
maximum speed for Y: 64mm/s

Maybe X can go up to 100 or 120mm/s. That remains to be tested.
Increasing layer height from 0.27mm to 0.5mm had issues with convex surface details but generally works.
Kommentar veröffentlichen