If an object with square sides is rotated by 45 degrees away from square with the bed, the X and Y components of the velocity vectors are reduced by a factor of root-2 over 2.
I have noticed significant changes in prints when changing the orientation by 45 degrees.
I can't imagine how rotating the object can help, but something you didn't mention: tighten or loosen the belts.
1
Thanks, modified the first post, forgot about the temperature as I was in a hurry.
I have photos and will add some comments to them.
From my experience, I have tried modifying jerk, acceleration and even adding weight to the plate, but did not see any improvement. In fact, lowering the acceleration and jerk too much produced worst results, as the print head would slow down too much and produce some blobs.
I don't see how changing temperature may affect this though.
If an object with square sides is rotated by 45 degrees away from square with the bed, the X and Y components of the velocity vectors are reduced by a factor of root-2 over 2.
Well okay, but you could instead just lower the velocity and acceleration, right? And if the problem is not related to velocity (likely true) but is only related to acceleration than it makes more sense to reduce acceleration than to rotate the part. But I see your point.
These are the results of the performed test. No it is not a proper lab test, but good enough as a starting point.
General print settings:
[*]Temp: 210 C
[*]Infill: 10%
[*]Wall: 0.8 mm
[*]Nozzle: 0.4 mm (stock)
All changes done via the Ulticontroller. The photo's show the front of the objects as it looks the most dramatic on that side.
Jerk
No noticeable differences between the settings.
Temperature
Printed within the advised printing temperature for this material. No noticeable differences between the temperatures.
Rotation
Although it does not really show in the picture below, the ringing appears to be a bit less, but for me not enough.
Acceleration
Changes have been made to the Acceleration setting (Control --> Movement --> Acceleration). Changed the default value of 5000 to 4000 and subsequently 3000 mm/s. At 3000 mm/s the ringing is reduced to an acceptable level. However the corner is still a bit deformed, but that is probably caused by some over extrusion.
I also tried to lower the individual Acceleration setting for the X and Y axis, but that gave no noticeable differences (no photo included).
Conclusion (for now)
The Acceleration setting helps reducing the ringing effect, however I do not completely understand how lowering this setting is affecting the overall performance of the printer. What are your ideas?
Furthermore I assume my belt tension is good, but to be completely honest I have some doubts. However, if I am able to improve the tension, the Acceleration setting appears to be the parameter to tweak with respect to ringing.
The individual axis acceleration settings are maximums for that axis. The overall acceleration setting is the one that is actually used, unless the acceleration maximum for any axis involved in the move is less than the overall setting, in which case that lower value is used.
Great work Peanudt
Thanks for your efforts. I'll give that lowering acceleration a go. I've been printing at 60mm/s so it may make a greater difference at these speeds.
Nice. Do you have a larger version of the final photo (the one that shows difference due to acceleration)?
f=ma. force = mass X acceleration. The mass is the weight of the print head. So there is a direct correlation between the amount of force on the belts and the acceleration. Twice as much acceleration means twice as much force on the rubber belts which means more stretching (potentially twice as much stretching).
When the acceleration stops suddenly at the corner it's like being in a car stopping at a stop sign. You really notice it when the acceleration stops suddenly at that moment when the car goes from .00001 miles per hour to 0 miles per hour because at that moment the acceleration goes through a very fast change.
Same thing when the print head goes around a corner.
This very high change in acceleration is called "jerk" by physicists but Marlin's meaning of the word "jerk" is different.
Anyway, at this moment of high true jerk, (the corner) the acceleration on one of the axes goes from stretching the belt to zero force. Like pulling back on a guitar string (the force) and letting go (no force). It vibrates, or rings at it's resonant frequency. You can lower the frequency by adding mass or loosening the belts. You can increase the frequency by removing mass or tightening the belts.
But note that adding or removing mass also increases or decreases the force on the belts so it's usually better to remove mass!
There are ways to mechanically dampen the harmonic oscillation - all involve increasing friction. For example if you glued a small box half full of sand (of the same weight as the print head) to the print head that would absorb a large part of the oscillating energy and it would probably only give you a single ripple (or not even that). You could design belts with material that don't store energy very well. For example belts that don't stretch. Think of a ball designed to not bounce very well. Like a wood ball. Or a tennis ball half filled with sand. If one type of rubber ball bounces to 80% of it's original height and another only bounces to 20% of it's original height - the second rubber would be the kind to use to reduce ringing.
Another way would be in software: decrease the acceleration gradually instead of suddenly at the corner. Aka "true jerk". It would be nice if Marlin worried about true jerk (optionally) for people who don't like the ringing and are willing to print a little slower. People tried this in firmware in the past and the problem was the print head pretty much came to a complete stop at every vertex. Fine for a square but for a "circle" with 100 segments this is crazy slow. But maybe cutting the acceleration in half is good enough for most people.
Right now Marlin defines "jerk" (roughly) as the maximum change in speed at a vertex. This lets it go around a circle fast as there is very little change in speed at the vertexes of a circle. Maybe Marlin needs also a similar setting for the maximum acceleration occurring just as you approach or leave a vertex (and max true jerk to achieve that acceleration).
I haven't read the whole thread yet, but I wanted to add that ringing is more related to the friction of the molten plastic coming out of the nozzle, than to loose belts. the "tighten your belts" paradigm can only be taken so far, and by now I am convinced that users are actually over-tightening the belts, instead of looking elsewhere for a problem.
the XY carriage is relatively solid, some play in the system, but not too much.
the print bed on the other hand, is one of the flimsiest constructions, in other words, the print head makes the print and the bed ring, not the other way around.
work arounds are Calins method to attach 2 rght angled extruded alu pieces lect and right to the arms, and use M4 or thicker screws to hold the print bed, and or increase the mass of the print bed by 200-500gr.
another work around would be a complete reconstruction of the Z stage.
Or clamp the frame of the UM to a 400 pounds weight.
Everything I wrote about the belts causing oscillations applies if it's the bed instead that is oscillating except it's much easier to fix the bed mechanically to make it stop oscillating than the belts and print head.
So acceleration of 3000 gives better results? it's funny that it used to be Marlin's default setting before 
I'll give it another shot because I was mostly messing with the jerk value thinking it was the most responsible for ringing.
Thank you for sharing. I tried to do similar tests a month ago, but for some reason, it was one of those days where nothing worked: print not sticking to bed, under extrusion, nozzle getting clogged...
Nice. Do you have a larger version of the final photo (the one that shows difference due to acceleration)?
Here you go, hopefully you can zoom in some more.
still tiny :???:
Sometimes technology tries to help too much.
You can click on "gallery" on the top left corner of this web page, then upload your picture there and then when you are making a post click on "my media" to insert it.
Update: second set of test results
Thanks for your thoughts. I agree with the reactions above that the ringing effect is caused by the movement of the print head and the resonance it produces that is transported throughout the whole Ultimaker (and bed) and the underground.
As the new fan duct is heavier than the stock version, so I wanted to check what the effect is of the weight difference. Furthermore the reactions above made me think about how the resonance is affected by the underground. Does a sturdy base prevent ringing or do you need some flexibility to absorb the forces.
And of course the print bed is a potential source of resonance. My assumption is that print at the front of the machine are more vulnerable to resonance effects compared to the back of the platform.
First of all, I tightened all belts and placed some belt tensioners. There was a bit of slack in the previous tests, but all tests were done with the same slack.
Stock vs Printed Fan
Althoug the results look the same, the top part is printed with an Acceleration setting of 5000 mm/s with the standard fan and the bottom part is printed at 3500 mm/s with the following fan duct: http://www.thingiverse.com/thing:69327. I will use the stock fan for now as the weight appears do affect the ringing in my machine significantly.
Base
My Ultimaker was placed on the corner of the desk. My idea of heaving a sturdy base and prevent wobbling in the middle of the desk.
For the bottom print the Ultimaker was placed on the floor. A wooden floor with some insulation beneath. The top part was printed with the UM located in the middle of my desk.
Surprisingly the UM in the middle of the desk produced the best results. Apparently the resonance of the machine is partly absorbed due to the fact that the underground is able to resonate as well (and thus absorbing some of the forces produced by the machine).
Location on the bed
I printed a test cube at the front of the print platform and one at the back. As you can see below the print on top (back of the platform) is less affected then the print from the front of the bed (bottom print).
Conclusion (after second set of tests)
The parameters to work with for my machine:
[*]Print objects preferably at the back of the platform (new Cura setting: load a model and place it in the back of the platform by default?)
[*]Do not use a stiff underground/base
[*]Acceleration settings
[*]And of course: proper belt tension
still tiny :???:
Sometimes technology tries to help too much.
You can click on "gallery" on the top left corner of this web page, then upload your picture there and then when you are making a post click on "my media" to insert it.
Used the galley, but I think the original is not that large.
Great research. You are such a scientist. In a good way of course.
I wish you tried clamping the UM stiffly to something immovable. I'm curious if that is worse than the "middle of your desk". I can see how putting the UM on for example a slippery surface could actually help more than hurt.
Another thought I've mentioned in these forums before is it would be interesting to put a weight on the long belts at the point opposite the clamp. The two weights on the X axis should add up to the weight of the print head and same with the Y axis. Then moving the head would not vibrate the UM itself nor the table it sits on, nor the platform. On the down side it would make for twice the work for the steppers.
Another thought I had while reading posts above was that oiling the rods might hurt matters. A little friction can help dampen oscillations.
I agree with Joergen,
You can see the bed flying around all over the place during fast head moves, and mine is screwed down so
far the the plate is actually touching at the back.
I would also suggest that we should try to mount idlers in the mid-span of the long belts to reduce resonance.
However I think for packaging reasons this will be tricky.
Regards
C.
Hi, you should do one more probe. Test squares varying only the intensity of the drivers. (I have prusa, but I don't know if in ultimaker it can be done). Excesive holding torque could affect, i think.
Needs more rigid belt. At least 4 times wide belt or 4 belts in parallel. Or pneumatic or magnetic oscillation damping...
ArunC posted a topic in UltiMaker Cura,
.thumb.jpeg.0b7a05eafc09add17b8338efde5852e9.jpeg)
Dustin posted a topic in Firmware,
Recommended Posts
gr5 2,170
I can't imagine how rotating the object can help, but something you didn't mention: tighten or loosen the belts.
Link to post
Share on other sites