Upgrading to ARM and 0.9 Steppers - Vertical Banding Issue

[int2str]

The banding seems about the same on my printer (see part on the right):
[img]http://ironcreek.net/photos/_s/1024/Hob ... G_6080.JPG[/img]

I thought that's just part of the process.
Could be extrusion pressure related patterns or rounding errors.

Never rally bothered me that much. Would be curious if there's a way to mitigate this though...

[barnett]

I have seen this banding even before I installed the magnetic arms. You can sort of see it in this post.

(Click the images to zoom in closer)...

This scrabble tile holder was also pre-magnetic arms, but the bands line up with the fill pattern, so I suspected that.

scrabble tile holder

But it can't only be the fill pattern because I always see banding on the single wall test cube.

single wall

I can't remember ever seeing banding in a spot where the layers don't stack vertically.

no banding

It's so minor I never tried to fix it. But now I'm wondering what it is. Could it be little nudges in speed from the steppers motors? Or the belt not meshing perfectly with the teeth on the pulley?

[mhackney]

I'm wondering if this is due to the line fitting algorithm in the firmware. The slight rounding errors would be reproduced layer to layer on an edge on these successive layers where the edge is the same on each layer. Make sense? This might be an artifact of the delta geometry calculations. On cartesian printers, lines on X or Y should be perfect since the math is simple. But any diagonal or curved line might exhibit this.

I will do some experiments tomorrow by drawing some simple geometries that break up the regularity of the side.

[xnaron]

I'm wondering if this is due to the line fitting algorithm in the firmware. The slight rounding errors would be reproduced layer to layer on an edge on these successive layers where the edge is the same on each layer. Make sense? This might be an artifact of the delta geometry calculations. On cartesian printers, lines on X or Y should be perfect since the math is simple. But any diagonal or curved line might exhibit this.

I will do some experiments tomorrow by drawing some simple geometries that break up the regularity of the side.

I agree and think it has to do with the delta calculations. I have tried rotating the prints on the build surface at various angles with the same results for the banding. I was thinking of trying the repetier firmware with the Arduino Due ARM board. This will be much faster than the 2560 we are using now. This means we could turn up the segments:

/** \brief Number of segments to generate for delta conversions per second of move
*/
#define DELTA_SEGMENTS_PER_SECOND_PRINT 180 // Move accurate setting for print moves
#define DELTA_SEGMENTS_PER_SECOND_MOVE 70 // Less accurate setting for other moves

I think I will go ahead and order one and a ramps-fd. I want to play with one anyways so this will be a good excuse to get one

[mhackney]

Yes, every segment on a delta has to be calculated with trigonometry so rotating the object might be expected to have the same artifacts. The X and Y movements on a Cartesian printer are very simple and not subject to rounding or other errors. I think we are on to something here.

Another test might be to lower the segments per second by 1/2 and see what effect that has. You would expect the vertical lines to get further apart. The comments on these two lines is revealing. On a movement, all you care about is arriving at point x,y,z and not how it got there and you want to get there fast so fewer calculations would accomplish that. For a print move, calculating every move in micro steps would be very accurate but require a lot of computational power. The controllers in CNC mills basically do this to get the smoothest moves possible - but they have a lot more horse power.

Now, I wonder what a delta segment is? If you are printing at 50mm/s does this mean that 50mm/s divided by 180 segments/s yields .278mm - that's actually pretty big. I think we are on to something.

[xnaron]

Yes, every segment on a delta has to be calculated with trigonometry so rotating the object might be expected to have the same artifacts. The X and Y movements on a Cartesian printer are very simple and not subject to rounding or other errors. I think we are on to something here.

Another test might be to lower the segments per second by 1/2 and see what effect that has. You would expect the vertical lines to get further apart. The comments on these two lines is revealing. On a movement, all you care about is arriving at point x,y,z and not how it got there and you want to get there fast so fewer calculations would accomplish that. For a print move, calculating every move in micro steps would be very accurate but require a lot of computational power. The controllers in CNC mills basically do this to get the smoothest moves possible - but they have a lot more horse power.

Now, I wonder what a delta segment is? If you are printing at 50mm/s does this mean that 50mm/s divided by 180 segments/s yields .278mm - that's actually pretty big. I think we are on to something.

I think the delta calculations really push the limits of the current boards. I think the segments can be bumped up past 180 a bit as well with the current boards (arduino atmel).

I ordered this http://www.ebay.com/itm/RAMPS-FD-shield ... 417b980b67 It is an ARM based board running at 84Mhz. The latest repetier 0.91 has Arduino Due (ARM) source. I am hoping that I can bump the segments up a lot. This board is running over 4 times the clock speed of the atmel processor we currently use. I'm not sure that will result in 4x the performance though. For $55 it is worth a try as I have another use for it if it doesn't work. At this point I have no idea if the arm code is functional or not.

[mhackney]

I just checked my .80 repeater and it has 200 segments/second for both print and moves. I think it's worth lowering the print value and seeing what that does. And maybe start tweaking it up and see if there is enough Umph to make a bit of difference.

[xnaron]

I just checked my .80 repeater and it has 200 segments/second for both print and moves. I think it's worth lowering the print value and seeing what that does. And maybe start tweaking it up and see if there is enough Umph to make a bit of difference.

Sounds like a good idea. I think 200 is very close to the upper limit from what I have read.

[mhackney]

Ok, so I ran a test this morning. I changed the DELTA_SEGMENTS_PER_SECOND_PRINT from 200 to 100. It does seem like the lines are spaced further apart. I tried to mark the lines on each cube and photo them:

[img]http://mhackney.zenfolio.com/img/s10/v9 ... 7393-3.jpg[/img]

The original 200 segment/s cube has a clear 5 lines on the end, the 100 segment/s cube has 3-4 and they do appear to be further apart. I compared it to the others I printed last night. The ones last night were all very consistent in the spacing. I'll print a few more and maybe lower to 50 segments/s.

[Jimustanguitar]

Another naive possibility.

Slight Temperature Variations.

If it's not the segments setting, are we seeing the PID pulses that heat the hot-end?

[mhackney]

I doubt it, they are too regular and unless everything were just right, any PID pulses would not produce such a regular pattern. My money is still on the math in the firmware. Could be simple accumulated rounding errors or the delta segments is still suspect given that it had a visible effect on printed parts.

[mhackney]

I think you may have surfaced a systemic problem. Take a look at this photo:

[img]https://s3.amazonaws.com/ksr/assets/000 ... 1371739182[/img]

from the kickstarted OpenBeam Kossel Pro. That sure looks similar. I'm going to peruse the web for other delta prints and see if a pattern emerges.

[xnaron]

I think you may have surfaced a systemic problem. Take a look at this photo:


from the kickstarted OpenBeam Kossel Pro. That sure looks similar. I'm going to peruse the web for other delta prints and see if a pattern emerges.

Yes It has been bugging me for over a year. I feel a little better knowing it is not something specific to my build. I hope as we move to faster controllers that the algorithms will get refined. Ideally if we could get rid of this artifact the DELTA printers have the potential for almost flawless surface finish.

I ordered the rev 1 A ramps-fd board from China. It is not the final version of the board from Bob but it should hopefully be workable. I also have started configuring the firmware and have successfully compiled an the arm version of repetier. Ideally I'd like to be able to turn up the segments to really high values and see what effect this has on the banding. Now the long wait for the hardware from China...

[mhackney]

Now that I'm looking for this, I am seeing vertical banding everywhere - even on cartesian printers: check out the purple box made on a Makerbot.

[xnaron]

Now that I'm looking for this, I am seeing vertical banding everywhere - even on cartesian printers: check out the purple box made on a Makerbot.

Yes I can see it very very faintly on my cartesian prints. It is so faint it is barely noticeable. Certainly not as bad as what you pictured in that link above. One theory I had was that the extruder wasn't linearly extruding with movement of x and y. I thought this might be due to the way the firmware makes the moves - perhaps a technical limitation of the microcontroller or arduino layer. I haven't looked at the code to be able to say if this is even a possibility. But I thought it might help to explain why.

[Nylocke]

If you need a print that breaks up surface regularities maybe try the khosht snowflake vase single walled (spiral vase)?

[neurascenic]

one thing that might be able to be done to test the rounding error hypothesis without having to do all the math...

Have the firmware send back the X,Y values to a host terminal..

Can the Host software be modified easily to accept and display such data?

Or, is it possible to send basic gCode commands through the Arduino terminal, as that is able to display data returning from the board.

Sorry, still naive suggestions...

[626Pilot]

Now, I wonder what a delta segment is? If you are printing at 50mm/s does this mean that 50mm/s divided by 180 segments/s yields .278mm - that's actually pretty big. I think we are on to something.

I've wondered about this for over half a year. What we are probably seeing is either pulsing from the extruder, or a Bresenham line-type effect due to limits in positional accuracy:

Bresenham.png (4.68 KiB) Viewed 23953 times

The effector moves generally in the right direction, but it bounces back and forth across the true direction vector because the coordinate system (either the firmware, or the resolution limit on the steppers) doesn't have enough precision. The Arduino library silently casts doubles to singles, so "upgrading" the AVR version of Repetier to use double precision numbers might be impractical. If it's the steppers, upgrading to something with more steps/rev could help. The current ones are 1.8 degrees/step, I think. You can get them over 4x finer than that, if you want to spend the money on it. I'd try 1 degree/step motors (I think that's the next step up), inform Repetier of the change and see if it alters anything.

[enggmaug]

would adding a gear reductor beween stepper and belt improve things ?

A gain of precision by this way would mean a loss in speed, and some play in the transmission... but a good quality reductor could be cheaper than a more precise stepper.

[xnaron]

Now, I wonder what a delta segment is? If you are printing at 50mm/s does this mean that 50mm/s divided by 180 segments/s yields .278mm - that's actually pretty big. I think we are on to something.

I've wondered about this for over half a year. What we are probably seeing is either pulsing from the extruder, or a Bresenham line-type effect due to limits in positional accuracy:

Bresenham.png

The effector moves generally in the right direction, but it bounces back and forth across the true direction vector because the coordinate system (either the firmware, or the resolution limit on the steppers) doesn't have enough precision. The Arduino library silently casts doubles to singles, so "upgrading" the AVR version of Repetier to use double precision numbers might be impractical. If it's the steppers, upgrading to something with more steps/rev could help. The current ones are 1.8 degrees/step, I think. You can get them over 4x finer than that, if you want to spend the money on it. I'd try 1 degree/step motors (I think that's the next step up), inform Repetier of the change and see if it alters anything.

This is good explanation of what I suspected. The extruder will move from a to b but not in a linear fashion wrt the movement of x/y. I do have some 0.9 deg/step steppers I can try. I also would like to try running the delta on mach 3 with a g540 controller. My arm based controller should be here early next week so I can try that. The other thing I want to try is running everything on 24V.

[JolietDelta]

I had those problems with various types of printers. My Kossel, with Kevlar lines instead of timing belts, seemed to be immune to that effect. Well, not totally but the surfaces always looked less "wavy".
I did some experiments with the RMax and exchanged the idler bearings with some special parts with GT2 toothing (wich are actually some 608 KF with pressed-on gear rims).

Red: stock idler bearings; blue: idler bearings with gear rims

The photo doesn't show the difference very well. But on larger straight and vertical surfaces I have noticed a visible improvement.
As these bearings were spare parts I've "borrowed", I can't provide any sources right now. I also don't know if this method is the one and only solution. Maybe one clever mind has a better one. :-)

~Jens

[xnaron]

My Ramps-FD got here... but as bad luck would have it they neglected to ship the arduino due arm board with it. I ordered another one from a closer source and it should be here soon. I was looking at some older prints today and I noticed the vertical banding was less visible. I think these prints were when I was printing at significantly lower temperatures. I'll see if I can replicate.

One thing I had recently done is add the vibration dampers to the steppers. The plus size is the noise reduction. The downside is I think the print quality is suffering slightly. I don't think the layer alignment is as good as it was without them.

I see this vertical banding is common to all printers. It's like the extruder is pulsing and pushing out uneven amounts of plastic. This goes back to the whole non-linear bresenham thing mentioned before. If I can I'll try hooking everything (or at least the steppers) up to a g540 controller and try to run a print with mach 3 pc software. It will be interesting to see if there is any difference.

[Generic Default]

This problem is caused by the stepper motors. 626Pilot got it right with the image on his post. The only difference is that the blocks would have smoother edges.

#define PULLEY_DIAMETER 10
#define PULLEY_CIRCUMFERENCE (PULLEY_DIAMETER * 3.1415927)
#define STEPS_PER_ROTATION 200
#define MICRO_STEPS 16
#define AXIS_STEPS_PER_MM ((float)(MICRO_STEPS * STEPS_PER_ROTATION) / PULLEY_CIRCUMFERENCE)

That means that each full step on our machines is ~0.157mm. Micro-stepping is not linear, and it leaves a gap between each full step that the stepper motor has to 'jump' across. This jump is what results in the vertical banding you see on flat, face type surfaces. The jumps can happen at a resonant frequency with the platform inertia, which is why some of them seem much larger or smaller depending on the angle of the surface, the velocity of the print effector, the position between the towers, and the carriage mass and rigidity.
The magnitude of each wave across the surface is about 0.001 of an inch at 16x micro-stepping. Cartesian machines have a slight advantage when printing parts with geometries aligned along their X and Y axes, since the axis parallel to the surface doesn't step.

You can reduce this problem by 50% by switching to 400 step/rev steppers, such as these;
https://www.sparkfun.com/products/10846

Twice as many steps means half of the wave amplitude (and twice the frequency, so twice as many vertical waves in a given space). Increasing micro-steps beyond 16 will help a tiny bit, but not much.

I think that within a decade most 3d printers will be using brushless AC servomotors and this wave effect will be nearly eliminated. (This is for future reference so I can say I'm a prophet:)


Another printing artifact similar to this is the larger waves (~0.002 to ~0.005) on sharp corners. They're caused by end effector inertia which increases with higher print speeds and acceleration. This CAN be reduced greatly by software, and in the newest repetier host versions there are functions that reduce it a little.
The corner effect is called "ringing ripples" in audio and electrical terms. All 3d printers with a moving effector show it. In fact, my Rostock Max with 12 inch magentic ball arms shows less of this effect than a high-end Stratysis Dimension printer that I used to print some of my first prototypes.

Here is a picture of the ringing ripples effect;
[img]http://i.stack.imgur.com/WNGaS.jpg[/img]

[xnaron]

This problem is caused by the stepper motors. 626Pilot got it right with the image on his post. The only difference is that the blocks would have smoother edges.

#define PULLEY_DIAMETER 10
#define PULLEY_CIRCUMFERENCE (PULLEY_DIAMETER * 3.1415927)
#define STEPS_PER_ROTATION 200
#define MICRO_STEPS 16
#define AXIS_STEPS_PER_MM ((float)(MICRO_STEPS * STEPS_PER_ROTATION) / PULLEY_CIRCUMFERENCE)

That means that each full step on our machines is ~0.157mm. Micro-stepping is not linear, and it leaves a gap between each full step that the stepper motor has to 'jump' across. This jump is what results in the vertical banding you see on flat, face type surfaces.
The magnitude of each wave across the surface is about 0.001 of an inch at 16x micro-stepping. Cartesian machines have a slight advantage when printing parts with geometries aligned along their X and Y axes, since the axis parallel to the surface doesn't step.

You can reduce this problem by 50% by switching to 400 step/rev steppers, such as these;
https://www.sparkfun.com/products/10846

Twice as many steps means half of the wave amplitude. Increasing micro-steps beyond 16 will help a tiny bit, but not much.

I think that within a decade most 3d printers will be using brushless AC servomotors and this wave effect will be nearly eliminated.


Another printing artifact similar to this is the larger waves (~0.002 to ~0.005) on sharp corners. They're caused by end effector inertia which increases with higher print speeds and acceleration. This CAN be reduced greatly by software, and in the newest repetier host versions there are functions that reduce it a little.
The corner effect is called "ringing ripples" in audio and electrical terms. All 3d printers with a moving effector show it. In fact, my Rostock Max with 12 inch magentic ball arms shows less of this effect than a high-end Stratysis Dimension printer that I used to print some of my first prototypes.

Here is a picture of the ringing ripples effect;
[img]http://www.pstracks.com/wp-content/uplo ... fteen2.gif[/img]

Thanks this makes sense. I do have some 400step 0.9/deg steppers I can try. I just need to check if I have more than 2 of them. The one issue I remember when using them in the past was they had less torque than the 1.8/deg step ones I normally use. I think they'll probably be ok on the rmax. I will try them and see what the effect is.

I also have some 32 microstep drivers I could try as well. I'd need to ditch the rambo though and use a ramps.

[Generic Default]

They will have less torque at higher speeds because of inductance. At lower speeds (such as using one as a direct drive extruder) you will get about the same torque as with a 200 step/rev stepper. Keep in mind that our printers only have to deal with inertia while printing, so your maximum velocity shouldn't be affected by what type of stepper you use (within reason). If you start skipping steps;

1) Increase the current in the firmware
2) Put fans on your stepper controller
3) Put fans on your motors (if they burn you)
4) Decrease your acceleration and/or jerk

Xnaron you're the magnetic arm OG around here. I'm not sure what setting's you're using for velocity, acceleration, and jerk, but in the newest firmware I can use maximum acceleration and jerk at a velocity of 200 mm/s printing and I have never, not even once, lost a step. The ringing ripples are terrible at those speeds though, and layer adhesion decreases.

While on this subject I should also say that the plastic you're using greatly affects how visible the surface waves are. I've been printing with trimmer line nylon for almost a year and I just made my first ABS print yesterday. The nylon is much more glossy and shows print defects more than ABS. If you want aesthetically pleasing models it is critical that you use a tiny nozzle and do everything you can to make the surface finish better.