Ghosting Defect

[ednisley]

18000 mm/min with low acceleration

The platform probably never reached top speed, because the moves were too short.

At 900 mm/s², the platform requires 50 mm to reach 18000 mm/min = 300 mm/s along a straight segment. However, it must also decelerate at the end of that segment, so segments shorter than 100 mm won't allow full speed; it hits full speed only in the short section between the 50 mm at each end.

So, for a 20 mm calibration cube, the maximum speed in the middle of each side will be only 134 mm/s = 8000 mm/min and the average will be much lower. A more complex shape with no long straight lines (like, say, the Alien monster) will run at a small fraction of the maximum possible speed.

That's why support structures cause so much trouble: the machine can approach maximum speed only when printing those long straight segments.

Basically, when you set the acceleration low enough to remove corner ripples, the maximum speed setting won't have much effect: the machine can't get there from a standing start.

For a constant top speed, the acceleration and distance have an inverse relationship: ten times more acceleration reduces the distance by a factor of 10. At a constant acceleration, however, reducing the top speed by a factor of 10 reduces the distance by only a factor of √10 = 3.2, which explains why tinkering with acceleration has much more effect.

Because the force applied to the machinery is proportional to the acceleration, you get a bigger payoff from limiting acceleration, not velocity.

Some path planners allow speed blending around corners by trading positional accuracy for smooth transitions, but that's a whole 'nother topic... [grin]

[pyronaught]

No zipping involved: the nozzle traces a continuous line from the inside of the object, around the edge, then along the outside of the object (or the other way around). Feed the G-Code into gcode.ws (or whatever toolpath previewer you prefer) and look at the layers to see how the nozzle moves.

If the object has only one opening on a layer, you'll see that there's a single thread going around the entire object, with no rapid (non-extruding) motion at all. You'll still see ripples on the down-stream sides of that opening, though.

Those particular ripples are due to the platform vibrating very slightly as it stops moving after the nozzle deposits the edge of the hole (parallel to the Y axis) and begins moving along the X axis. The M2's platform rests on three springs compressed by the alignment screws, so it has a tendency to shake very slightly after high-acceleration stops.

I added silicone plugs inside the springs to damp that motion:

http://softsolder.com/2013/05/31/makerg ... ilization/

That worked reasonably well, but there's still plenty of mass lashing around.

As other folks suggest, you can reduce the Y axis acceleration by a factor of two or four, which reduces both the force and the ripples. You can also reduce the speed on perimeter threads to get much the same effect: less speed requires less force to speed up and slow down.

That makes sense, if it was the whole machine vibrating then the extruder would move in unison with everything else and would not cause this problem. I think the nature of the build platform design makes it prone to vibration even with stiffer springs. With the platform being supported only on one side and the whole thing hanging from the relatively narrow mounting plate that the spider attaches to, it is just not rigid enough to avoid vibrating. If you push on one corner you can see a considerable amount of flex before the springs ever start to compress.

McMaster does sell these wave springs so it might be possible to also buy stiffer versions of the same spring as an alternative to making the silicon plugs:
http://www.mcmaster.com/#compression-springs/=vepi63

If I ever scratch build my own printer I would support the platform on both sides the same way the M2 has on only the left side, then use linear rails on both sides of the platform instead of one in the middle like how most CNC milling machines are built. I realize the M2 is trying to keep costs down so more people can afford to own one and those changes would really impact the selling price, but if building your own you might not have that same design constraint.

[jsc]

Then you run into the problem of over constraining. Printers that attempt that often are prone to binding if things are not absolutely aligned (and they never are).

Edit: summary

[pyronaught]

Then you run into the problem of over constraining. Printers that attempt that often are prone to binding if things are not absolutely aligned (and they never are).

Edit: summary

I think "racking" is the problem they are trying to describe, which is easily preventable with the proper design & assembly. For example, if you were going to put a pair of guide rods on both sides of the Z axis, you would also want to put a stepper with the ACME screw on both sides to prevent racking (I see the Rambo board already has the extra Z driver for doing this). If you had a single pair of rod type rails for the Y axis, you would put the ACME screw at the exact center to prevent racking. The M2 is already doing this on the Z axis, and there are thousands of home made CNC machines out there using that same kind of arrangement. Bearing selection is another factor-- some handle side forces better than others. Those cheap self lubricating type slides will bind very easily from forces that aren't parallel to the line of motion, whereas the more expensive kind that employ ball bearings are more tolerant. Binding is only a symptom of improper design or assembly, not a limitation inherent in constraining unwanted play or backlash.

[pyronaught]

Finally got around to printing the part in the picture again and realized the extruder is actually coming at the holes from the right side and then reversing direction once the hole is reached. So it doesn't actually cross over to the other side, but rather bounces off the edge and returns the way it came. So the vibration causing the ghosting must be from the jerk created on the sudden path reversal. I've got acceleration lowered to 800 and doing outer perimeters at 50% speed, but still getting the same problem.

[jimc]

pyro, your not going to get rid of it all together. you will get it to a somewhat reasonable level then thats as good as it gets. even if your print speed and acceleration is low there will be a slight ripple.

[pyronaught]

It actually looks worse in the picture than in real life. I tried to highlight the defect as much as possible with the light reflecting off of it. It still looks exactly the same after the changes I made though.