Symmetric belt-driven extruder to remove stepper motor from X carriage

Hi folks,

I want to remove the stepper motor from the X carriage for Prusa Mendel, to reduce weight and increase precision at higher speeds. I think the X motion and extruder can be controlled by two independent timing belts with a stepper motor on each side.



I have uploaded more pictures and a Sketchup model here:
[www.thingiverse.com]

The X carriage is based on VertX (vertical X axis) by watsdesign:
[www.reprap.org]
[www.thingiverse.com]

The original idea for a belt-driven extruder was posted by imitation:
[forums.reprap.org]
[www.thingiverse.com]

If both belts move in the same direction, the X-carriage moves in that direction too. If both belts move in opposite directions, the extruder moves but the X-carriage stands still.

x_motion = (stepper1 + stepper2) / 2
extruder = (stepper1 - stepper2) / 2

We'll need new firmware to control x_motion and extruder independently by setting the correct speed for each stepper motor for the two belts.

stepper1 = x_motion + extruder
stepper2 = x_motion - extruder

Do you think this could work? How would you improve it?

Thanks,
Johann

The issue that jumps to my mind is timing.
How are you going to keep both stepper motors in time with each other? And at a reasonable cost?
And what about belt slippage, balanced tension in the belts, and slippage in the motors?

I think you will find that this is one of those designs that looks good on paper, but is over complicated/expensive to implement,
but it would be cool if you get it to work reliably.

I don't want to sound negative, but I can't mentally picture how this will work. For some reason this design stuck in my brain and I keep thinking about it. Perhaps it is my lack of imagination. When you apply the differential movement, I don't know what law of physics says that the carriage will stay in the same place, but the gear will turn.

I think I realise what is troubling me, it looks like the same mechanism as the differential in car. However, in a car you actually want one side to slip when the forces do not balance. In this case the force will not be balanced, and there is nothing to constrain the carriage... so the carriage will move.

I also would like to suggest a better way to drive the extruder "remotely", I have been racking my brains, but I have not though of anything practical.

The belts hold the carriage in place. It will move at the average of the two belts speed, and the extruder will turn at the difference of the two belts speed. So far as I can see, the only problem with it is that it will probably add backlash to the X axis.

I think this design may actually reduce backlash because in normal operation both belts will continue to move always in the same direction. To drive the extruder and push filament into the hot end, they will move in the opposite direction from each other all the time. Either one of them will move slightly faster or slower to move the X carriage back and forth, but they may never need to reverse their direction unless we want to retract filament.

I assume that the X carriage would stop when retracting filament, so both belts would change direction at the same time, and their backlash would cancel out.

I guess the only way to know for sure if this works is to start building it, implement the necessary changes in the firmware, and then put a video on YouTube, so that's what I'm planning to do. I have also ordered some high-torque GT2 belts and pulleys (3mm pitch) from sdp-si.com to see if their special rounded tooth profile is really so much better than the common XL belts, and if belt backlash can be eliminated almost completely.

Am curious about this design too.

I suppose the 2 belts would need to be closed loops ?

Johann Wrote:
-------------------------------------------------------
> I think this design may actually reduce backlash
> because in normal operation both belts will
> continue to move always in the same direction. To
> drive the extruder and push filament into the hot
> end, they will move in the opposite direction from
> each other all the time. Either one of them will
> move slightly faster or slower to move the X
> carriage back and forth, but they may never need
> to reverse their direction unless we want to
> retract filament.
>

One belt will need to reverse during slow X fast E moves, as will happen when moving along Y axis.

Andrew Smith Wrote:
-------------------------------------------------------
> The belts hold the carriage in place. It will move
> at the average of the two belts speed, and the
> extruder will turn at the difference of the two
> belts speed.

Well, I still can't get my brain around how that will work with all the forces involved, but it would be great if it does!

One concern will be that whenever you're extruding, you're imposing a rotational force around the axis of the feed roller, which could easily cause our cheap linear rails to flex and/or bind. Even if it doesn't cause binding, it's likely to cause the nozzle to move horizontally, at least a tiny bit.

A vertical axis will be much more resilient to binding than a horizontal axis. Ive be toying with the exact same idea but never left my head onto paper (let alone a CAD model). Another way to combat this would be to slightly extend the x carriage. Since the stepper is no longer mounted on it the carriage can be made longer without loss of build area. The increased length will reduce the force since the torque is acting on a longer arm. (F=T/L, .: increase L to decrease F .: reducing the chance of binding.)

Surely the control is just one line that works out the difference in rotational velocities of the extruder and x axis steppers and outputs accordingly.

and this would also provide a more balanced X axis w.r.t the Z axis steppers as they will both be carrying a fixed load.

This is a neat design. I was doing a few calculations on weight saving which is your main goal here. The motor moves off the axis saving about 400 grams, but you do add 4 bearings and fasteners about 150grams. If you drive the axis directly with one of the belts and the extruder with the other, you eliminate 2 bearings on the carriage which is an improvement over your design.

Overall though this doesn't seem to offer any advantages over a bowden extruder other than it reduces extruder filament hysteresis. From my understanding, with care and setup of extruder filament retraction this problem can be virtually eliminated.

This indeed could work, but there is an issue that the X carriage would essentially be loose, and friction to another direction might be greater than another.

BUT, this could be used with just a single Nema 17 to drive both X-carriage and extruder, by using solenoids to either lockup the gear or by locking up another of the belts/rack.

Another of the belts (or can be replaced with rack) would not be attached to that gear at all, it's simply used to control if X-carriage is allowed to move, or is all force towards extruder.

The one braking the gear would allow deciding if we want to move extruder at all

Either one of them will need to be able to control the friction, but as that is unreliable, motion will need sensors, which makes too complicated (=expensive) hardware easily. Basicly the movement of extruder AND x-carriage would need to be measured, and autocorrected braking or motor forces applied to get the correct one (this can be autolearning so accuracy increases over time according to statistics).

Driven with 2 belts like this, i would expect some unreliable movement of extruder and X-carriage due to surface friction differences. How large? Only testing will tell.

Neat idea tho to drive like this!

you could use two belts, another driving x-carriage, another the extruder, with extruder having automatic tensioner (Ie. see motorcycles and engine timing belts) to keep belt tensioned even when x-carriage moves, this will however create the problem that the x-carriage wants to move towards another side and more force is needed to travel to another direction. Also, it would require quite long belt and lots of space.

In any case i would be worried of the bearing surface frictions with this kind of "loose" x-carriage arrangement because it is entirely possible a bearing (or gear even more likely) will have just slightly more friction towards another direction than the another, wear will smooth it out somewhat tho. It is possible this could work for the intended purpose with good bearings and gear.

Test, test, test! That will show.

No matter what anyone says, one needs to test it and see how it works

First off any concept that make you think about the mechanics, physics, firmware, ect like this is a good one and I think this concept is great.

I'll join the rest who think the key challenge is in the control of the steppers. Steppers aren't torque control but position control, sort of.

Isn't the trick to control belt tension when moving the X axis? Tension would represent the torque applied to the extruder drive bolt.

To test your ability to control tension/torque, why couldn't you mock up two steppers and a belt that has a pair of springs in the center. Could also try one spring. The goal would be to move the spring to any location while maintaining some preset length of the spring induced by the initial belt tension.

Until you are able to maintain a constant force on that spring over all locations I'd say you will either experiance poor position or extruder control.

Above all don't give up! Get the solution and let others worry about how to make it cheaper.

on second thought... the spring wouldn't represent the extruder very well. It would simply react to the tension on the belt and remain constant over the range of positions. In the end you will probably have to test it with a real extruder. Just thought that I'd offer something to simplify the proof of the concept.

Edited 1 time(s). Last edit at 01/22/2012 09:36AM by dean448.

This could possibly work better with the help of an encoder.

Either an encoder on the extruder, or a linear encoder on the X axis. (It would be best to use an linear encoder on the X axis.)

This way, the controller controller will know exactly where one of them are and can then calculate exactly where the other is.