People have used motion belts or shafts to "carry" the extruder drive force before, in order to utilize direct-drive extruders without putting stepper mass on the carriage. What's amazing is how much simultaneous work is occurring in this space, all apparently independent. A few examples:
A serious problem with the CoreXE design is racking. First off, the "unknown plotter" design of parallel Cartesian gantry (also called Dualwire Gantry) has the same racking problem as HBot. Ok, fine, you can work around that with good mechanical design and make it stiff enough to resist racking. Or you can invert the belt path on one side to make a non-racking parallel Cartesian variant. See:
https://groups.google.com/forum/#!topic/3dp-ideas/vWlaCGTHKhg
The problem is, when you multiplex the extruder drive over one the drive belt like this, the extrusion force has to be provided by differential belt tension. One belt leg going to the extruder gearing must have substantially higher tension than the other. If your drive belt takes the path used in CoreXE, it will put a large racking force on the gantry. Now the Y axis mechanism must be significantly stronger to resist that racking force. Torsion rod twist and belt stretch will allow quite a bit of racking motion and thus cause Y axis error proportional to the extrusion rate.
Back in April, Scott Booker gave me the same idea for a belt-driven extruder and I came up with a different implementation. Mine is much more similar to CoreXY, has non-racking belt forces, and allows dualstrusion (unlike CoreXE). Since the extruder axes are usually called A and B, I'm calling it CoreXYAB. Basically each CoreXY belt is doubled back in a loop and each independent loop controls one stepper. Here's the early concept drawing showing one equivalent CoreXY axis:
There is no racking force because each belt loop is routed on a fully parallel path, such that the increase in tension on one leg is exactly offset by the decrease in tension on the other. The other CoreXY belt route can be a single belt (at twice the tension) for single extrusion, or another looped belt for dualstrusion.
Yes, printer sizing will be a challenge on account of the need for a continuous loop belt. Basically the belt must be 4*(x travel + y travel) long. A 2218mm long endless loop belt from SDP/SI would allow a >200x200mm build area in a 300mm frame, which is a good size for a desktop printer.
Then there is the control system. The CoreXE concept and my initial CoreXYAB concept both require new firmware to multiplex two motors into a shared motion + drive system.
- CoreXE drives the gantry by moving one stepper and drives the extruder by turning both.
- CoreXYAB as shown above drives the gantry by moving both steppers in the same direction and drives the extruder by moving both steppers in opposite directions.
Same principle, different implementation.
But this really kind of sucks. Multiplexing two motors via firmware is a bad implementation. Why?
- You need custom firmware! Forking firmware into a custom variant is bad.
- If the extruder skips, the gantry will lose position registration. Which means your extrusion volume must be spot-on for the entire print, and it's very risky to use excess plastic squash on the first layer for better adhesion.
A superior option is multiplexing the motion and drive via hardware gearing. That way, each motor only does one thing, and gearing combines the motor output into a multiplexed belt motion. I think this is really clever -- you can use "phase-shift" (or "summing") planetary gearboxes to combine two motor inputs into a shared output. One motor drives the ring gear, one motor drives the sun gear, and you take the combined output off the planet carrier. Here's the concept drawing:
This would be a big gearbox using a mix of printed herringbone gears and probably some large bearings to support the double-sided ring gears. Not simple, but it's an elegant solution to the control problem. The double-shafted drive stepper moves the belt sides in the same direction, giving carriage motion. The extrude stepper moves the belt sides in opposite directions, giving extruder drive. Stock firmware can then be used simply by adjusting the steps/mm values according to the planetary gear ratio.
Plus, since the steppers are actuating separate motion modes, it will not lose position registration when the extruder stepper stalls. The drive stepper can maintain position even if the extruder jams.
The final challenge is converting the belt path to a horizontal-axis rotation to turn the drive gear. With non-twisted belts, the carriage pulley will spin on a vertical axis. The CoreXE team has designed a small right-angle gearbox to convert pulley motion to drive gear motion. Alternatively, you could use spectra line in a non-planar path and turn the carriage pulley sideways to drive the extruder directly. That's probably what I'll do to start... but I have another extruder idea that is the subject for a separate thread.
