While changing to black filament, I measured the force required to pull the (natural PLA) filament through the translucent guide tube arching over the M2’s chassis from the spool to the extruder:
A strike-anywhere kitchen match (bet you can’t buy those any more!) provided more than enough heat to bend the end of the filament into a loop suitable for the pull scale:
The results:
- Tube reasonably straight: 0.5 lb = 2.2 N
- Tube arched to middle of X axis: 1 lb = 4.5 N
- Tube sharply bent to X axis nearest spool: 1.5 lb = 6.7 N
The force increases slightly while tugging filament off the spool, as the spool does not rotate freely on the printed arm jutting out from the frame, but those numbers are in the right ballpark.
The effective diameter of the extruder drive gear is about 11.5 mm, so overcoming the tube friction requires somewhere between 10 and 40 mN·m of torque. That’s applied at the one point in the whole system most likely to show the result of uneven loading, because it directly affects the pressure of the molten plastic behind the nozzle.
That’s considerable motivation to get rid of the filament guide tube…
The Smell of Molten Projects in the Morning 
It would serve a better purpose if the filament was pushed through the tube by a stationary feeding mechanism, rather than pulled by a one moving with the extruder head.
That’s what I was thinking…
Extruders seem to work best with the main drive gear directly at the hot end, in order to get rapid retraction. A much smaller drive that unspools the filament and maintains an arch over the extruder, without embossing the filament too severely, should be do-able. That hardware could signal end-of-filament, detect snags, and maybe even measure filament diameter.
All that will be easier to tinker with under LinuxCNC, because we’ll have more I/O and control.
Was giving some thought to dual drive, and the issue I’d worry about is keeping the filament from twisting. The filament guide tube would help with this, but you’d need something to keep acceleration mismatches (between the supply and the extruder) from causing more trouble. Perhaps a springy termination at the hot end? Sounds like re-inventing a bike’s brake system.
I’ve been working with barbed wire, and its ability to tangle in awkward ways focuses your attention.
FWIW, any thoughts about moving the spool to the centerline of the of the printer? It would make the tugging forces symmetrical at the two extremes, and could be tried as an interim solution before trying the dual drive. Painful for footprint if mounted at the back, unless you mounted the spool above the printer…
That turns out to be no problem at all. The spool rotates to dispense the filament, so it doesn’t impart a twist: the filament starts from its natural orientation on the spool and maintains that curve all the way to the extruder.
That said, the lowest-entropy state of spooled filament is a hairball exactly filling the entire room surrounding it…
The simplest solution seems to be a loop of filament, just like in those old tape drives with vacuum columns.
The Y axis travel would put the spool way out beyond the existing structure, which would unbalance the chassis pretty badly, and it’d need a tower to get the spool high enough above the chassis for an overhead mount to work. The M2’s arrangement on the end of the X asis travel is just about as good as it gets for a self-contained unit.