Belt drive extruder vs "belt idler" extruder

[Ryan Carlyle]

This is interesting. "Vaeder extruder" by Martijn Korevaar.

https://plus.google.com/110183027524080326690/posts/4xWC7GCoWUu

http://reprapworld.com/newletters/newsletter_12.pdf (on end of p1 and most of p2)

He's driving the filament with belt friction. What's different here (as opposed to previous belt-drive examples like this Ultimaker extruder) is the large-diameter wheel that the filament wraps around as an idler. That provides much larger contact area and more even pressure than a "tank tread" style belt drive extruder. So that's good, as long as your filament isn't particularly brittle. (Old 3mm PLA probably won't work so well.) 

But, I think the real potential in this isn't using belt friction as the drive mechanism. I think the big wheel should be hobbed+driven and the belt used as an idler. That would provide an enormous amount of contact area for the teeth to grip. You would then use widely-spaced and sharp teeth to get the necessary tooth contact pressure for adequate bite. 

Regular bearing idler designs or dual-hob designs only apply contact pressure to the filament over maybe one square millimeter, probably less. No matter how good the hob grip is, there's a finite amount of shear strength in the plastic in that small area. A belt-idler approach really doesn't need a lot of contact arc to beat that -- even, say, 20-30 degrees of contact on the hob is vastly more area than a single 608 can provide.

The design challenge here is the tradeoff between belt tension and hob diameter. The radial force applied by a belt wrapped around a circle is inversely proportional to the arc radius and proportional to the belt tension. So we want a small hob to maximize contact force due to belt tension... but that means the filament is more likely to break going around the turn. A very large hob would need very high belt tension to apply adequate contact pressure. (And it would need enormous motor torque.) I think there's probably a compromise hob diameter that will work well, but it would take some experimentation to find.