I should also clarify this:
There are at least two ways to run a motorized rewinder.
For my setup, I run the rewinder fast enough so that it always “outruns” the ERCF gear motor during rewinding. There is then slip happening on the roller and the spool. This approach puts stress on the rewinder motor’s speed and torque – I had to do calculations carefully to choose the right motor and roller diameter, etc.
There is another way to run the rewinder, namely, by disengaging the ERCF gear completely and let the rewinder do all the work of pulling.
This method works best if you have a gate sensor (which ERCF does not have in a stock build), or you must rely on the encoder to detect how much the filament has been rewinded — might not be reliable and accurate if you encoder is not reliable enough. In both cases, you’d need more advanced Python (or even klipper firmware) code to handle the filament movement.
The upside is that you don’t have the problem of “varying filament diameter on spool”.
But, the downside is that you get a lot slower rewinding when the spool is almost empty.
If you go with the bambu AMS gearing, the loading/unloadign speed might turn out to be slower compared to stock ERCF (which is >200mm/s easily), as I’ve seen from various AMS videos. Therefore, it might be that you have to end up with the second approach, or the first approach if you reduce the ERCF gear unloading speed to something very slow.
One thing I do like about the AMS first stage feeder design is that it can actually feed filament. With this, you can move to a Type-B MMU system (Happy-Hare/doc/conceptual/typeB_mmu.png at main · moggieuk/Happy-Hare · GitHub) and ditch the ERCF altogether! This is what really interests me about motorized rewinders. If you can get the first stage feeder to work, I’d be happy to contribute on coding support, and I might build one myself 
ok now it gets ambitious.
MMU’s like the ERCF are definitly more complex, complicated, and a mess to calibrate.
but the ERCF works.
Ditching the ERCF means reverse engineering the bambu AMS completely. Definitly possible, but it takes a lot of time. more sensors. completely new software.
and there are possibly legal problems.
i would stick to the box for the ERCF at first, thats a good basis, improvements and extensions can be done later on. and i think many users of the ERCF would like an addon without getting rid of their MMU
First test was a success. Rubber Coating works so far. Needed a few iterations to get the rollers fit right to the axle. Holders will get some clips, at the moment i can pull out the rollers too easy.
Well, guys, i have a working prototype!
Took me countless tries, development of several tools…you really need to be precise to be able to press the gears onto the shaft…
It is grippy, has power, and it just slips under load.
I’m gonna build the next 8, then i’ll do a testrun with manual switching.
May some tinkering with the voltage is enough to get it move as fast as i need
Edit: Seems like some chinese guy built a electric rewinder based on the same gear mechanism…just with more volume: 3Dprinter-Electric-rewind—filament-rewind—multi-color-printing/封面.jpg at main · muzixiaoyang/3Dprinter-Electric-rewind—filament-rewind—multi-color-printing · GitHub
just found it cause he made an issue in happyhare on github. Idk about his chinese stuff, but there’s a .cfg file inside, i’m curious about his strategy
1 Like
Well, mechanics are done! After trying to make the prototype 9 times i figured out i get a precision problem when i’m just mounting it into the base plate. So i redesigned it to be modular, fully enclosed, with a plug, and added some bearings…mounted it in a milled polycarbonat sheet.
Electronics are already on my desk, an rp2040, some transistors and some already tested Dual H-Bridges. Anyway, university and work are sucking my time till summer, although i would like to finish it soon, i have no choice than giving it a break.
