Saturday, April 18, 2009


Bowden extruder concept

Anyone who's designed a 3D printer knows that one of the primary constraints is the geometry of the extruder. If the extruder is mobile then it has a big impact on the final volume of the ‘bot. I've always been a fan of making the final printer as small as possible (cup cake’s footprint’s pretty sexy), so to that end I have been looking at options for tiny extruders.

I think the key here is to decouple the nozzle from the drive mechanism – it is the motor and its transmission which takes up a big proportion of the extruder's volume. Using some PVC sleeve (3.3mm ID) I decided to follow Forrest’s previous attempts at using a bowden cable approach to make a flexible link between the two. The motor is anchored to the side of the bot and drives the 3mm PLA filament (A) through one end of the tube (B) down the flexible tube (C) which terminates in the hot nozzle (D).

Decoupling massively reduces the necessary volume of carriage which in turn enables a much smaller machine. It also takes a lot of weight out of the moving parts.

Two trade-offs are:

- Increased power will be needed from the extruder motor to overcome the friction in the system (though we’re only running our NEMA 17 pinch-wheel extruder motor at 30%, so we have some capacity there. Also, we could use Teflon sleeve to reduce the friction co-efficient as a last resort).
- Stretching hysteresis in the sleeve may affect print results (this is down to finding a sleeve stiff enough but still suitably flexible for print movements).

Unfortunately I’ve knackered the electronics on my machine so I can’t test this yet, but finger tests feel good. Thankfully Adrian’s offered to fix me up with some of Zach’s latest boards while I’m away for Easter so watch this space.

Love the concept! Where did you get the PVC sleeve? Hopefully we can find something feasible for this that won't be another special order part. How is the play in the cable? For example does pushing 1mm from rest yield exactly 1mm out the other end? I'd be interested in knowing the response time as well. If there is a delay between then the pinch-wheel extruder turns and then the filament finally starts getting pushed through the heater it may effect print accuracy. Either way, I like the concept and I could definitely see something like this working.
Damn, eD! It would be great if you could get that idea working. I'm holding thumbs hoping it works for you.
If we put a pinch wheel and a shaft encoder at the business end of the cable then we could compensate for any inaccuracy.

A bit less elegant, but still a lot smaller and lighter than a stepper motor.
In fact, if the first pinch wheel puts a pattern on the filament we could measure its motion at the other end of the cable with an optical mouse sensor.
On the other hand we could just design an extruder that didn't weigh very much. :-)
I just tried some polyethylene icemaker tubing (4mm I.D.). Very flexible and easy to push filament through but, over a 60cm stretch, there's about 6mm of play.
Something similar to bicycle shifter housing should work very well with very little stretch or compression. Modern index shifting housing uses wires parallel to the length of the housing to prevent stretch and is generally teflon coated. The problem is I don't know of any large enough in diameter to pass 3mm filament. Perhaps housing from a motorcycle?
Other possible sources: throttle cables for go-karts, lawn mowers and garden tillers. The two I have here both have 1.5mm cables but they're on fairly lightweight machines.
While this may increase the print area of a Darwin, it seems to me that it's straying from the long-term goals of RepRap: to produce a self-replicating printer.

The conventional design has the hardware of the print head mounted on a printed piece. This design adds a long stretch of additional vitamins without substantially increasing the self-replicatability (is that a word?) of the system.
Bob: Actually, I think this design in fact is a bit of an improvement in the ability of the machine to print it's own parts. Simply because with a much smaller extruder, you can start to stack multiple extruder heater-heads in parallel easier(since the current limiter is the weight of all those motors and electronics being pushed around by the xy axis motors). They'll be running off of driver motor(s) that are situated off from the print area and taking the weight away. This means you can run the extruders in parallel, making the whole thing faster for making duplicate parts. Which will be a great thing cos alot of the pieces are simply one piece, copied 4 or 8 times.

Along that vein, you could have one motor driving plastic for multiple extruders, a definate possibility given the new drive designs seem to have plenty of spare torque thanks to their stepper motors.
I suggested something similar to this in the comments back when Zach first blogged the pinchwheel prototype - At the time, I was thinking in terms of aquarium tubing, and I still don't know much about the different cables that exist, but something I came up with as a possibility was McMaster part no. 5733K54, a hard, clear (so you can see the filament) tubing with a 4mm ID, minimum of 1" bend radius, rated to 500 degrees F (260 degrees C). There's also a 3mm ID of the same at McMaster, IIRC.
You might not need an optical mouse sensor: those have to be 2-d, while this is a 1-d problem. The sensor off of one axis of a ball-driven mouse might work, with appropriate illumination.
If the force on the filament (i.e. the nozzle pressure) inside the cable changes significantly with head position, then you could just map out that force and modify the extruder rpm appropriately.

This extra force on the filament, due to change in length (smaller radius of curvature of bowden cable = retraction of filament back into cable), would be modelled by a scalar field, where the xy coordinates of the head are transformed into the length of filament retraction.
Knowing the head's velocity you would then get a value in mm/s which must be added/subrated from the extrusion rate.

The elasticity of the bowden cable would have to taken into account, and would be proportional to head acceleration (which may be constant which would be cool). Or use a v-stiff bowden cable.
For the encoder, you just need a roller from an old mouse or something. Using a physical roller instead of an optical one seems more reliable and compatible.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]