Idea: Belt-extruder

Hi everyone,

i've been thinking a lot about how to make the Mendel faster, i.e. getting weight off the axes. I know there are bowden extruders, but i also know they don't work perfectly for everyone and are prone to breaking. So i had another idea that involves moving the extruder motor off the x-carriage. Enter the belt-extruder.

Let me explain what you see above, starting with the known parts:

• orange: x-belt, Prusa style
• light grey: x-rods
• dark grey: x-carriage (not an actual representation of the final product)

Now these are new:

• light blue: extruder belt,
• green: some kind of bearing
• dark blue: extruder pulley

Basically, the motor driving the extruder belt sits on the opposite side of the x-motor and moves the belt along with the x-motor, but minus the extruder movement. The extruder pulley would pick up only the difference (= wanted extrusion) and then drive the actual extruder gears.

Now the obvious advantage is that the extruder motor is off the carriage while still keeping the lenght of tensioned filament short.
But: Driving the extruder this way requires an extra belt (=money, not much though) and firmware support, but that should be a one-liner, something like

- MoveExtruderMotor(deltaE);
+ MoveExtruderMotor(deltaE - deltaX);

I haven't exactly looked through the firmware but i guess it can't be much more complicated than this.

Please share your thoughts on this concept!

Thanks,
-imitation-

Would also require 3 extra bearings. I don't think a belt on small pulleys would give you anywhere near the precision or torque of the extruder gears.

The standard wade gears are 11 over 39 teeth, meaning a torque increase of 3,54:1. I'm assuming your gears will be near 1:1 meaning you'll loose a whole lot of torque.

But that assumption on torque probably has to be thrown out the window because you're going from a one-degree-of-freedom system to a two-degrees-of-freedom system. With a regular extruder gear train your extruder motor is responsible for one degree of freedom: push and pull filament. In your design that one motor will now have to participate in two degrees of freedom: follow the X-axis motor, and push and pull the filament.

Calculating this goes way beyond my mechanical skills because it gets complicated pretty quickly. If your filament moves in the same direction as your X-axis the X-axis motor will have to deliver all the torque required for moving both the X-axis and the filament (which will either require a much stronger X-axis motor or much less torque exerted on filament and X-axis individually.) In this case your extruder motor won't be able to help since any extruder belt movement will make the filament movement either faster or slower (depending on the direction in which the belt moves) while giving more torque. I don't see how your extruder motor can provide more torque without affecting filament velocity.

And this doesn't even take into account the ability of your extruder belt to do this kind of work.

Anyways, I'm out on a limb. Hope most of what I said is actually correct.

(Edit: I'm now assuming my naive explanation reveals a misunderstanding of how stepper motors actually work.)

(Edit 2: disregarding all of the above, a bowden cable is a much more efficient way to take the extruder cold part off the X-axis.)

(Edit 3: typos)

Edited 3 time(s). Last edit at 10/01/2011 04:20PM by devijvers.

I bent my mind a little more around how complicated this would really be to implement in firmware.
In a simple non-accelerating firmware this would be somewhat simple. Assumed that the same pulleys would be used on both x and e-motor, the movement of the x-axis could be compensated by simply substracting the steps, after calculating the steps required for the x and e movement.
Since i don't know how Sprinter and the likes really accelerate, i can't say how complicated this is going to get. If they split up accelerated moves in small linear moves, read above. If not, if have no idea.

Before i move on: Of course there should be some wade-like gearing in the carriage. More concerning that matter furter on.
Now on to torque. The way i understand it is that stepper torque decreases when speed increases.
Let's look at what happens when we do a linear move along x while extruding (simplest scenario). For the extruder to be, ahem, extruding, there needs to be a difference in belt speed = motor speed. So either E will turn slower than X, meaning X will have to generate the extra torque needed for extrusion and E will be generating "negative" torque. Or, when E turns faster than X, E will have to accelerate the carriage as well (shouldn't be much of a problem, since it's now way lighter) and X will be generating negative torque, slowing the carriage down.

Either way, now that i've written all this, i don't think torque as is will be a problem. The force required to accelerate the carriage should be quite small, and the torque for a wade gearing even with the carriage added should also not be a problem. Right now, i can extrude a lot faster than my axes can move without getting all wobbly while printing.
BUT: There will be problems when E moves faster than X, since, like i wrote further up, stepper torque will have reduced by quite a bit, and while i think the motors i'm running would handle it, it might cause some problems for others.

So here's another idea: Add more gears and have the belt run faster! So if we add a simple 1:2 transmission we'd cut the torque that gets transferred to the carriage (and therefore through the belt and against the x-motor) in half! Since i'm assuming that we're running a small 8 tooth pulley, we could either use a large (i'm thinking 30-40 teeth) pulley as a direct drive for the hobbed bolt or use a smaller pulley with 16 teeth and utilise the standard wade gearing afterwards.
And you can always up the initial gear ratio to, say 1:3 or 1:4 and make the pulley on the carriage a bit larger.

Concerning the extra bearings: You can buy 50pc 624 bearings on ebay for 21€ including shipping (and i don't mean those directly from china), so those shouldn't be a problem. The extra belt is a lot more expensive (~5-7€ + shipping) than the bearings.

Edited 1 time(s). Last edit at 10/02/2011 02:34AM by imitation.

Not commenting on the entire torque story - is still unclear to me - I don't see how cutting the motor torque by half is desirable.

How much does it help removing the mass from the x axis, when you are still limited by the mass of the Y axis?

@devijvers:
I guess my last post was a little confusing in regards to the whole "gearing" matter. My initial thought was to keep belt speed "synchronized", that is that the pulleys on both X and E Motors would be the same size and the needed E steps would simply be

Esteps = Esteps - Xsteps

Now with using a larger pulley on the E motor (or extra gearing before the pulley, but that seems overly complicated), the belt would transfer, say, only half the torque, but double the speed and would then be geared-down in the carriage again, either through a wade-like gearing or with a larger pulley driving the hobbed bolt directly. This means that the X-motor has to work against less torque generated by the E-motor and that the E-motor wouldn't have to turn as fast when pulling the belt in the same direction as X.
This obviously gets belt speed "desynchronized" and the necessary E-steps would be something like

Esteps = Esteps - ( Xsteps / GearDifference )

with GearDifference being the ratio of the pulley size on the E-motor to the size on the X-motor.

I like the idea of having the pulleys do the gearing. With a 16 tooth pulley on the E-motor you'd have halve the torque transmitted to x, and when using a 57 tooth pulley on the carriage to drive the hobbed bolt, you'd have almost the same gear ratio as a standard wade.


@Andrew:
I'm used to having a Sells Mendel, and i can confirm that the y axis is also quite heavy, especially when using a glass plate (=300g). The Prusa Mendel has a lighter bed afaik, but still heavier than the x-axis when using the glass plate atop the heater PCB.
When directly printing on the PCB, the y-axis get a lot lighter, with the main springiness still introduced by the bed springs. I think you could cure that by ditching the springs, adjusting the bed with screws that you can tighten with nuts on both the y-chassis and the bed. Of course, you'd then need some sort of (magnetic?) pop-out security mechanism on the hotend.

But let's say we keep Y like it is for now and just make X faster. If we could achieve this, then we could set acceleration (and possibly maximum speed) a lot higher for X, making printing faster at least in the X direction while maintaining acceleration and speed for Y. Say you could run X at twice the speed than before, that would speed up your print not by the factor 2, but by roughly 1.5.

@imitation:

I think that one of the best improvements that might come from your idea is the reduction in the size of the x-carriage. I'm finding that the size of the extruder is one of the largest factors limiting my build area.

My concern about it is the sideways pressure that it would be exerting on the carriage. Basically, now your x-carriage motor has to be able to overcome the torque applied by the extruder.