Idea for a lighter extruder

Hello,

I recently got interested in the idea of making a 3D printer and I'm becoming more
interested in the ideas and philosophy of RepRap. I started to try and think of ways
to reduce the amount of mass that needs to be moved in the extruder and looked
at various ideas such as the Bowden variant. Whilst this design removes mobile mass
it also introduces problems due to hysteresis. But removing the stepper motor is a very
good idea as these are heavy.

I thought of three ideas for removing the motor that I haven't yet seen around - but maybe
thats because they don't work - or I just haven't seen them yet - I don't know.

The first idea was to move the motor to a stationary part of the machine but keep the large
gear that drives the grip wheel mounted on the extruder and driven by a smaller gear. The
small gear would be driven on a long shaft but be free to slide along it. This may however
add unwanted friction.

The second idea was similar to the first except that the small gear is driven from a flexible
shaft ie, one that bends readily but maintains tortion - or a segmented shaft coupled with
universal joints. These designs may suffer from "flapping" when the extruder is accelerated
rapidly and a flexible shaft may introduce hysteresis.

The third idea which I think has the best chance is to drive the large gear with a toothed
belt in the manner of a rack and pinion. The belt would form the "rack" and be held against
the gear with another wheel on the extruder. The belt would be driven from the motor mounted
at the end of the axis of travel and go around a pulley at the other end. This motor would
need to take account of the movement of the extruder assembly. The pumping action of the
extruder would come from the difference in movement between the assembly and the belt -
the difference being precisely controlled in software. I can't see that this add's any friction
or hysteresis problems and moves the differential problem trivially into software. It could
also be implemented easily using no extra "vitamins", as I've seen them called.

[forums.reprap.org]

There are 2 moving assemblies: [xcarriage+extruder] and [ycarriage+heated bed]. Or better said assembly of all that moves on each axis, bearings included, hot end, etc.

As far as i can figure it, only the highest mass of the two does impact in speed and acceleration machine could perform overall (given that both axis have same motor type, which they probably should).

If you have a 2-3kg heated bed assembly and like 0.8-1.2kg extruder moving assembly, then what would be the advantage in trying to make the extruder ligher.

Probably to figure this exactly, one needs to estimate or to actually weight the 2 moving assemblies. That could tell you exactly which one needs to be lighter, and can give you an ideea with how much.

I dont think it can be generalized, because on Y some use only pla bushings and a slim board, or others use heated bed with a heavy tile on top of it, which ofc its very different. And from picture one can only form impression, not facts. But i think in most cases the biggest weight would be on Y axis. At very least that should not be omitted from the general picture.

Edited 2 time(s). Last edit at 10/07/2011 01:57PM by NoobMan.

Does the Y carriage have to be in the form of a moving bed? I was thinking more in terms
of using some sort of rubberized fabric conveyor belt. That would have far less inertia and more
utility than a moving bed wouldn't it? You could then print objects longer than the actual printing
area, i.e chains / toothed belts etc that hang off the end of the machine when they are finished.

The main goal for me is to come up with ideas to reduce inertia of every moving part and be able
to have a faster machine.

I was wondering, is it possible to calibrate the pump speed of the extruder against the travel of
the tip to gage the thickness of the extrusion? E.g speed up the travel of the tip to stretch
the extruded plastic to make it thinner and thus achieve a higher resolution?

The build platform has to be rigid enough to resist part warping (a lot of force)

Ultimaker is about as fast of a machine as is possible with FDM. You should start there.




Beetlejoose Wrote:
-------------------------------------------------------
> Does the Y carriage have to be in the form of a
> moving bed? I was thinking more in terms
> of using some sort of rubberized fabric conveyor
> belt. That would have far less inertia and more
> utility than a moving bed wouldn't it? You could
> then print objects longer than the actual
> printing
> area, i.e chains / toothed belts etc that hang off
> the end of the machine when they are finished.
>
> The main goal for me is to come up with ideas to
> reduce inertia of every moving part and be able
> to have a faster machine.
>
> I was wondering, is it possible to calibrate the
> pump speed of the extruder against the travel of
> the tip to gage the thickness of the extrusion?
> E.g speed up the travel of the tip to stretch
> the extruded plastic to make it thinner and thus
> achieve a higher resolution?

---

www.Fablicator.com

Quote

If you have a 2-3kg heated bed assembly and like 0.8-1.2kg extruder moving assembly, then what would be the advantage in trying to make the extruder ligher.

Two advantages:

- The total inaccuracy is the inaccuary of both parts, added together. So, if you improve one, you improve the whole.

- Inaccuracy due to acceleration not only depends on acceleration steepness, but much more on the stiffness of the carriage. As the extruder sits on a much weaker assembly than the bed, saving weight on the extruder should be the primary goal.

If all were just about getting up to speed as fast as possible, installing bigger motors would be a simple thing.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

I have a slight different point of view about that. If we take a portal machine / moving gantry machine were X and Y are "packed together", i believe that indeed the error will "add up".

We have X and Y independent, which means positional errors are also independent around same targeted point. Which means the result is not a simple "addition". The errors will "compose" (dunno if its proper english word but maybe represents the ideea). E.g. they compose and have a very good chance to partially overwrite each other out. We consider a point A that is targetted by both axes, then a small grade circle (like a temp graph) would be the Xf(a) probabilities for that target, and for the Yf(a) will be same thing / looks. But coz the axes X and Y are independent, then the circles or ellipses will partly over-write each other - more or less.

The stuff is in relation to ~ mass * (velocity^2), but the top speed for a diagonal move will practically be same speed for both axes, so we cant care for the speed. The differences can be made with the factor mass / weight. Then lets say we have a bigger mass on a axis and a smaller one on the other. A bigger probability "circle" or "elipsoid" for one axis and a smaller one for the other axis, then the "cancelling" aspect changes more. Basically if only one axis has a bigger "circle" like that, than the other axis impact becomes less and less, e.g. it ends up being neglectable. The chances to add an error to the result become at the limit of their probability distribution extremes. Sort of speaking, when the second circle is smaller, it will fit "easier" inside the bigger one, so will count less and less.

I think this is the very reason why separated X and Y axis are recommended for "beginners" or when materials are "lesser" ones, because its basically easier to achieve a better positional accuracy in systems with independent X and Y. For a gantry / portal system, all the errors will add up, because are all packed together, even Z will add up, so these machines will need good and precise mechanics components.

So i think that in our case, for Mendel / Prusa the X errors adds up directly with Z, because we pack these two together, but wont add up directly with Y because that one is independent from the rest.

If we take machines that pack X and Y together, upside like dunno, darwin / rapman or many others, or lowside like cupcake, those will add their errors directly. One axis registered error becomes the reference point of the next circle / shape probability of the other axis.

I hope i am not too far off with this.

Edited 3 time(s). Last edit at 10/09/2011 11:54AM by NoobMan.

I looked at the Ultimaker - that is a fast machine. There are some good ideas there
to start with. A fixed build bed, with both X and Y axis' moving but the motors are
stationary. That is similar to another idea I had using a single belt driving two axis'
and stationary motors.

Now what I'm thinking, hypothetically if you could have a machine with no inertia in
its moving parts (I know thats impossible - but just suppose), then what would the other
limitations be on speed and accuracy? I'm not considering control as I have drivers that
divide a full step into 64 so effectively I get 12800 steps/rev which I think would give a
higher resolution than is practically possible in the machine mechanism.

What are the physical limitations to depositing a stream of plastic onto a surface? How
thin can you make the stream and how fast can you drag it around before things start
getting messy? Can you use differences in extrusion speed to printing speed to vary
the thickness of the plastic deposited - i'e stretch the extrusion to make it thinner? I
would like to acheive something that looks like spider web in consistency and from
experience I know that you can stretch hot plastic to make it very fine.

Edited 1 time(s). Last edit at 10/09/2011 06:08PM by Beetlejoose.

When you talk about errors adding up, are you referring to step loss in the motors?

I haven't looked into the control systems of the various repraps, but I've heard a few
of them and the motors often sound like they are being used in full or maybe half step
mode. Under those conditions if you are pushing the limits of acceleration and top speed
then you are likely to lose steps.

Quote

We have X and Y independent, which means positional errors are also independent around same targeted point.

Where would this targeted point be? I think it's fair to assume this target is the part's surface and this part moves with the Y carriage.

Sure, if you're lucky, X and Y bend in the same direction and both errors eliminate each other. It isn't exactly trivial to achieve this situation for all movements, however, if at all possible.

Quote

When you talk about errors adding up, are you referring to step loss in the motors?

I think we're talking about lost precision due to the apparatus bending. Forces make even the stiffest assembly bend, and instant speed ( = infinite acceleration) even more. In this light it isn't wise to put two axes on top of the third, as you even experience bending on the two axes when only this third axis is moving.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

Beetlejoose Wrote:
[...]
> Now what I'm thinking, hypothetically if you could
> have a machine with no inertia in
> its moving parts (I know thats impossible - but
> just suppose), then what would the other
> limitations be on speed and accuracy? [...]

If there is no inertia, then nothing opposes changes to the system's state. Then the object could accelerate in space from 0 to speed of light with no cost, and in 1 microsecond. Also, the conservation of energy law would probably be bent, meaning that energy would either go away or be created out of nowehere. Usually the stuff is like it is because there is always something that constricts it to that place, state, level, etc. Take the temperature as a variable from our environment and we couldnt breath oxygen because it couldnt be a gas, since temp would not restrict it to gas state anymore. In a similar way, we take out inertia then there is nothing to restrict the object to 0 speed, or to any speed in general which is weird to suppose what would happen. The object could decide by itself it can accelerate at c speed or other speed at any time. It could exist at all speeds in between, or would stop to exist at all as we know it. Quantum effect at macroscale. We then might have a very different perception of what speed is, or have no perception at all. How would we perceive changes in pressure if the temperature itself is inexistant as a measure - we probably couldnt have density either. I cant imagine things without inertia, it would suppose too much redesign of our world. At least way more than i'm capable of even imagining. For starters about missing inertia and speed, accelerating to a speed wont cost anything, but its first subtle effect means that the kinetic energy gets created at 0 cost. Further than that just too many things get involved.

If we have a 1kg mass with a 10mm/s speed and we have to stop it, we gotta deal with inertia, i dont know of any way to avoid it. The inertia will be inevitably the same as value, cant escape, cant dodge it, cant affect its amount. Its "fatality".

I think there are quite a number of times when we get generic stuff that we dont like, but we cant do nothing about. What we can try, is to try control it, redirect it, dissipate it, dilute it. For example we dont stop the object in 1mm space, we do it in a 10cm window. (e.g. "ramp deceleration"). Or we take part of it, put it in a spring to release it later, or even further we have a device controlling the rate of spring release (e.g. car shock absorbers). For shear forces within 2 objects we gotta use lots of screws to assure overtransport from one piece to the other, (e.g. the industrial pipes 40cm dia with 80 bolts of M30 all around, like 1 each to smallest possible inter-distance). Probably more or less like that. When we reach a point where we cant do anything more about it, and its amount becomes overwhelming, its energy turns into a plastic deformation (~2*m*v) coz there is nothing else which can absorb that energy, and at least we try to control that then (e.g. car impact crashes with studies of its plastic deformations).

And missing steps are not an issue, only unreliability at micro-scale. Missing steps are outside of the "default" premises, thats just "tuning".

Again i hope i am not too far off. Also so you know, you should question everything, even what i say. I am typically one of the dumbest fish in the pond, and there is just so much that i dont know.

Well, I suppose you gave the correct answer for the question that I asked,
and I managed to make it look like I don't understand the physics...

Inertia is proportional to mass. So my statement implied a machine with
zero mass - which granted - is not possible. But that isn't really what I meant.
I meant that what we need is a machine with a mobile mass that is
negligible with respect to the forces driving it. Or at least a mechanism that
adds very little inertia to that already existing in the motors - imagine a motor
flapping a feather around - not very useful, but the feather has close to no
inertia as far as the motor is concerned.

I'm trying to think of ways to drastically simplify and remove mass. If you
think of how small the mass of the plastic being deposited is then most
machines that I've seen I regard as disproportionately massive in comparison
- although they are very clever and non trivial to design and build in your
garage - so my hat is off to everyone who has successfully built one!

I want to achieve an extremely simplified mechanism using light but stiff
materials - like carbon fiber tubing instead of steel rods for example.
The design should be such that all parts that have to have significant mass
such as motors are not mobile. The path generation algorithm should
produce paths that take advantage of properties of the machine - whatever
they are. The goal for me is to produce a high resolution machine but that
can still produce parts in a reasonably quick time and is not too expensive
to build. Why? Because its a fun challenge!

Really sorry in my post i do usually think things step by step because it helps me alone. Its how i get to understand things. Also doing it little steps points out weaknesses and mistakes and gives me a change to get corrected by others if anything is wrong.

If it would of been on that road before, i would get to be a person able give a short specific answer that sums all things up in a wise manner. There are a few who post like that, but those belong to ppls smarter than me. I think because they done the learning part till now, and they just want to help. I aint there just yet. Most of these roads are new to me. So i am learning, my posts are longer, exposing all steps and opinions can give me a chance to get corrected, reinforced, put things in proper order, etc. This how i have to do it atm coz i havent found other way that would work better for me. So by now we can say my motivation is more egoistic than you thought, but certainly had no other objectives than that.

Cheers

No need to be sorry Noobman. Intelligent people usually do think things through
in small steps in order to gain a more thorough understanding. And I would think
that people just starting out in thinking about something complex like this
(such as me) appreciate detailed answers to their questions or pointers to where
the information.