Larger Reprap?

Can the reprap be scaled up to make larger items?

A bicycle frame, for example?

Or a computer desk even?

Could you just build a larger box and extend the X and Y dimensions?

Thanks!

Joe

I wouldn't want a bike frame from abs or a desk really but theoretically you could scale the reprap up to size. Just have to reengineer the supports.

You could, however the larger you make the box, there will be a noticeable droop or sag in the X and Y axes which will interfere with the Z.

RyanHendrickson Wrote:
-------------------------------------------------------
> I wouldn't want a bike frame from abs or a desk
> really

why not?

Grogyan Wrote:
-------------------------------------------------------
> You could, however the larger you make the box,
> there will be a noticeable droop or sag in the X
> and Y axes which will interfere with the Z.

That should be overcomeable.

Instead of having the rods support the X and Y actuators (or whatever they're called), you could put them on rollers and have them move back & forth in a piece of C-channel (which could be supported from the outside) and the rod would only provide the motive force.

... for 'really big' repstraps you'll use much sturdier/stiffer materials (there are big cartesian 'bots' for handling ship-containers with some ten meters dimensions) or you'll try with parallel kinematics and lightweight rods made from carbon.

I made a tripod setup with hinges made from magnetic orbs ( [builders.reprap.org] ), that could easily be scaled up or down (here i scaled down ), but the controlling software is something more complicated ...

Viktor

I'm sorry if I seem a bit silly, but I can't visualize how that won't stop the sag, especially when the carriage is in the middle of either the x or y axes.

You could in theory go up to 1m for the x and y axes before the sag becomes problematic.
You could though use Adrian's extruder where the pinch drive mechanizim is exterior of the carriage, and therefore is less of a problem

I've got right angled aluminium to hold up both Y and Z axes on my RepStrap, spanning 0.5m, the sag isn't very noticeable, but if I went to 1m it would be useless, instead of C channel, try getting box extrusion aluminium (available from most hardware shops) span it across a gap and measure the sag when loaded with weight in the centre.

Why I mention box extrusion, is that its a lot stronger and U channel or right angle extrusions and should be quite resilient against sag

Hi Grogyan,

there are some methods and materials for building really stiff and big frames - e.g. big and high precise CNC-mills are built from stone, not from steel or aluminium.

Similar DIY-frames are made from hollow structures out of steel or aluminium and filled with sand-epoxy-mixtures, what's a really rigid compound sometimes even better than real stone and much cheaper and easier in handling.

When building lightweight, then the frames are made from glass-fabrics or carbon hardened with epoxy and complex hollow structures, sometimes filled with foam.

A good material for rigid frames is dibond (a sandwich of two metall-plates and foam or plastic between).

Parallel kinematics are often made with hollow carbon-tubes what's stiff but lightweight - as all applying forces run in axial direction of the tubes, you didn't have much bending or sag ...

Stifness over large spans using single elements is often acheived using an I-Beam cross section.

I am thinking here of building construction (Steel Girders and Masonite joists) and the sort of overhead gantry crane that is common in engineering works.

For full sheet size cartesian machinery, a moving gantry style bot is arguably the most cost effective, with the gantry spanning the smaller dimension (X and Y).

Darwin is in effect a moving gantry Cartesian bot (X and Y) with an added Z axis the work bed.

aka47

---

Necessity hopefully becomes the absentee parent of successfully invented children.

Thanks, but I was merely suggesting problems and options with a fairly small scale RepRap setup becoming larger.

Assuming a larger than normal Darwin, would need more strength on the rods or removing some of the strain from stuff you can buy readily, box extrusion.

Unless of course the guy wants to build a large greater then 2m x 2m cubed bot, then yes other methods will need to be found

Also, one of the design objectives of the RepRap project was to achieve self replication. Building larger makes this more difficult using the current plastic extruding head.

Another option might be to look at it like a "RepRap for metal" that can take the high temperatures needed to cut or extrude metal so that it can form its own pieces on a larger scale. Imagine having a RepRap like machine cutting out your aluminum tubing for your bicycle. And yes I know that cnc machines do this already, but the whole point is to get a cnc machine design to self replicate. Then it can sit comfortably at the dinner table with RepRap.

I'm working on something like this, with the hopes of expanding the number of working mediums RepRap can deal with.

Of course, the wider the span, the stronger the structure has to be.

I've seen C-channel in widths of up to 12" (for building the under-structure of manufactured homes).

Of course, that would be over-kill for spanning 3-4 meters, but you get the idea.

as for printing with metal, what about using something like a wire welder? Of course, you'd need some way to deal with the slag.

Or you could make a really cheesy frame, and simply compensate with the programming.

When I get to the point of making my metal cutting machine (I expect in a year or so), the programming for that will allow for compensating for sag in the frame, and bends in the frame, and a bunch of other stuff.

That machine will have a fairly rigid frame though, to allow it to cut steel. It will not be an accurate frame though. The programming will fix the inaccuracy of it.

Tony

Just out of curiosity, how exactly will the programming fix inaccuracy in the frame?

I only know of one way to do that and that is by building a digital model of the characteristics of the frame for each step and mapping that to the actual xyz coordinates on the bot.

Thats the easy part, but what kind of sensor are you going to use to get millimeter positioning accuracy? Actually I think I've answered most of my own questions.

In the history of the blog post there is a link to the blog comments

where I made the below comment.

I hope this is enlightening. Tell me though your solution.


You reminded me of anoto dots [www.anoto.com]. They are ridiculously simple to make and can be printed on a piece of paper. using a cheap camera(mouser.com has them for $1.56-$3.14) on white paper, you could get accurate tracking for less than $10.

Furthermore, you can use a free program like context free art [www.contextfreeart.org], and this source code,

startshape dotpage

rule dotpage{
100* {y .2} {
100* {x .2} dotgroup {}
}
}

rule dotgroup{
DOT{x .2 y .2}
DOT{x .2 y .3}
DOT{x .3 y .2}
DOT{x .3y .3}
}

rule DOT{
CIRCLE { size 0.01 x 0.01}
}

rule DOT{
CIRCLE { size 0.01 x -0.01}
}

rule DOT{
CIRCLE { size 0.01 y 0.01}
}

rule DOT{
CIRCLE { size 0.01 y -0.01}
}

right now. (work is a bit slow at the moment)

Why is sag a bad thing? Seriously?

Set up a feedback loop so that as a bottom-middle point of the tray "droops" per a laser-bounce sensor, you reduce the amount of downwards tray movement to compensate.

Obviously there's a limit but this would at least let you scale it up a considerable amount...?

Groygan,

I completely agree with you.

You bring up the excellent point that sag is a big problem with big machines and there is a quick to implement design change that can fix it for at least the next factor of two bigger.

An alternate design change would be to replace the relatively thin solid rod with wide tube. This would not be quite as stiff as box extrusion, but it would be quicker to implement since the design is currently using rods. Either way, the design could be modified to use a box extrusion or a wide tube relatively quickly.

And a quick, good enough solution is what we need.

I am currently building a RepStrap that has a 22"x22"x18" (X x Y x Z) build area.

But I also intend on using it as a small cnc router

My solution to the sag problem are some half inch Thompson Rails and slides.

My X and Y axis are assembled but not mounted to there frame yet, all parts are made out of aluminum and the entire 2D Cartesian system weighs about 30lbs with motors factored in. Sitting on my work bench I have no noticeable sag, I may take some measurements later and see if there is any.

My solution to the sagging problem is a granite slab, resting on pontoons, floating in a tank of water. The z-axis actuator is a water pump. There are solutions to every problem.

Reminds me of the milk scanner

[www.etre.com]

R.J.Bowmaan Wrote:
-------------------------------------------------------
> My solution to the sagging problem is a granite
> slab, resting on pontoons, floating in a tank of
> water. The z-axis actuator is a water pump. There
> are solutions to every problem.

And heat the water to keep the "raft" from warping!

One easy possibility for eliminating the sag is to use rotating arms. A long horizontal arm that rotates in the xy plane will have the same droop/sag no matter where it rotates to (being the same distance to the end of the arm no matter how you rotate it). The software would then have to adjust for a curvature in movement for the other axial directions, but you'd eliminate the effects of differing sag.

No gantry use a floating magnetic head that floats across a thick steel plate. No stiffening required.

merlz Wrote:
-------------------------------------------------------
> One easy possibility for eliminating the sag is to
> use rotating arms. A long horizontal arm that
> rotates in the xy plane will have the same
> droop/sag no matter where it rotates to (being the
> same distance to the end of the arm no matter how
> you rotate it).

Unfortunately, your assertion (constant sag) is not always the case; I suspect it's rarely the case. Consider a SCARA arm (two links, with rotary joints both vertical), the outboard link always sees the same bending moment, however, the inboard link sees a combination of bending and torsion, depending on the angle between the two links. In this case, the sag is *not* independent of position.
For any robot with more links, the situation is more complex, and even less likely to be constant (configuration-independent) sag.

Sorry,

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]

... this 'position-depending sag' could be solved by a calibration-matrix when the sag ist the same in every specific position - run a measuring job over the working area and add/subtract the measured disposition ...

Viktor

For a larger reprap, just build it like a suspension bridge, have an arch from which is suspended the gantry.

Don't know about you guys but this looks to me like a mechanical design (Structural Engineering) challenge rather than a challenge that needs a software repair.

Making up for poor mechanics via software strikes me as being a false economy.

I could probably think of lots more useful things to dedicate the processing cycles to.

Like perhaps resolving Cartesian coordinates into angular coordinates or some such.

Sorry to muddy the conversation up a bit.

Cheers

aka47

---

Necessity hopefully becomes the absentee parent of successfully invented children.

... online-position-feedback should solve most issues with sagging.

I was thinking about applying rotary encoders and some wires as in my string-tripod Alex Jonis Toy ( [dev.forums.reprap.org] ) for triangulation of the toolhead - so with cartesion or angular positioning you can measure the position of the toolhead regardles of structural or other displacement ...

Viktor

aka47 Wrote:
-------------------------------------------------------
> Don't know about you guys but this looks to me
> like a mechanical design (Structural Engineering)
> challenge rather than a challenge that needs a
> software repair.
>
> Making up for poor mechanics via software strikes
> me as being a false economy.


Tony, et al,


This is a matter of opinion, however I'll argue that using software (and sensors, if available) to compensate for an imperfect mechanism is a *great* idea -- especially if we want reprap to be widely adopted.

Real mechanisms, made out of real materials, have finite stiffness.
In the presence of gravity, acceleration, or machining forces, such mechanisms will deflect. (In a highly repeatable manner, in most cases, although the model may be very complicated.) Friction and wearing away of material are also issues for mechanisms -- harder to model, but still present.

By using more/harder/stiffer materials, such deflections (etc.) can sometimes be kept small enough that we can neglect them. However, they're still there, so if we want to achieve greater precision (or at least repeatability), using sensors, feedback and software may be the only way to get higher performance from a given mechanical design. Alternatively, better sensing/compensation could lead to acceptable reprapping, using a less expensive structure/mechanism.

Software and sensors, IMHO, are often a more (cost) effective way of improving mechanical performance than throwing more/better metal at the problem -- and are *far* easier to share with fellow reprappers. Contrast this to the hassles of fabricating and shipping mechanical parts (especially big/heavy/stiff and/or tarriffed ones) around the world. It's much easier to send software improvements!

I would also argue that most reprappers have better software and electronics skills and (access to) tools, than they do mechanical ones. So, software and/or electrical improvements (read sensors) are more likely to be broadly adopted than (for example) stress-relieved, flame hardened, hand scraped machine slideways.
Even I'm not going to that level for my repStrap, and I probably have more mechanical tools available to me than 95+ % of reprappers. (Plus somebody willing to lend me scraping tools, if I were that ambitious.)

I'm in favor of improvements in any area that make reprap work better, whether they're mechanical, electrical or software.

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]

I was labouring under the impression that we were able to printout a well designed mechanism.

First port of call for a structural engineering problem should be the engineering of the structure. For me personally the clue is in the name.

If we are not able to engineer, design and print out a sufficiently well designed mechanism then I would agree perhaps a potential route to fixing it is the use of Increased resolution and positional accuracy linked to computed compensation.

These too have their associated costs which may not be neglected. You gets what you pays for.... (and other addages like you don't get owt for nowt etc)

Given that we are striving towards sustainability (ie local provenance of cost effective raw feedstock, which translates directly to home produced components) then the argument of shipping components that we are able to print does not hold. It is a substantial part of the point of RepRap.

Not sure where the more metal idea comes from RepRap is a plastics printer (As Yet).

On the whole I think RepRaper's are pretty much as mixed a bunch as are the constructional techniques required to make a functional machine. Each will have their preference and is wise to make best use of what they already have. To presume all are much of a muchness is neglecting the diversity that is out there.

Similarly to assume that best fit is acheived for all by skewing a design solely in the direction of Computation & Sensing or Even Purely Mechanical is missing the point.

Horses for courses sums it up with the addition of a healthy lump of engineering best fit and compromise.

A full design will have areas, each of which is best fitted to a particular class or type of solution. Given all of the above I am still currently for engineering structure for structural engineering problems. But am equally happy to consider alternatives if their costs/benefits add up in real terms.


Cheers

aka47

---

Necessity hopefully becomes the absentee parent of successfully invented children.

Assuming that potential sag could be fixed by a stronger structure, what would be the minimum software changes necessary to use the larger print area?

I am guessing that I would have to go to the preferences section of the RepRap host, under "globals" and increase the values of "working X" ,"working Y" and "working Z" to whatever the larger reprap is going to allow. But it can't be that simple, or does it? Am I missing something here?

Brew

... it's really that simple ... the only limitation is the variable-size for the data containing the actual position.

When moving/positioning with an stepsize of 10 microns and using long Int you have a max. linear count of 4.63 kilometers per axis

A much more serious problem is the possible speed - imagine an object of 1x1x1m size printed with an accuracy of 0.1mm (not really fine ) with a fabbing speed of 10mm/s ... this would take about 3251 years building time - when no error occurs ><

So you have to dimension your machine in respect to needed accuracy and estimated building times ...

Viktor