Software for huge reprap

I've decided to try building a LeBigRep [reprap.org] type printer.

This will take a while (about 6 months I think) but first I want to ask something really important: Will the reprap software be able to handle a 1m build area? Replicator G? Skeinforge?

not a software issue, but I'm wondering if warping becomes unmanageable at some size. Has anybody actually built one of these 1m^3 repraps?

I was thinking that it should be a rather minor modification to make the y axis of the mendel "arbitrarily" large (well maybe a meter. you'll start running out of torque at some point, and ensuring flatness over that distance might be difficult).

this would help if you wanted to print something that was long on one axis, like a paddle or an airplane wing, or a boat model. with an unheated bed, i don't think it would work, and it will take a lot of watts to run such a large heated bed.

it doesn't look like he's using t slot aluminum, which i think would be a much better choice.

The software and firmware can pretty much handle any size. The only issue is if you use microstepping, where the number of steps to reach the extremes could possibly be too big for the firmware to handle, though I'm not totally sure on that.

not a software issue, but I'm wondering if warping becomes unmanageable at some size. Has anybody actually built one of these 1m^3 repraps?

http://reprap.org/wiki/MegaMendel

---

-Sebastien, RepRap.org library gnome.

---

Remember, you're all RepRap developers (once you've joined the super-secret developer mailing list), and the wiki, RepRap.org, [reprap.org] is for everyone and everything!

With a heated bed and a perfboard nothing should warp.

Seeing as this is a huge size thing I think I'd go with threaded rod rather than belts.

Also another problem presents itself: big rolls of plastic. 10-15kg rolls should be available for this thing to work.

Threaded rod may be a bad idea. If the size of the bed, in steps, is over a certain amount (varies by firmware) it will overflow, causing problems. Threaded rod often means many, many more steps per mm, thus a smaller maximum supported-by-the-firmware build size.

The fived_on_arduino firmware uses a signed 32-bit int for position (in steps) so the max bed size is 32766 steps.

The reprap.org 5D firmware uses a "long" which I *think* on the 8-bit ARM platform is also 32 bits, so the size limit would be the same.

There are more firmwares, but those are the only two I'm familiar with at all.

---

--
I'm building it with Baling Wire

I knew there were some limitations.

I'm worried about slipping and general inaccuracy if I'm using such large belts.

Do the math before giving it up. The key is "how many steps per mm?"

Standard mendel, toothed belt 5 mm per tooth, 10 toothed pulley.
full step is 4 steps/mm = 8191mm max
1/16 step is 64 steps/mm = 511mm max

somewhere between those settings is the optimal for your machine. It looks like you could do a 1 meter machine with 1/8 microstepping without running into the 32767 limit. 1/16 microstepping would cause problems though.

Threaded rod, 1/4-20 (200*20 = 4000 steps/inch /25.4=157.48 steps/mm)
full step is 157 steps/mm = 208mm max

You won't be able to successfully make a 1 meter machine with 1/4-20 threaded rod.

Threaded rod 3/8-16 (200*16 = 3200 steps/inch /25.4 =126 steps/mm)
full step is 126 steps/mm = 260mm max

Can't do it with 3/8-16 threaded rod either.

I'm not familiar enough with metric thread sized to calculate for metric threaded rod, but it's probably pretty similar. Looks like you'll almost have to use a belt instead of threaded rod for a 1 meter machine. Either that or re-write the firmware to use int64_t for position instead of int32_t.

---

--
I'm building it with Baling Wire

Rewriting the firmware int64_t shouldn't be too tall of an order. Obviously I should read up on memory requirements first but it's doable.

126steps/mm means 126000steps/m. That's doable.

Quote

The fived_on_arduino firmware uses a signed 32-bit int for position (in steps) so the max bed size is 32766 steps.

A 32 bit in will store 2^31-1 = 2147483647 and a 16 bit int will store 2^15-1 = 32767.

---

[www.hydraraptor.blogspot.com]

OH! Right. You're correct, I recalled the number for 16 bit ints and applied it to 32 bit ints.

Everything I've said in this thread is wrong, and no firmware will have a problem with steps sizes even with 1/16 stepping on 1/4-20 threaded rod and sizes MUCH bigger than a meter.

my bad.

---

--
I'm building it with Baling Wire

Using 1/4-20 rod is incredibly slow. I had them on my McWire for a short while before converting to belt driven.

perhaps i should have said, "have any ~1m^3 objects been built", or any useful objects been built that could not have been built on a mendel?

I guess my caution is to make a machine that fits your ordinary needs, not your outlier needs.

because, after all, this thing prints plastic. I don't see a square meter of plastic being very useful, and i doubt you'd want to experiment with building something, as the resources become cost prohibitive. you are going to start small and scale up, and at some point, plastic is not the proper material.

(I used to sell camping equipment, and some people used to want to get 0 degree F sleeping bags, when they never went camping after September. i didn't earn commission, so i would tell them that it was a waste of money. most of the year they would be way too hot, and it would just be extra weight and bulk they would have to carry. it was not something they had considered, and i usually saved my customers money and earned their trust by suggesting a sleeping bag that fit their average needs.)

My cnc converted mendel is chugging along quite nicely at the moment so the big reprap is just the next step for me.

Saw a video of one guy printing chairs. He was using a granule extruder (which had pulsing problems, causing a rough surface finish) mounted on a huge robot arm. About a 2cm wide extrusion. Yes, that's 2cm. I'd estimate that his extrusion head weighed about one or two hundred pounds.

Not something any reprap I've seen is prepared to do yet. But something we could do. You'd really have to redesign your print head, and then redesign you frame to carry such a huge weight!

---

--
I'm building it with Baling Wire

Yes, and his robot arm costs in excess of $100,000.

---

Bob Morrison
Wörth am Rhein, Germany
"Luke, use the source!"
BLOG - PHOTOS - Thingiverse

Yes, and so did 3D printers, a few short years ago. Some 3D printers still do. Of course, they get lots better results than our sub-$1K machines. Looking at his speed of extrusion, and the inaccuracies caused by inconsistent extrusion pressure, I'd bet you could get better results with a sub-$5K machine. Probably a lot less than that, really. Probably enough less that you could build such a machine and sell it for $5K.

I do wonder about warping though. The design he was printing was such that I don't think it'd be subject to much in the way of warping stress.

---

--
I'm building it with Baling Wire

A heated bed and a perfboard platform should prevent bad warping. I do think we'll see warping on the order of a few mm but nothing more.

I found some thermal videos of FDM printing and it looks like the plastic strings cools down very fast anyway: [www.youtube.com]

Quote

I found some thermal videos of FDM printing and it looks like the plastic strings cools down very fast anyway:

Yes that's the problem. Ideally you want it all to stay hot (just below the melting point) until the end of the build so that it all cools and shrinks at the same time. That way you just get an object slightly smaller instead of a curled up object caused by the new layers bonding to the layer below and then shrinking, the next layer doing the same and so on.

---

[www.hydraraptor.blogspot.com]

It's obviously not a problem since an industrial FDM machine works by cooling the strands (the one in the thermal video)

AKAIK industrial machines all have a heated chamber at about 70-80C. So yes you want the ABS filament to cool quickly from 240 to 80C, i.e. just the point where it freezes, because while it is a liquid the thermal contraction does not affect the layer below. But when it is solid you don't want it to shrink any more until the end of the build.

---

[www.hydraraptor.blogspot.com]

In any case that's a hardware issue. I'll cross that bridge when I get to it.

Edited 1 time(s). Last edit at 01/26/2011 11:51AM by Tramagust.

Good luck with this project. It would be wise to use a heated bed/chamber with that Size.A moving extruder at the top and the platform only moving in Z direction. That way you can have all printed parts and motors outside the heated area.
You will probably want to look into larger nozzles ant thicker filament as well. 4mm Filament is readily available for around the same price/kg as 3mm. This gives you 77% more volume per mm of filament length. [www.orbi-tech.de] should give you an impression, however there are cheaper suppliers i think. 2,2 Kg on spools seems the usual size even for welding machines, but maybe you can have them make larger spools for you.

I do wonder if that 4mm ABS is good for 3d printing. I've read that it must have certain properties for it to work. In any case orbi-tech is very expensive.

Edited 1 time(s). Last edit at 01/26/2011 02:40PM by Tramagust.

They are, this example was rather to show, that 4mm is readily available at no extra cost. I have not yet tried their material.
When ordering 10 Spools the price becomes more reasonable - but then again that is a lot of printing to do. I did not intend to recommend them.
The point is rather that you will have to look into other feedstock than 3mm if you build that printer. Thicker than 4mm will be a problem with spooling, granule-extruders are heavy, and thus no solution when having a moving extruder.

... for printing with 4mm-filament you need more torque or bigger/stronger motors.

When i first mentioned Orbitech 3 years ago, they only had 4mm in stock ... with time the demand for 3mm rose (all for reprap), so they developed 3mm filament too

Now there are interests for changing to 1.7mm-filament - here you can extrude with lower torque or weaker/cheaper motors ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

It is not a given that he needs more torqe, that depends on the extrusion diameter, if he prints ,08mm diameter or even 1 mm a nema 17 might be enough

Hi Tramagust, I am currently trying to slice single stl models that max out the build area of my machine (400 x 400 x 320mm) with skeinforge/replicatorg. I am running into some out of memory issues slicing parts that large and I am trying to figure out how to go about this (see thread). This would definitely be an issue with 1m^3 parts and is something you should consider.

---

Jim Smith
GrassRootsEngineering.com

I was afraid of that. Maybe netfabb slicing would be more friendly

Is it the *size* that's the problem, or the *complexity*?

I doubt that shear size is really the problem. There may be a limit on the total number of layers, but layer thickness on a large machine may be 2mm (or more) and so. There may be a limit on the total number of g-code moves that can be generated. That's a complexity issue that can also be fixed by reducing the total number of layers.

Having said that, skeinforge *shouldn't* run out of memory no matter what you throw at it, and the fact that it does is clearly a bug in skeinforge. You might try sending a bug report to the maintainer Enrique. Make sure to include your preferances and the model you're having trouble with. Also upgrade to the latest version and verify it still happens, of course. SOP when reporting bugs against anything, I suppose.

---

--
I'm building it with Baling Wire