Mine is bigger... I guess you know that one
Size matter... maybe that one too
The one you do not know is ... the bigger the size, the bigger the errors, hence... precision matters, and to obtain precision, the rep rap FDM models are out in space...
I can build you an FDM printer the size of a shipping container but...
If it is built by the standards to make it cheap, the output will be not commercial quality, and if I build it to make you classy cozy commercial quality output prints, than the printer will not be chap, at all.
Where lays the difference?
One thing is you cannot have shaking, belt dis-alignments, tooth skipping, belt elasticity, and precision in the same design. And what do the most FDM desk top size machines use? Belts. Why? Cheap, easy to work with, no engineering needed to study it.
On the other hand same size of a cheap FDM desktop, say 500 bucks ready to print, if you go ball screw motion powered, with encoder steppers, we talk 4 gran minimum, and the quality of the output goes up a lot, I mean some 200% or more.
Scale that to a cube the size of a 40Ft container, and you have belt technology on wheeled porters giving you a price of maybe 10 gran (just throwing numbers here) compared with ball screw technology that on a 6 meters length with 2.5 meters width might cost you 100 gran like nothing, plus assembling it.
How do I know that? Well, I am preparing to make me one, that is how.
Nice project! I want to follow your lead.
Getting high accuracy inexpensively is the right goal. The issues, of course, go well beyond belts versus ball screws or even stepper motors versus servo motors. The driver software is also a player.
I totally agree with you presentation on the parameters that are effected by size. But their are other parameters that you are not considering. The biggest is the basic nature of the means of supporting the printing head and obtaining the motion. It appears that your assuming a Cartesian approach.
Also in your mix should be delta type with three input triangulating to a position. Delta is a fixed linkage and achieves position by moving one end of the linkage along a linear path. For all approaches one end of the linkages are attached to the printing or cutting head. Commercial CNC machining has fixed linkage ends but vary the length. To get angular the second set of linkages are not just offset from the first at the ends but also vary in length.
Maslow CNC router has one of the axis being gravity thus direction is fixed and the connection of that linkage moves in the plane of the router, cuts typically 4'x8' sheets. The other two linkages of the triangulation have fixed ends so the length of the linkage is changed to define a position. Steel chain with sprockets along with weight is used to minimize errors. One issue being worked on is that position changes the load, portion of the weight carried by each chain. This approach could be scaled to doing sheet the size of over-sea containers.
Third is the hanging approach that has made using stair wells 3D printed items as tall as a person. The DIY approach comes directly from the founder Torbjorn Ludvigsen and a variation smaller is the Arcus-3D-C1 cable 3D printer. The size is effected by the stiffness of the cable and weight. Four linkages are used, three pull the printing head down and to a location on a plan defined by the fourth cable that is typically much longer from the ceiling, defining a curved surface, if the cable is held fixed in length. The control of the cable length can be done a either end. I believe that to improve accuracy the latest design has encoder measuring length and a motor that drives the drum that holds the cable that can have multiple layers. With the measurement of tension the stretch of the cable can be calculated. and the chainette length can be also calculated, math call catenary.
The Arcus approach is more like a mix between the delta and the Maslow where the three cable links hold the printing head up and gravity from the weight of the head the fourth axis. Basically turning Torbjorn Ludvigsen machine up side down. The TV cameras for sports events use the approach to give a from above view of individuals and plays in action and a pleasing point of view.
You may thing the accuracy of a cable approach is not that good. Consider that for the very large telescopes, both optical and radio with fixed mirrors or minimally movable mirrors the instrument that collect the image etc. are all cable system. So the accuracy is needed within ~ 1/4 the wave length. and for long exposures that accuracy is maintained while the instrument sweeps across the mirror.
Like any computer geek can tell you, software is not a problem, because software is always more advanced than hardware, it is the hardware that needs to catch up both in costs and in quality.
I am the hardware man, if you want to jump in the boat let me know.
Emil Pop, I am also a hardware guy. Designed a spectrograph instrument chassis, part of a team. The spectrograph rides on those cables for the South Africa optical telescope. Thermal expansion of the chassis is important as the air cools down at night. I would like to see the software and hardware that move the item riding the cables and the cross check instruments. My bet is that lasers are used to provide information to tweak the position which is why within the instrument itself we had to use materials that minimized thermal expansion. Talking about the wave length of light. That is the level of dimensions for tweaks.
Thomas,
Is there any public information on how the cable system for the South African telescope works?
Emil,
I am not a software guy. I am familiar with a company that makes very large milling machines, big enough to handle a work piece as big as several 40 ft containers. Vibration is the enemy of any machining operation. Long spans (arms, gantrys, etc) make vibration control extremely difficult. I know this company has put a lot of effort into software. I suspect they have figured out a way to reduce the effect of vibration through software. They claim to hold some pretty tight tolerances on these very large work pieces.
The stiffness issue may not be quite so relevant with a printer. I expect the forces on the print head are pretty small. But, will the work piece come out with the same dimensions indicated by the position of the print head or will there be a lot of variation in the spread of the material as it is layed down and becomes solid? If the material's change in dimension from liquid to solid is constant and accounted for in the software what happens when you change material? I assume each material would be treated as a separate object within the software with its own characteristics.
The variation in the final output of the machine is the sum of all the sources of variation in the machine. But, since they are variations they add by the square root of the sum of their squares. That is good news because it means the largest source of variation has the largest effect on the final result by far. As a demonstration of how important this is, take two sources of vibration, one of 1 unit magnitude and one of 5 units magnitude. 1 squared + 5 squared = 26. The square root of 26 = 5.099
It looks like all you have to do is stop the 1 and you are in fat city. Not true. Total elimination of the 1 source gets you a reduction of only 0.099 units! But if you took the 1 out of the 5 source, variation is now the square root of 17 = 4.123.
The point of all that is wherever you are in the development process, identify the largest source of variation and reduce that. Do not spend resources on the little ones. Of course, as the machine gets better and better some of those little sources of variation become the largest sources left. Then they must be dealt with. But you will not know which until you get to that stage. Eventually you get down to the wave length of yellow light like we use with an optical flat in flatness measurement.
I would love to participate in your project at whatever level you desire. I have my own machine shop (hobby - I don't do work for money) and I have been an engineer for 52 years. I have a lot of manufacturing and product development experience. But I don't write code, yet.
Well Pete, is moving the machine that holds me back, the rest of it I figured it out in time.
I mean this is not a router or mill, level of vibrations induced by the main cutting tool are zero, there is no cutting tool.
The highest level of vibrations is induced by sudden change of direction in the X axis and Y axis while printing the infill sections, and that is no big deal on a 2.5 meter span X axix traveling on 6 or even 12 meters long Y axis as a bridge crane.
As long as I stick to ball screw motion most of them vibrations are gone up to a tenth of a mm, below that here are still vibrations but they do not affect the final output unless I want to go with 0.1 mm diameter nozzle, which I might at a certain point.
If I go belt and pulleys vibratins go up to even 2 mm amplitude with a frequency of up to 10HZ, and that is trouble for the final output.
I need a good C++ coder that chews Marlin to give me a hand with this, and things will come together.
Emil, That is a really big machine. i have been thinking about a router but nothing like your machine. Ballscrews for a 12 meter run are going to incredibly expensive, if you can even find someone who can make them. I am not your guy for software. Sorry.
Torbjorn Ludvigsen is a programmer and has a foundation that should help you. I would learn as much as you can about his design. He had to deal with the problems discussed here.
https://www.seeker.com/open-source-prototype-turns-any-room-into-a-3d-printer-2313142390.html
He has started a web sit where you can download 3D print parts for the third version of the Hanging 3D Printer. https://www.linkedin.com/in/torbjorn-ludvigsen/
Software site: https://github.com/tobbelobb/hangprinter
Printed parts site: https://hangprinter.org/doc/v3/
modification design: https://www.youtube.com/watch?v=S9vbJFpzeTQ
another variation with bigger changes: https://hackaday.io/project/26938-arcus-3d-c1-cable-3d-printer
Timing belts used in 3D printers are chosen not for the small size of the printers. Timing belts have different materials used for their cables to minimize stretch. Steel cables are used in larger belts. So your going to need belts that have tensions sufficient to over power the loads, effective fixed length even adding dynamic motion. That is the point of lead screws. That screw for the loads applied has insignificant stretch.
You do not have height in a container. so consider high tensions cable or big belt system. again this is already being used.
https://drmrehorst.blogspot.com/2018/08/corexy-mechanism-layout-and-belt.html
https://www.instructables.com/id/Build-a-Parallel-Straightedge-Drafting-Table/
I have included the second site showing that the printer is just a version of the older solution use on drafting tables, decades old and proven.
The hanging printer used the tension of the ceiling cable to make the dynamic loads have a minimal effect on position. In the above approach the cables or belt are held at a high tension also minimizing the effect of dynamics and the metal frame in compression similarly to the ceiling cable just balance the cables and also not significantly effected by dynamics.
Another and more typical drafting board cable layout closer to the printer. https://www.rexart.com/media/alvin_1101_2201_instructions.pdf
Cable and pulleys on a container size printer might work, but due to long distances cable might need to be thicker, and that requires larger pulleys with the space constraints consequences derived.
Another way I am looking at it is the reverse of ball screw system.
A rack worm; a screw spinning against straight rack rail to move the Y axis (actually 2 of them, one each side of the bridge) the way those Russians did on their vehicles
It seems it is called a Worm Rack Drive
If you don't receive the email within an hour (and you've checked your Spam folder), email us as confirmation@grabcad.com.