The Most Self-Replicating RepRap Yet

The goal of the RepRap project was always a machine that could replicate itself. The project began with the RepRap Darwin, a machine with a frame made nearly entirely of threaded rods, and progressed to the Mendel, with a slightly higher proportion of printed parts. Around 2011, the goal of self-replication fell by the wayside after some money was thrown around. The goal now, it seems, is to create the 3D printer with the best profit margins. That doesn’t mean there still isn’t a small contingent of RepRappers out there trying to improve the status quo and create a printer that can truly self-replicate. [Revar] is one of those tinkerers, and he has just released the RepRap Snappy, a snap-together 3D printer built nearly entirely out of 3D printed parts.

Other 3D printers designed around the idea of self-replication, like the RepRap Morgan and the Simpson family of printers, use strange kinematics. The reason for this is that Cartesian bots can’t print up to the limits of their frame, yet self-replication requires all parts be replicated at the same scale.

[Revar] is setting a new tack in the problem of printer self-replication and is joining parts together with snap fit connectors. The entire frame of the Snappy printer is built out of small parts that interlock to form larger units.

Another of the tricks up [Revar]’s scheme is reducing the number of ‘vitamins’ or parts that cannot be 3D printed. This includes belts, motors, screws, and electronics. You can’t really print machine screws yet, but [Revar] did manage to eliminate some belts and bearings. He’s using a rack and pinion system, all made with printed parts. It’s a technique that hasn’t been seen before, but it does seem to work rather well.

[Revar] has made all the files for the printed parts available in his repository. If you have enough filament, these files are enough to print 73% of the RepRap Snappy.

Thanks [Matt] for sending this one in. Video below.

46 thoughts on “The Most Self-Replicating RepRap Yet

    1. The most amusing way I can think of to have that question answered, would be to see a machine 5 prints deep (So pictures of a Snappy printed from a snappy which was printed from another snappy which had been printed on another snappy printed off of a first generation snappy which was printed on any other machine)

          1. Considering that we’ve managed to build machines more precise than smashing rocks with other rocks, I would hazard a guess that it is indeed possible to build a machine with greater precision than that of your tools.

          2. @Poose
            There’re plenty of things harder than granite. it’s 1/3 quartz (mohs 7) and 2/3 feldspars (mohs 6-6.5). Any of the following will easily polish granite: corundum (aluminum oxides), certain garnets, topaz, and a whole host of man made abrasives including commonly available chromium oxide.

        1. You can use a similar trick to check the squareness of a table saw fence or guide. Cut a scrap piece 4 times, each time you use the last face cut as the reference. At the end you should have something pretty close to a perfect rectangle. Any deviation will result in additive error. Making even small errors into something measureable with basic tools.

      1. Every printer is (or should be) calibrated before printing usable parts. If you 1st generation printer has been tuned so that a 20mm cube is really 20mm on all sides, the extrusion volume calibrated properly, and the usual quality gremlins worked out, then the 2nd generation should have an equal chance of being calibrated to the same quality standards.

        The photocopying a photocopy generational degradation analogy doesn’t work here.

        1. A print of a scan that’s a print of that scan that’s a print of that scan that’s a print of that scan that’s a print of that scan does apply though. (it looks pretty cool by the way).

          1. Not exactly. All of these would be printing the original document (using your scanner analogy), all of the printers would be printing the original scan just the printer would change. The only variables would be if the snappy could print high enough quality to make a part up to the specifications, and your ability to put it together properly.

    1. On a relatively fast machine @.2mm layer height, depending on infill, expect 100-130 hours and close to two Kilos of plastic.

      I would go with smaller layers for sliding surfaces and use higher infill on some structural parts, which would probably push it up over the 150hour mark, and then you have mistakes, time between part changes, etc… Many printers can’t even go that long without serious maintenance either, so expect a couple weeks worth of your spare time.

      I wouldn’t consider this an easy build by any means, in fact, I would rank it up there pretty high if you expect good results. Remember, if you mess up gluing a part, it messes up a bunch of parts that need to be redone.

      1. Yeah, i simply divided the volume to maximum extrusion rate and got to 70 hours. But then i had no idea what correction factor use… 2 is too optimistic? or is it as bad as 5? I just thought that at 200 hours of print time it would easily require 10 of assistance to the printer.

    1. It’s harder to make an anti-backlash system without intricate gears and springs and bearings.

      Think about how the pinion gear is. There’s play between the gear and the rack that depends on how high it rides, and also in the pinion shaft hole. The rack also has to be perfectly parallel with the carriage rails or else there’s more backlash at one end than at the other, making it hard to compensate it with software. The simple straight rack and pinion also doesn’t advance entirely smoothly – it speeds up and slows down slightly as the pinion turns between teeth.

      All told, it’s quite sloppy and difficult to construct right. To make it operate smoothly and tightly, you usually need some kind of helical gears.

    2. Rack and pinion has been used before. I’ve seen it laser cut and printed. I personally use sections of belt fixed to my axis edges to make a “rack” my pinion is a standard toothed pulley. I’ve eliminated hundreds of parts from my design by getting rid of regular belts in the design.

      1. Using a belt that’s fastened to aluminum angle or square tubing would help me quite a bit with a current project. It needs a conveyor-type mechanism and I think using a fixed belt could reduce some errors with slippage and whatnot. Do you affix the belt with tension or is it glued or something?

        1. https://www.youtube.com/user/peterthinks/videos I tack glue one end, add glue to the length (ca glue) fix it to the length of the platform and keep it pressed to a sheet to a sheet of glass till it cures. Check out my youtube to see the design, I make damn good prints with that monster. better than a $3000 commercial. as for backlash there isn’t any, I cant the motor a tiny bit so it engages the belt at a tiny angle, it is fully engaged in either direction.

    3. I have seen it a couple of times. I think there are other things better than it, like for example simple fishing line on the axis of the motor. A good belt and aluminium gear would not be that expensive, but they don’t fit the bill of “as much as possible 3d printed” … so that is why they are not popular.

  1. Percentage is not a great metric to measure against… If I were to print same exact model but tack on 10x as much plastic in unnecessary parts, I’d get 96% self-printable..
    Maybe it should be measured in how many/which parts are still non-printable, instead?

  2. Does his path algorithm need to be better optimized?
    It looks like you’d get rebound (not sure if that’s the technical term; I mean loss of accuracy due to momentum of the x-y table & a large object) issues doing all that rapid infill on objects that fill the whole print volume.

    Since it’s snap together solvent welding would increase strength appreciably. Probably not too much increase in rigidity though and I’m not sure if that’s in the spirit of self-replicating.

  3. I’m all for Reprap, it’s a nice long term goal, but it always boils down to one question: Would you rather print the entire printer, or would you rather have a better/longer lasting printer.

    Printed slides, which are the core of precision, are a long ways away from matching something you can get off the shelf for a few dollars in terms of accuracy and longevity. While some will say this allows anyone access to one, it hardly changes anything. In the end, you still need motors, which are either bought, and shows you have access to linear rail/sliders, or they are ripped out of old scanners/printers which have linear rail/sliders right in them.

    1. It seems to me that the ‘machine which can print itself’ thing would be better served using UV curing SLS technology. With more weight moving around rigidity becomes more and more important, so using a light weight laser as the moving print head and a simple water tight container as build volume seems to limit the negatives from using snap together parts. Also a heated head and build plate designed to melt the plastic your machine is made from seems like a bad thing. Am I missing something or has it already been done? Is it just the cost of UV resin which is keeping this from happening?

  4. neat idea and attractive execution, and it looks like it’s printing OK, but…
    HOLY CRAP !!! I can hear the backlash. It no doubt is translating to accuracy and precision. Maybe over time it can be improved.

    1. Thanks for the link. I had never looked into herringbone (double helical) gears before. I imagine that they would make gear trains much quieter than they are with normal spur gears. And with 3d printing, the past machining difficulties are eliminated.

  5. I think the problem with the self replicating goal is two much goals much higer on the 3D printer buyer’s list are cost and quality. You get a much cheaper and better quality printer frame buying some aluminium extrusion or laser cut plywood than you do paying someone for 100’s of hours of printer time (or the 2KG of filament).

    I’ve wanted to print the Tantillus for a while though, as a portable second printer.

  6. I was always under the impression that the real reason the printers built using articulated arms were built had nothing to do with pint volume, and EVERYTHING to do with accuracy. People have been dividing up large objects into smaller ones for a long time, so this project comes as no real surprise.

    That being said, Such as system is inherently less accurate, due to the impossibility of building anywhere NEAR as smooth rails and rods with fused filament directly. What comes out will always have small bumps and lumps, and these will transfer to whatever they are used to print at a larger scale. They may not be observable with the human eye here, but give it a few generations and they’ll render the printer unusable.

    The articulated arms however, only have a sliding surface at their joints, and the smoothness issue is solved with some bolts from the hardware store. Because they are often tensioned with fishing line wound around the stepper’s shaft, this also eliminates the horrific backlash, slop, and general crapiness of any 3D printed rack and pinion or screw based system.

  7. perhaps attach the rack separately, so that when it wears out you only need to replace the rack. Or alternately, add a thin slither of material, like a ribbon to sit between the rack and the gear. Simply replace the ribbon and gear when they wear. Naturally selecting the ribbon to be softer than the rack.

    1. says on the website, about $300 in parts + printer + time to print.
      Which I think is not that great, since here in EU you can get complete kits for 350 EUR already….which saves you the huge time to print the parts and the need of a printer in the first place.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.