Saturday, April 5, 2014

Today’s Biggest 3D Printing Challenges


Devil – Skyler
The state of 3D printing today reminds me of the days when the Commodore 64 ruled the PC world. As much as we struggle with today’s printer technologies, we also love the doors it has opened to us. Yet just as we eventually moved beyond the Commodore, the capabilities of each successive printer generation overcomes some of the limitations of those that preceded it. Unimaginable as it may seem, we may wake up one morning, shuffle down to the kitchen and have our Jean-Luc moment as we utter the command, “Earl grey tea, hot.” In the meantime, lets take a moment to briefly discuss the really big 3D printing challenges we’re now up against.


Glacial Print Speeds

I give a fair number of talks and attend many a school event introducing 3D to teachers, students and parents. I often hand out colorful 3D printed mushrooms with a hole in the stem at these events so that kids can zip tie them to their backpacks and parents can bury them in the junk drawer after a week of wondering, "What am I supposed to do with this?" I’ve probably produced some 500 of these tchatchkis. Perhaps enough to fill a 5 gallon Home Depot bucket. Each mushroom is an assemblage of 3 pieces — a stem, a head and a cap. My lab is converted into a cybernetic mushroom farm the night before an event as shrooms are ganged up 10 or 20 to a printer. The average build time for the 3 parts that go into each mushroom works out to about 45 minutes. 

That’s right, it's taken something like 375 build hours to fill a bucket with 3D printed mushrooms. As amazing as it may seem, I’m not troubled that it takes as much time as it does to produce small batches of parts. This is inherent to the very nature of production in an industrial age. A primary advantage that 3D offers is not faster production times, but rather innovation times. Knowing this, I arrange my production schedule to replicate batches of parts while I sleep or while I’m off doing other productive things. The incremental march toward ever more advanced printers will certainly result in faster machines. But I’m just as interested in innovations in post-production automation that allow multiple batches to be run unattended, thus extending the scale at which a custom production business remains competitive. Printers are called “bots” for a reason. Shouldn't they also cleanup after us?

What I find unacceptable about 3D printers today is the amount of time it takes to iterate during the design process. I come from a software background where a key metric of programmer productivity is how long it takes a developer to make an incremental code change then re-compile and get back into the program debugger to confirm that the bug's been squashed and move on to the next challenge. Production printers should be chosen based on the requirements of production, such as cost, capacity, reliability and product quality. Printers used during the product design process are different animals. They need to be fast — as in one or two orders of magnitude faster than they are today.


High Cost

With a budget of $300, I get to choose between several different models of Chromebooks. Acer has one for $200. I see on Amazon that Dell has a netbook running Windows XP that can be mine for just $99. When I’m ready to splurge I can get a new iPad for as little as $400. A top of the line iPad Air will run me about $900, about the same price as a 13” Macbook Air.

For more than three times the price of Dell’s XP netbook, $350, I can zip tie together a Printrbot Simple kit with a wood frame that doesn’t work very well nor last very long. Makerbot’s lowest cost 3D printer, the Mini, costs $1,400 with a spool of filament. The mid-size Replicator is more than twice that price at $2,900. And the “massive” Z18 is again more than twice the price of the Replicator, flying out the door at a massive $6,500.

If we look back we see that high prices are the norm when disruptive products are first introduced. The original 128K Macintosh was initially priced at $2,495, which amounts to $5,595 today after you adjust for inflation. The cost of printers will come down as capabilities improve. But don’t expect the same price curve, because Moore’s Law doesn’t apply to machines with moving parts. Still, that we’re buying printers with wood frames would seem to imply that the innovation cycle is rife with opportunity and that there’s ample reason for optimism that prices have a long way to fall.


Small Build Volume

The PrintrBot Simple has a build volume of less than 4” cubed. Which is too small to print a teacup. The Makerbot Mini is an inch taller. Even the Z18 is only one foot square and a foot and a half tall — big enough to accommodate a teapot. The problem is, few schools can afford a Z18.

I contend that we shouldn't let the small build volume of school printers be a limitation. We must look at it as an opportunity to design beyond the bounds of a single print job through the use of joinery, assemblies and modular designAn overriding curriculum objective is to advance students' modeling skills while offering a path to success that’s long enough to support students’ imagination yet wide enough to include every kid regardless of their innate abilities. I view joinery as a skill central to real-world success. It's also the means by which to develop at scales beyond the bounds of the build volume. Real world objects are an assemblage of parts connected together via screws, pegs, slots and cotter pins. Sliders, levers and ball joints allow parts to move. Weight and other forces are distributed using beams, rods and cables. Modeling curriculum must teach the use of kits of standard parts like these.


Low Reliability

In 1995 I bought an HP 5MP laserwriter printer. It's been about 5 years since I was last able to find toner cartridges for it. So I don’t use it regularly any more. In fact, I don’t even keep it plugged in except when I’m occasionally feeling nostalgic and print a page to verify that it still works. And 19 years later it does still work.

The first 3D printer I bought was a dual extruder Makerbot Replicator Dual clone out of China. The mean time between failures on that first unit was 30 build hours. That’s right, some piece of hardware hard-failed about once a week on average. It got to the point where I was so experienced repairing it that I could strip the unit down to the motherboard and put it back together again in about an hour. The most difficult problem though was sourcing parts. The only supplier of parts was the manufacturer in China. So I sat through a lot of downtime waiting for parts to arrive. When I realized how unreliable the printer was I ordered one of every mechanical and electrical part I could get my hands on, and two of those that failed most often. To their credit, the company felt so bad about my experience that they proactively sent me another printer. They said I could pick the price I chose to pay for the second unit. I told them I’d pay full price if they agreed to give me free support, including free express shipment of replacement parts. We had a deal. But 3 hours into my first print job on the new unit the whole gantry exploded in a shower of parts because the factory hadn’t tightened any of the screws keeping it in place. I emailed asking to return both printers, but never heard from them again.

What portion of new users are having buzz-kill experiences like mine? How is the overall market potential affected by these poor quality machines?


Design Complexity

In many ways, the world of 3D printing today is similar to the dawn of the computer age in that modeling tools offer a limited set of operations on a small collection of primitive shapes. It’s the geometric equivalent of programming in assembly language. Modeling programs need to evolve and fresh ones introduced that widen the path to success to include a greater proportion of students and non-professionals.

Another aspect of design complexity that slows the innovation cycle of non-professional designers is the lack of standard components. Must every designer be required to reinvent ball joints and hinges instead of using standard, well-designed parts that have been fully vetted by a global crowdsource community? Just as software frameworks are the fuel that power object-oriented programming environments, 3D modeling innovations will surge with the availability of standard geometric component kits that designers can pull primitives, parts and assemblies from for use in their designs.


Limited Materials

Various forms of soft plastic are the most common artifact of a 3D print job today. These materials are certainly colorful, which is important to an emerging technology like 3D printing. The objects produced are strong enough for some applications (like selling 3D printers) though they can hardly be referred to as durable. A steady stream of exotic feedstock materials with interesting properties are regularly announced. But few make significant inroads as compared to the PLA and ABS stalwarts. Users are deterred by the downsides of each new and improved material, whether the downsides be cost, durability, appearance, toxicity or something else. From a materials standpoint, electrical conductivity on par with aluminum, durability on par with polycarbonates, and food-safety (and dishwasher safeness) are each generally considered “killer apps” that will propel 3D printing to the next level.

2 comments:

  1. I can't agree with you any more. All those limitations are crucial when planning any 3d printing workshops, especially for kid whose easily get impatient and lost attraction. I also like your idea of 3D Bucks, maybe I will try it in my next class.

    ReplyDelete
  2. @ Low Reliability

    Well I had way better experiences with an Ultimaker of the first hour.
    I needed to take it apart only once in two years to put in more accurately machined axles. Besides stronger motors where necessary. The rest where rather small details.
    I print a few hours per week.



    @ Small Build Volume

    An Ultimaker has ~20^3 ccm building volume and ~35^3 ccm housing volume.
    Except with propeller blades I haven't yet hit the limit with my projects.



    @ Limited Materials:

    I also recently made the experience that there's not necessarily conservation of misery
    [ http://en.wikipedia.org/wiki/The_law_of_conservation_of_misery ]
    when changing to a different type of plastic.

    I recently tried PET-G:

    + shrinking is almost as low as with PLA (no heated printbed is necessary)
    + a lot higher softening & melting point than PLA
    + higher viscosity -> stringing less than PLA - smoother surfaces
    + no deformation under high long term load (months) like PLA
    + similar hardness but higher toughness than ABS (almost like unchewably hard chewing gum) => the filament can not be incised & broken | cutting with a knive is almost impossible
    + almost no smell (but probably more toxic than PLA)
    - absorbs a bit of water (first layer a bit bubbly) ?
    - made from fossile oil & not biodegradable



    @ Design Complexity

    Well imo OpenSCAD is already like the C-language of 3D printing.

    I'd like to see a domain specific language as elegant as purely functional programming with minimal syntactic shugar i.a. no brace madness (like Haskell / Curry) though.

    An intuitive point and click interaction (e.g. by mouse) in the graphical preview
    would be nice but it needs to be fed back upwards the "decompression chain" in a controlled way into clean manually controlled code. Ugly auto code generation is imo to avoid like the plague.

    Here is what I think of being the "decompression chain":
    high level domain specific language <->
    constructive solid geometry graph <-> (quadrik mesh) <->
    triangle mesh (previewd with e.g. OpenGL) ->
    toolpath -> motor & other digital one bit commands

    I'd imagine intuitive point and click interaction like a combination of say Google Sketchup, Inkscape and Geogebra. In textual OpenSCAD code I frequently encounter occasions where I don't want to specify the evaluation direction (what the freely adjustable parameters are and which parameters follow from them) so I don't have to rewrite the code lateron. Example: planetary gear assembly parameters. This goes into systems of (potentially nonlinear) equations that - dependent on the situation - need to be solved (or iteratively approximated) in different directions - this is CAS stuff and can become veery complicated.

    There's Christopher Olah's ImplicitCAD but it still has serious limitations like no quick preview (and low focus on the native haskell interface ?). The problem seems to be a mismatch between computer architecture and the "decompression chain". OpenSCAD starts lower in the "decompression chain" and thus does not suffer those problems.

    I did a minimal experiment some time ago named (miniSageCAD) to explore the single (unrelated) aspect of elegant semi-infinite volumes (aka algebraic varieties):
    http://www.thingiverse.com/thing:40210


    About standardisation:

    I'am thinking about creating a screw library that takes into account the typical limits of FDM printers. I'am somehow reluctant to force users to adhering to certain parameter combinations though.

    Gears: http://www.thingiverse.com/mechadense/collections/gear-generator-libraries
    Screws: http://www.thingiverse.com/thing:8796
    Screws: http://www.thingiverse.com/thing:45314

    ReplyDelete