Solid freeform fabrication: DIY, on the cheap, and made of pure sugar
In February we gave a sneak preview of our project to construct a home-built three dimensional fabricator. Our design goals were (1) a low cost design leveraging recycled components (2) large printable volume emphasized over high resolution, and (3) ability to use low-cost printing media including granulated sugar. We are extremely pleased to be able to report that it has been a success: Our three dimensional fabricator is now fully operational and we have used it to print several large, low-resolution, objects out of pure sugar.
The general idea of our build process-- that of stacking solid two-dimensional printed layers-- is actually common to most solid freeform fabrication methods. Our machine employs what we believe is a fairly novel low-cost technology to accomplish this: selective hot air sintering and melting (SHASAM).
The printing process begins with a bed of a granular printing media that has a fairly low melting point. Using a narrow, directed, low-velocity beam of hot air, we selectively fuse together the print media, forming a two-dimensional image out of the fused grains. We then lower the bed by a small amount, add a thin flat layer of media to the top of the bed, and selectively fuse the media in the new layer, forming a two dimensional image that is also fused to any overlapping fused areas in the layer below. By repeating this process, a three-dimensional object is slowly built up. At the end of the build, the bed is raised to its original position, disinterring the fabricated model, while unused media is reclaimed for use in building the next object.
Our process is very much like a low-cost version of Selective Laser Sintering, or Selective Laser Melting, which are commercial processes used for plastic and direct metal printing. Rather than using a high-power CO2 or YAG laser ($5k and up), we use hot air created with the help of a $10 heating element.
Trading off a laser for a heat gun gives us lower resolution but at much lower cost, and is typical of our approach to 3D fabrication. We have taken a very different approach from most other fab projects (e.g., Fab@Home and RepRap) in that we have a comparatively large printable volume, but less need for precision and high resolution. Our fabricator is not designed for prototyping machine parts; it's designed for fun, for large-scale 3D illustration, for sculpting, architectural models, and other applications where resolution isn't the only important factor. We estimate the total cost to build a machine with similar capability to be in the neighborhood of $500. Realistically, the cost of any project like this is not a fixed number, and since recycled components are involved, the actual cost could range up or down by a factor of two depending on how resourceful the builder is.
There are a number of different print media that may be suitable for use with SHASAM fabrication: many types of plastics and waxes have low melting points and are available in granular or powder form. Beyond that, there are a number of interesting foods-- chocolate chips come to mind-- that can be used with the process However, one of the most interesting possibilities is using table sugar.
Granulated Sugar: low cost print media
The price of sugar compares quite favorably to the polycaprolactone (a low melting point polyester) used by the reprap project which costs about $4.00 a pound. As it turns out, even $4.00 per pound is quite inexpensive compared to the media for many other solid freeform fabrication systems.
Beyond just lowering the media cost of a given fabricated object, using a low-cost medium can be leveraged to make large-volume printing both practical and economical. Our fabricator has a maximum printable volume of 24 x 13.5 x 9 inches (61 x 34 x 23 cm)-- 2916 cubic inches, or 1.7 cubic feet, and holds a little more than 100 pounds of sugar, which costs about $37 retail.
For a fun exercise, look up how much it would cost to make a similar model on a prototyping industry standard $20,000 Dimension ABS 3D printer-- if it could print objects anywhere near that big. (Hint: it's more than $0.93.)
Mechanics and Electronics
The big idea of the mechanical system is that we take a hot air gun and move it around a bed of sugar, selectively fusing a set of points before lowering the bed of sugar and adding a new layer.
Our hot air gun is based upon the design of a hot air rework station. However, we have heavily modified it, and learned how to make an equivalent system inexpensively. The heater design now essentially consists of a 500 W, $10 air heating element and a small air pump-- a $5 aquarium air pump works well. At a minimum, use of the heater element requires a housing to be constructed, the air pump and a control system that can provide a the chosen amount of power to the heating element. We have seen that the element can be driven directly from 120 V, with duty cycle controlled by an inexpensive digital relay. The heater element is hardly new technology; it's the baby sister of the one in your hair dryer. None the less, it's well designed and quite suitable for this application.
The original head was not designed to operate at both high temperature and low air flow; it tends to overheat easily. One improvement that we made that has been hugely beneficial is to mount a cooling fan right next to this structure, keeping it cool on the outside while in use.
The X and Y axis motion control systems are based on belt drives and quadrature-encoded motors recycled from two old HP plotters, a large one and a small one. This is one of those places that your resourcefulness can save you a lot of money: The number of old-generation inkjets and plotters out there is truly stunning-- go find a couple, and make them do something useful again.
In order to control the quadrature-encoded motors that came on our printer parts, we designed custom digital servo circuits that cost about $10 each to build. The circuits are based around a high-power analog output stage and an AVR microcontroller that accept position commands. The position commands are sent using a higher precision version of standard hobby servo PWM control code, where the position command is encoded in the width of a positive pulse between one and two milliseconds long. We will be writing up and releasing the hardware design as well as the source code (under the GPL) for these servo controllers in the near future.
The hot air gun is mounted to the belt-driven carriage on the Y axis of the printer. The Y-axis belt-drive system is mounted, on one end, to a linear bearing that slides along a steel rail. That bearing is pushed by the belt-driven carriage on the X axis, through a rubber band low-cost flexible rubber coupling. The other end of the Y-axis belt-drive system is supported by a free-rolling rubber wheel from the hardware store.
Here you can see the model as drawn in Sketchup, and the base that we constructed from that model. If you want to take a closer look, you can download the model here. (144 kB ZIP archive of sketchup .skp document)
Besides the three motion axes, there is also a heater controller that is used to control the power delivered to the hot air heating element. Together, the four controllers (X,Y,Z, Temperature) require four axes of computer control.
Canvas Liner
Wrapped around the wooden base is a flexible canvas liner that prevents sugar from leaking out in strange places and assists in recovering unused media. Canvas is a good choice for this application because it is strong, durable, woven tightly enough to contain granular media like sugar, and washable. We got ours at a fabric store for $7/yard, in 60-inch width, and we needed about five yards. If you're trying to save costs, you might be able to do better elsewhere, e.g., buying canvas drop cloths intended for painting.
Software
There are several different layers to the software needed to control a three-dimensional fabricator, and they are implemented in our system with a variety of different techniques. We begin with a 3D model generated in (or imported into) POV-Ray, and then render the POV-Ray image as a set of two-dimensional bitmaps of slices through the image. The bitmaps are generated in such a way that they directly represent which points will, or will not, have the printing medium fused. We then take the bitmaps and use them to "draw" with our hot air gun at all of the black points on the bitmap.
Here is one of our 3D models, along with one of the generated 2D bitmap slices through that object:
Operating the 3D fabricator requires precision motion control in three directions, which is potentially difficult. Computer control and interface are provide through a MAKE Controller. Presently we are using an old student version of LabVIEW to control the MAKE Controller-- reading in a 2D bitmap, parsing it into a simple rastered toolpath, and converting that to position commands, sent to the MAKE Controller using UDP packets. Labview is, of course, not free software, and any suggestions about open-source solutions that would do the job nicely are welcome. (PD and Processing seem like possible directions, but we'd like to hear what you think in the comments.)
While the Make Controller has many remarkable capabilities, we are hardly taking advantage of them here; it is strictly acting as a computer-controlled device to output four servo-motor control code signals. Budget conscious builders may want to instead consider using a dedicated servo controller, like this Micro Serial Servo Controller, from Pololu, a precision 8-channel servo interface starting at $17.95.
Making things with the fabricator
Now that we've got all our parts together, let's fab some sugar objects. The effective horizontal resolution of our fabricator is presently limited to around 2 mm by the very one-point-oh design of our hot air nozzles, but can in principle be made much higher even while using granulated sugar as the print medium. The resolution is determined by a number of factors, including the air nozzle size, the air temperature and flow rate, and (obviously) the position step size in the three directions. Printing at a higher resolution takes longer, so we have actually been operating it in a low-resolution mode in order to produce some sample objects-- quickly-- before the Maker Faire. All of the objects on this page were made with pixel (well, voxel) size 2.5 x 2.5 x 2.7 mm (10 x 10 x 9 DPI), where the 3D models have been properly quantized to account for the larger vertical step size. Even at this low resolution setting, the total number of printable points in our fabricator is over 2.6 megavoxels.
Finally, here is a group of three objects that we've made out of pure sugar: A little dodecahedron, the toroidal coil, and the twenty-inch-long wood screw
So how does it taste?
Like praline, no doubt.
While our process has incredible potential for making interesting food, we are still in the early stages of prototyping and we have not yet worked with the sugar under conditions that could be construed as proper food handling procedures. We are instead at this point treating the sugar as a relatively safe (but not edible) industrial chemical and prototyping medium. There is no fundamental obstacle to food-safe 3D fabrication-- however we still need to carefully audit the system and make sure, for example, that the air pump for the hot air does not contain any substances that could contaminate food.
See it at the Maker Faire
Our completed fabricator will make its public debut next week at the 2007 Bay Area Maker Faire. (Our Maker Faire program entry is here.) We will be bringing the machine itself and some of our fabricated sugar objects. We've decided to spend our time at the faire showing off the printer and its parts, rather than actually using it to fabricate objects. One reason is safety; we have discovered that the First Law of Laboratory Work (Hot glass looks exactly the same as cold glass) holds true for molten hot sugar as well.
We made the sign from recycled and scrap acrylic for a total cost of about $20, cutting out the letters and segments on a laser cutter before cementing them in place.
We will continue to document the CandyFab 4000 as time goes on. In the mean time, you can find more pictures of the system in this flickr photoset.