Sweet results with Heater 2.0

Saturday, September 01 2007 @ 06:46 AM PDT

Contributed by Windell

We've been test driving the new heater design on the CandyFab, and things look pretty sweet indeed. Quick summary of the results: The new design allows a significant increase in the effective resolution of the CandyFab; as a rough estimate, our effective pixel size reduces from about 1/5" to about 1/16". Oh, and the output is edible. =D
Read on for a few more pictures and details.

Let's start with a baseline. Remember our vast array of little dodecahedrons?
We made the dodecahedrons as objects for destructive testing (objects to be destroyed?), but they do serve as a useful representative of our print quality with the original heater design. These certainly look less defined than many of the other objects that we've fabricated. However it's just an illusion: these are just small objects that let you see all the flaws that also exist at the larger print scale.
To some extent, this is intentional. When we made these, we were trying to make sturdy, identical objects that were close to the limit of how small of an object we could make on the CandyFab. So having a bunch of these around has turned them into our de facto resolution test objects (although we have had some good ideas for more formal future test patterns).

But, let's not digress too much. The following four pictures compare one of the little dodecahedrons from the cookie sheet to a new dodecahedron that we've printed using the exact same 3D model but with the new heater as the print head. The horizontal resolution for these models is 20 DPI in each direction, the vertical step size is 1/16".

(You can click on the pictures above see them in higher resolution.)

Here's one more picture of just the new one:

There are still some remaining defects and kinks to work out. One of the interesting things is that the coloration of the sculpture is much more controlled than with previous versions of the heater; it's much more precise. You might note that there is still some variation, which comes across as (faint) darkening along the sides of (for example) the lower, front pentagon. This is actually reproducible in nature; the same level of darkening was independently achieved on many different layers to achieve that set of stripes. That means that changes to the control software could almost certainly compensate for the effect by spending less time heating those edges.

We have one other partial model that was printed out at a different scale; a piece of a larger dodecahedron:

What is not obvious is that these new models, with such small beam cross sections (less than 1/4"), are quite fragile. The large dodecahedron above, if it were completed, would break very easily. For a model of that size, we probably need a cross section of about 3/8" to be sturdy enough to be handled. We will have to build up some increasingly sturdy models to print out, in order to compensate for the smaller effective pixel size. Of course, printing larger objects with higher resolution means that print time gets large!

The new heater also allows us to begin to address some of the culinary implications of CandyFab technology. One of the things that has been rather obvious from the start is that no one wants to eat a twelve-pound sugar sculpture. In fact, twelve grams is pushing it. So what can you do?
To help decorate a birthday cake this month (chocolate hazelnut... mmmm....), I CandyFabbed some two-dimensional snowflakes. These were printed at high resolution, using only a single layer build. The snowflakes were mostly clear with mild caramelization to the light amber stage.

From past experience, we've seen that the first layer in any build is always the most difficult, because we're just blowing air at a layer of pure sugar, not at a thin layer of sugar atop a solid model below. Therefore it is difficult to get consistent color, depth of fusion, or even consistent fusion, especially in a layer that is trying to be as thin a 1/16".
We tried to fab 15 snowflakes in a single batch. About ten of them were acceptable, with all six points actually attached-- quite a feat for such tiny beams as 3/16" x 1/16" -- and no webbed feet. (An unreasonably good yield.) A few points on some of the snowflakes were a bit darker than others, again due to the difficulty of fabing single-layer objects. One thing that might be worth trying is to fab single-layer objects in the future on a very thin layer of sugar sitting atop a piece of (e.g.) parchment paper. That might give the benefit of a hard surface right below the thin layer of sugar, but come off cleanly after printing.

Besides changing the heater and tubing and inspecting our pump, we have had to make a few other changes in order to produce food-grade output on the CandyFab. The most fundamental change is really procedural: treating the sugar like food. That means handling the sugar with clean hands, putting it into clean containers, and so on. We established a separate set of storage buckets for "clean" food grade sugar and for the sugar that we've been less careful with over time. We also constructed an easily-removed (and therefore easy to clean) inner fabric liner (the "culinary liner") for the print bed that goes over the main canvas liner. (Read more about the main liner here.) In the photo above you can see the white fabric of the inner liner on the edges of the sugar bed. In this model, sugar that spills over the edge of the culinary liner falls into the collection system of the main canvas liner and is not returned to the "clean" sugar supply. (It's not wasted-- it will be used with the other not-for-use-as-food sugar to make larger scale models that won't get eaten.
One other thing to note here is that some of the color variation is reproducable between the snowflakes, indicating that the cause is programmatic; almost certainly caused by the same hinting errors that caused the stripes on the dodecahedron above.

So how does it taste? Pretty damn good, actually. Like hard rock candy with its characteristic mineral note, but with the addition of the least hint of smoky, sweet caramel. It reminds me of what you get when you peel the layer off the top of the creme brulee and taste it alone without the custard... not that anyone would do that of course. ;)

The implications of CandyFab as a cooking technology are potentially interesting, and we are definitely interested in finding other applications for the process. The CandyFab Culinary Applications Forum is the best place to ask questions and get your food-related ideas heard. (And we have other forums for other topics!) And as always, if you're interested in helping out with the CandyFab project-- in terms of design, engineering, applications, or building one, we can always use your help.

Comments (16)

The CandyFab Project
http://www.candyfab.org/article.php/heater2results

	Let's start with a baseline. Remember our vast array of little dodecahedrons?
We made the dodecahedrons as objects for destructive testing (objects to be destroyed?), but they do serve as a useful representative of our print quality with the original heater design. These certainly look less defined than many of the other objects that we've fabricated. However it's just an illusion: these are just small objects that let you see all the flaws that also exist at the larger print scale. To some extent, this is intentional. When we made these, we were trying to make sturdy, identical objects that were close to the limit of how small of an object we could make on the CandyFab. So having a bunch of these around has turned them into our de facto resolution test objects (although we have had some good ideas for more formal future test patterns). But, let's not digress too much. The following four pictures compare one of the little dodecahedrons from the cookie sheet to a new dodecahedron that we've printed using the exact same 3D model but with the new heater as the print head. The horizontal resolution for these models is 20 DPI in each direction, the vertical step size is 1/16".


(You can click on the pictures above see them in higher resolution.) Here's one more picture of just the new one: There are still some remaining defects and kinks to work out. One of the interesting things is that the coloration of the sculpture is much more controlled than with previous versions of the heater; it's much more precise. You might note that there is still some variation, which comes across as (faint) darkening along the sides of (for example) the lower, front pentagon. This is actually reproducible in nature; the same level of darkening was independently achieved on many different layers to achieve that set of stripes. That means that changes to the control software could almost certainly compensate for the effect by spending less time heating those edges. We have one other partial model that was printed out at a different scale; a piece of a larger dodecahedron: What is not obvious is that these new models, with such small beam cross sections (less than 1/4"), are quite fragile. The large dodecahedron above, if it were completed, would break very easily. For a model of that size, we probably need a cross section of about 3/8" to be sturdy enough to be handled. We will have to build up some increasingly sturdy models to print out, in order to compensate for the smaller effective pixel size. Of course, printing larger objects with higher resolution means that print time gets large! The new heater also allows us to begin to address some of the culinary implications of CandyFab technology. One of the things that has been rather obvious from the start is that no one wants to eat a twelve-pound sugar sculpture. In fact, twelve grams is pushing it. So what can you do? To help decorate a birthday cake this month (chocolate hazelnut... mmmm....), I CandyFabbed some two-dimensional snowflakes. These were printed at high resolution, using only a single layer build. The snowflakes were mostly clear with mild caramelization to the light amber stage. From past experience, we've seen that the first layer in any build is always the most difficult, because we're just blowing air at a layer of pure sugar, not at a thin layer of sugar atop a solid model below. Therefore it is difficult to get consistent color, depth of fusion, or even consistent fusion, especially in a layer that is trying to be as thin a 1/16". We tried to fab 15 snowflakes in a single batch. About ten of them were acceptable, with all six points actually attached-- quite a feat for such tiny beams as 3/16" x 1/16" -- and no webbed feet. (An unreasonably good yield.) A few points on some of the snowflakes were a bit darker than others, again due to the difficulty of fabing single-layer objects. One thing that might be worth trying is to fab single-layer objects in the future on a very thin layer of sugar sitting atop a piece of (e.g.) parchment paper. That might give the benefit of a hard surface right below the thin layer of sugar, but come off cleanly after printing. Besides changing the heater and tubing and inspecting our pump, we have had to make a few other changes in order to produce food-grade output on the CandyFab. The most fundamental change is really procedural: treating the sugar like food. That means handling the sugar with clean hands, putting it into clean containers, and so on. We established a separate set of storage buckets for "clean" food grade sugar and for the sugar that we've been less careful with over time. We also constructed an easily-removed (and therefore easy to clean) inner fabric liner (the "culinary liner") for the print bed that goes over the main canvas liner. (Read more about the main liner here.) In the photo above you can see the white fabric of the inner liner on the edges of the sugar bed. In this model, sugar that spills over the edge of the culinary liner falls into the collection system of the main canvas liner and is not returned to the "clean" sugar supply. (It's not wasted-- it will be used with the other not-for-use-as-food sugar to make larger scale models that won't get eaten. One other thing to note here is that some of the color variation is reproducable between the snowflakes, indicating that the cause is programmatic; almost certainly caused by the same hinting errors that caused the stripes on the dodecahedron above. So how does it taste? Pretty damn good, actually. Like hard rock candy with its characteristic mineral note, but with the addition of the least hint of smoky, sweet caramel. It reminds me of what you get when you peel the layer off the top of the creme brulee and taste it alone without the custard... not that anyone would do that of course. ;) The implications of CandyFab as a cooking technology are potentially interesting, and we are definitely interested in finding other applications for the process. The CandyFab Culinary Applications Forum is the best place to ask questions and get your food-related ideas heard. (And we have other forums for other topics!) And as always, if you're interested in helping out with the CandyFab project-- in terms of design, engineering, applications, or building one, we can always use your help.