Week 6: Building a Solar Barbecue Grill


I had been thinking of using the power of the sun to grill a steak or cook a delicious stir-fry a lot this summer in Redding, CA. With clear blue skies and temperatures well over 100°F almost every day, I think it's hard not to think of all the things you could burn with the concentrated power of our home star – other than our own skin. Solar grills are actually a real thing (not surprisingly, you can find lots of solar barbecue enthusiasts in the South-West) and for this week, I wanted to build one myself for Harvard's astronomy club so that we finally have a reason to also meet up and share our passion for astronomy during daytime. Parabol solar grills can by far reach the highest temperatures, and their impressive design made a perfect challenge this week.

Designing the Parabolic Mirror

The general idea for the mirror was to create a wooden paraboloid support structure, coated with reflective aluminum sheets on the inside. I started by designing the basic layout for the parabolic mirror that would form the core of the grill. I used Fusion 360 to model my grill. Unfortunately, Fusion does not natively integrate a way to create parabolas (or other forms) from functions. In this case, I decided to instead create a rectangle in the middle of my sketch and use the top-left and top-right points as anchors for a conic curve with a rho = 0.5, forming a nice parabola. The dimensions of the rectangle defining my parabola were 100mm x 25mm, giving me a focal point at (0, 25). At this point, I trimmed down the edge of the parabola by 5 mm to place the focal point outside the mirror, since this would make it easier to design a secondary element to house the grill surface.

I also divided the vertical (y-axis) space between the midpoint and the end of the parabola into 8 similar-sized segments that each would house a supporting ring (one for each planet in the solar system).

I continued by extruding the relevant profiles of the 8 planet rings, as well as the inside and outside ribs that would later hold everything in place. The most challenging part of the design came next, which was about designing the press-fit system that locked the 8 planet rings to the 6 inside and 6 outside ribs. At first, it was quite hard for me to picture this fit in my head, until I realized that the system is actually quite simple: the length of the incision on the outside part determines how far it goes into material. To make modelling easier, I chose to only model the incisions I wanted to create in the ribs – and then later cut the planet rings with these incisions. In retrospect, this made the entire design dependent on the 3D arrangement of the components rather than their parameters, which was harder to work with. I should have just modelled the appropriate incisions for the press-fit into the circle sketches defining the rings. This whole design is quite hard to describe in words, so I hope the pictures below help you understand the process better.

I also realized very late in my design that although my design was parametric, it was extremely hard to impossible to change dimensions and have them propagate upwards. The biggest problem came from the constraints created to define the conic curve, which also had several attaching constraints set onto it. Due to this, whenever I tried adjusting the dimensions of the rectangle defining the conic curve, I got an error that the attaching constraints were still applied. This made my prototyping extremely slow and very inflexible. Unfortuntely, this problem only became apparent to me when I did try to update the dimensions of the base parabola to reflect the dimensions of the material we were given. In future, I should make sure to have appropriate sizing from the start, and avoid circular dependencies created on conic curves (although, I think, in this example, it is a design flaw in Fusion).

Exporting and fixing glitches

Just when I thought I was done with designing the parts and arranged them for export, I ran into a lot of glitches that I still do not fully understand and that I had to give my assigned wood working spot up for. The problems always became apparent after I imported my exported DXF file into Aspire and tried cutting the toolpaths. Aspire often did not accept my export because of "unjoined" profile vector paths. I tried to work myself around this by just auto-merging nearby vectors, but that did not solve the problem. It was only by accident that I zoomed into one of the parts too much and realized that all the lines were doubled and that there were lots of double-vectors. I also realized this happened in a few other places of the design. Investigating the issue, I realized that I should better export the top faces myself manually, and arrange the parts into one file in Illustrator. Once I switched to Illustrator to fix the exported file, my iterations got much faster.

It took me a lot more of prototyping time to figure out the best way to make sure the board dimensions would match all my joints. As I could not change the board thickness with ease in Fusion, I instead hacked my way around it by scaling the model as an SVG after export in Illustrator.

Finally, I got one cut working with the right dimensions that I prepared successfully in Aspire and got ready to cut out of a 24 x 18 x 0.5 inch panel.

Cutting the Parts & Assembly

For cutting my parts, I used the desktop-sized CNC machine ShopBot Desktop with a 1/8 inch drill bit. The final cut of the file I prepared in Aspire went smooth and without issues. I used 4 screws that I drilled in around the perimeter of the board, one for each edge. This seemed to have fixed the board enough in place. I then homed the machine and did an aircut of my design to see if the generated machine paths aligned with my expectations and did not cut any nails. I then executed the cut job and waited for the ShopBot to finish it.

During assembly, I realized that in most cases, my joints were very tight and so I could not simply press-fit them. If that is a bug or a feature is still up for debate. Regardless, most joints could be fixed in place with a hammer and some force. In the end, I was quite happy with the assembly and got a final result with a diameter of around 1.2 m on the outermost circle (effective diameter).