Falling Fall

Recording Data

A phone can capture accelerometer data while falling. But you can’t just strap a phone to a leaf, it’ll fall too fast.

Solution: make a huge “leaf” that spreads the phone’s mass across greater surface area. Then drop it off a balcony thirty times.

Iterations

The original “leaf” fell two stories in about a second. When recording accelerometer data at 25hz, that meant our data was too sparse. To slow down the leaf, we quadrupled its size to increase surface area and air resistance.

Fologram, the app we used to record position data, lost wifi signal during falls because the phone was travelling so far. So we went off wifi and moved to new locations.

Without wind, the leaf doesn’t spin: and a huge leaf needs a ton of wind.

Our data models leaves falling on a calm day, without flipping.

Data Processing

We recorded 30 “Leaf Falls”. Fologram captures at roughly 25 frames per second. We used the same name convention for all takes so that we could iterate through them programmatically

Grasshopper Definition

We moved the data input scripts offscreen and created a set of “panels” that would allow us to interact with the data naturally. We created a formula that takes an integer slider as input and spits out the full filename of the relevant “leaf drop” take. This allowed us to iterate through all our takes effortlessly without the need for multiple versions.

Next we needed to clean up the data. We started our recordings while still holding the Big Leaf, and ended each recording a few seconds after it hit the ground, so we needed to remove these non-moving frames from each individual take. We made a helper function that let us figure out where each “drop” actually started and ended.

Then we fed these values into a switch statement that displayed only the planes that we actually cared about, depending on which Leaf Drop we were looking at.

Data Cleanup

Data-cleaning scripts flipped the plane axis because Fologram records planes perpendicularly to the phone screen, and offset the (X, Y, Z) values of the Leaf Drop so we could center them around the origin – because recording in the real world is messy, and each leaf drop was recorded meters away from every other one.

Only once all the drops were centered around (0,0,0) and displaying the relevant frames could we begin to truly play around with the form and shape of the leaves falling!

One of the joys of writing your own creative tools is that when you need a feature, you just go build it. After the data was cleaned up, we realized that the leaf drops looked best if we stopped them before they hit the ground – so we wrote a helper function to cut off the bottom of the Leaf Drop wherever we dynamically thought it looked best.

Displaying Meshes

Our first idea was to render a 3D scan of a leaf, and map one to each plane recorded from the leaf drop.

But when we fed this mesh onto the planes we’d captured (plus interpolated planes between them), it looked like a dinosaur spine, not a leaf falling.

Displaying Lines

Our next idea was to display a loft between the outlines of the leaf.

We modulated the size of the leaf based on its speed, so that the leaf got bigger as it slowed down, so that it would take more “visual space” when it took more “temporal space.”

This looked intriguing, but again we’d lost the leafyness. Instead of dinosaur bones, we had somehow made dinosaur muscles.

Final Form

Ultimately we realized that a leaf’s path needs to visually show the beginning differently than the end. So we changed the script that modulated the size based on the leaf’s speed, and instead based size on the leaf’s position in the whole drop.

This created the visual “swoosh” effect we were looking for.

We want this sculpture to be a series of 1.5m-tall hanging lamps that can hang from tree branches throughout winter and bring warmth and light to the colder month. That’s beyond the scope of what we could produce in a 2-week assignment. We scaled back our idea to create two prototypes using slightly different materials and processes. We chose a maple leaf because it is recognizable, has a complex edge, and actually loses its leaves in Fall! The leaves are connected with a “whoosh” prism based on the actual fall speed data

Manufacturing

We laser-cut leaves in increasing size, and 3D printed struts taken from the raw fall data of our favorite Leaf Drop.