Light boxes are notoriously ugly and intrusive items to have on one’s desk. For our responsive light box, Melody and I attempted to prototype something that would be, if not less intrusive, more welcome to sit on one’s desk. We went with a pyramid shape meant to sit at an angle from a user’s laptop. By placing the light box at an angle, the 90 degree angle forms a visual congruency with a rectangular desk. In turn, this shape also gives the user a cue to place the device at an angle from their laptop, which ensures they are receiving indirect light (as direct light is potentially damaging to the cornea). 

 

The box stands 10 inches tall, and is 15 inches wide, dimensions that suit it for placement next to a standard 13”-15” laptop screen. 

Melody and I planned on having two different modes for our responsive light therapy box. We wanted an orange light for night sessions, and a full-spectrum white light for day sessions. Given the two different jobs, these lights needed to be independently controlled. 

 

Our other parameter was form factor—we needed everything to fit in our physical box, we wanted to have only one physical plug for the user to worry about, and we wanted to use only one arduino.

Working with one repurposed lamp cord meant splitting both wires in the cord into separate connections to both the light bulbs and the relays. In short, a bad idea.

 

After several attempts involving the stripping of wire and dangerously patching with electrical tape, I finally created a relatively safe circuit with the help of the fine folks at SVA’s Visible Futures Lab. The circuit shows how the two bulbs, and the two relays are connected to one arduino, and one AC input (a wall outlet). 

To accomplish this, only one part of the live wire was tapped into, using a set of wire strippers to move the insulation forward a centimeter or two. I soldered another strip of live wire to that open piece, and ran both of them into their respective relays. You can see the heat shrink in the picture above, marking where that splice was made. Other than that piece, I didn’t have to split any wires (which raises resistance), use any other electrical tape, or otherwise compromise the quality of the lamp cord we were using. 

 

So Melody and I achieved our form factor wants of having one arduino and one cord from the box to the wall, while operating two separate light bulbs.

 

We started with a clear piece of acrylic, laser cut to the dimensions we needed for our box. We then diffused the acrylic using an orbital sander with medium grain sandpaper to scratch the surface in a manner that was both pleasing, but also obfuscated the lightbulbs within the encasement. After this we created an illustrator file with the manipulated image of a zen sand formation we found online, and fed this into the Roland printer program. From here we printed out the image onto white adhesive vinyl, with the help of the folks at Visible Futures Lab. 

 

After applying the vinyl to the acrylic board, we placed the board into the inset we had created on our MDF board (the material we used for the encasement), using a CNC router. 

A responsive light box needs both data and an algorithm to act on that data. To achieve this we used a google spreadsheet that automatically updated whenever the physical fitbit was synced with the user’s computer. Melody then worked tirelessly to get this data into Processing, where we could parse it according to our needs. She ended up working from a tutorial from the inimitable Jer Thorp. Once we had the data in Processing, and we could access individual cells, we were able to manipulate that data according to our needs. 

Primarily we used FitBit’s proprietary activity score to determine how much light therapy the user should receive. We used the number of minutes of recorded sleep to determine how much orange light therapy the user should receive. We created multiple conditions for how to use and read the fitbit data. 

For example, if it was after 6pm, and the user had slept between 6.5 and 7.5 hours, then the orange light would glow for 30 minutes. If it was before 6pm and the user had an activity score at the low end of our range, the white light would glow for 45 minutes. 

The real frustrations in this stage of the product were to get all of the programs to talk to each other. For instance, at first we had trouble sending more than one character from Processing to Arduino, limiting our ability to employ the additional LEDs we had attached to the box. These LEDs were an exploration into visually showing the user how much light therapy they were in for—when all 4 were lit, the user was in for the maximum amount of time, with each light going out, in order, after a quarter of the time had passed.