This post follows on as a back story to our engagement, so read that first. The basic premise was a locked box that knows where it is, and will only open in the right location. You get told the number of tried left and the distance to the destination - then it’s up to you to narrow down where to go!

Sarah’s birthday (when she was presented with the finished box) was in August, and I didn’t get all the electronics until June. Roughly tested and working, the second challenge was finding a box the right size, with enough space inside to hide all the electronics but without looking cumbersome. I found this 1900’s tea caddy on eBay and set to work taking it apart.

The software, a custom board and some of the electronics were provided by Mikal Hart, and you can read details of how he designed the original box on his website.

Below you can see the finished inside of mine, along with a silk lining to hide all of the gubbins. From left to right are: Servo motor to open the latch, Arduino board (blue), Mikal’s board that connects everything together (green), a GPS receiver and of course AA batteries (where would we be without them?). Behind the credit card sits the display connected with good old serial cable. The switch and latch are visible at the bottom.

There are a few variations made from the original design. I wanted a very plain looking exterior, without any fixing or screws showing, so as to add to the intrigue of how it could be opened; so I mounted the GPS unit inside the box and not on the lid. The box is wood so I didn’t think this would be a problem and it worked fine as long as you were outside or near a window.

The screen was mounted from the inside, held in place by the credit card so there were no fixings on the outside. To get it flush the wood had to be carved out to leave a thin, 1mm lip around the display. This was done painstakingly with a dremel multi-tool. I also covered the USB connector (used for programming and as an emergency backdoor), with an enormous grub screw made from the sawn off end of a large bolt, and slotted with a saw.

The latch was improved upon from the original chopstick and metal eyelets, to a sliding brass rod and hinge. By removing the hinge pin and using a rod narrower than the hinge pin diameter, a reliable yet solid sliding lock could be made, operated by the servo. This meant that it was doubly important to test the box out, as once locked, and with no external screws, it would be very hard to open without significant damage. There ensued a lot of trial locking and unlocking of the box with the lid open before finally closing it up - a process that made me feel quite sick!

I also reduced the radius of error from 2 miles down to 40m. I knew exactly where I wanted the box to open, and I knew that the carriage ride would stop within 10m of that location, but I was a bit nervous just how accurately the device would be. GPS has an error of around 10m anyway, plus we were in a built up area with the receiver inside a box. Imagine after all that effort if it wouldn’t open when I needed it too? Alternatively, if the radius was set too big it might open before we got there. This involved a little testing, setting the box to open nearby and going for a walk. I eventually settled on 40m as a compromised between the right place and a guaranteed open.