If anyone reading this has chatted to me over the last half-year, you might be aware that I’ve been working on quite a big secret project – and no more is it secret!
Inspired by various family members needing solutions that didn’t have a monthly subscription cost, didn’t automatically phone for an ambulance when one wasn’t needed, and didn’t need the fallen person to press a red button (that sort of thing’s hard if you’re unconsious or have broken limbs!), we started work on a solution. It occured to us that perhaps we weren’t the only ones with this problem – so in April of this year, three friends and I founded a Living Electronics Ltd, to facilitate development of FallAlarm: a fall-detecting, relative texting alarm for those prone to falls.
The idea is simple: the alarm detects a fall with an accelerometer, a microcontroller then interprets the signal, sends a message to a far away server, which alerts the relative through a text message service or push notification on their phone.
A quick dig in the empire-of-dirt and poke with the soldering iron and the first prototype was made: an accelerometer, ESP8266 and LiPo battery. The basic formula is the same as we’re using now. That was nice – but there were a few issues – it was quite bulky, we wanted to add some other features (better buttons, battery charge & boost), so I was cajoled into designing a PCB….
Version 1 of the boards arrived courtesy of OSHPark (an excellent hobbyist PCB fabrication for those who don’t know). After burning ourselves several times soldering small pieces of metal, mistakenly ordering metric components (they were very very small – dust in a packet), we had a soldered PCB. There were quite a few errors: pins being pulled the wrong way due to a slight lack of I2C knowledge, my fantastic attempt at capacitive touch being entirely dysfunctional, not being able to program the chip because you couldn’t get to the right pins, using the wrong regulator, and lots of traces being connected wrong. Other than that it was perfect!
After some ‘creative edits’, we eventually got a working V1, which was enough for Ann to do some code-fu and get some working firmware going. After much jumping on beds to test it, it worked! However, it was apparent some changes needed to be made. The board was way too large, the alarm kept false-triggering if it was moved suddenly, and – well – there were clearly quite a few PCB errors!
In response, I spun up this board:
This time we included an Altimeter (to reduce false alarms by verifying that the unit had decreased in altitude), a transistor (to trigger the reset pulse more reliably), and in response to the it needs to be smaller comment, I squeezed it all into a tiny 3x2cm PCB. This seemed like a great idea on paper – the board looked nice and large on my screen! However, when the boards came through, it was incredibly hard to solder. There were vias under chips and pins to save space, components on both sides (a nightmare for hot air soldering), and even a few accidental components-on-top-of-other components. We got another almost-working version of the board – and as we found out later, the hardware was actually fine, just the layout of the board made it nigh-on impossible for a human to make.
I was determined for the next version of the boards to be a working version – and it looks like they are; I received a joyful Slack during Hackaday Uncon to let me know that Dan had boards outputting data to a debug spreadsheet over WiFi!
The hardware on these boards is almost identical to before. The only changes were that we moved the voltage regulator back to a bigger package so it could cope with more current, installed headers for an FTDI programmer and LoRa radio breakout (there’s a team in Norwich working on getting a Things Network running in the city – this would be an ideal use case), and added a switch to set the chip into program mode. All the components are laid out better, with the vias nice and far from the chips, and all the components on one side so we can attach a LiPo battery to the other side. Also, I put our faces on the silkscreen – ’cause why not?
Now we’ve got a working board (and 40 of them thanks to DirtyPCBs!), we’re hoping to set up a beta-testing trial with users near Norwich, so that we can get feedback and develop the boards into a saleable product. We’re just waiting on some more Accelerometers and switches to arrive on the slow boat so we can begin building more units for trial. Barny is also working on a backend solution so that we can move to a push-notification based system instead of paying for a text service.
I’ve learnt a huge amount so far – and the whole team is excited for what comes next!
Photos in this article are mostly thanks to Dan – who remembers to take photos far more often than I do!