The competition ends tomorrow and I have fallen in a heap a week ago. I need to re-group on what I can achieve with my abilities, and be content with that. It will not be the perfect build I was hoping for, it may only partly touch on the concept I am aiming for, but such is life.
Recap
The idea is to build a smart lock box for digital devices to help control digital addiction, more on the idea can be found in part I and coding with LPSDK and accessing a fingerprint scanner in part II, and mechanical stepper motors and vault lock mechanisms in part III and Bluetooth and BTStack in part IV:
- Sentinel Box - Part I - the plan
- Sentinel Box - Part II - back to C
- Sentinel Box - Part III - Stepper Motor and Vault lock mechanism
- Sentinel Box - Part IV - Bluetooth and BTStack
Perfect Build
Looking back honestly, I have learnt an enormous amount across this challenge. I attempted a fully hermetic Rust build, maybe next time. I did, with the help of the amazing Element 14 community, get to successfully drive a stepper motor, fingerprint scanner and connected Bluetooth all for the first time. My final piece was a vault mechanism that look great in initial previews and I was genuinely excited about completing.
The perspex work is where the gap between concept and reality opened up. Precision mechanical assembly requires tolerance, and I underestimated it badly. Only today I had been watching Clickspring’s extraordinary series on building an Antikythera mechanism (YouTube) and there is something humbling about watching someone fit-and-file a part to within a tenth of a millimetre with hand tools. My lock mechanism needed that same care. It did not get it. I cut fast, missed a few steps and thought it better try when I had the right tools and patience to actually get it right and build a mechanism that was “square”, worked and worthy of display.
The final week did not help. A hectic schedule, a weekend away from home and tools. Even though I was excited about the mechanical build, the perspex was not as forgiving as re-burning a new build onto the MAX32630FTHR.
What I did get started on is expanding the Bluetooth communications layer with the browser. As the theoretical setup would need stepper motor setting, reading a number of finger prints and setting up a consensus framework for who can co-operatively open the lock box, doing this via a web site seemed the way to go. The foundation is super sound:
- React 19 for a modern frontend
- Lit 3 hybrid with the React to allow for a custom web component (
<sentinel-ble-manager>) - shadcn + Radix UI + Tailwind CSS for styling
- Vite 8 as the build
The part I am most pleased with is that the browser and the firmware speak exactly the same binary protocol. The MAX32630FTHR stores its config in a 100-byte block in INFO flash, and the web console encodes and decodes the identical layout — offsets and all — using DataView and Uint8Array:
view.setUint16(0x00, CONFIG_MAGIC, true) // magic 0x5B5A ("SB" LE)
view.setUint16(0x02, cfg.vaultSteps, true) // vault_steps
buf[0x04] = enrolledCount // fp_count
buf[0x05] = cfg.unlockPolicy // unlock_policy
buf[0x06] = cfg.relockMinutes // relock_minutes
// fp_name[10]: 8 bytes each starting at 0x08
// fp_role[10]: 1 byte each starting at 0x58
view.setUint16(CRC_OFFSET, crc, true) // CRC-16/CCITT over bytes 0x00–0x61
To make sure nothing gets corrupted in transit the browser runs the same CRC-16/CCITT the firmware uses, so a tampered or truncated blob is caught at both ends:
export function crc16(data: Uint8Array): number {
let crc = 0xffff
for (const byte of data) {
crc ^= byte << 8
for (let i = 0; i < 8; i++) {
crc = crc & 0x8000 ? ((crc << 1) ^ 0x1021) & 0xffff : (crc << 1) & 0xffff
}
}
return crc
}
It is not the kind of JavaScript you write every day, but it is satisfying when the numbers match on both sides of the BLE link.
So the build is not what I imagined. The vault door is not ready, the full orchestration of fingerprint plus Bluetooth plus stepper motor is het to run end-to-end. But the pieces exist. The concept holds. I know now exactly which joints need filing to tolerance and which parts of the stack are solid. That is not failure, that is a prototype.
Next
Well it’s crunch time, next step the full write up at Sentinel box: consensus based lock box.
