joey castillo

@joeycastillo.bsky.social

purveyor of Oddly Specific Objects

The story of our time is this: the bad guys win, and it's not even close. Remember the whole "It gets better" campaign? That was wishful thinking: a hope, not a promise. It Does Not Get Better. YOU get better at handling the bad thing. The sooner you embrace this reality, the better off you'll be.

wait so you’re telling me the NFC championship game is sponsored by TurboTax… and the AFC championship game is also sponsored by TurboTax… which means that no matter who loses, TurboTax wins? Again: this feels very on the nose.

Solving a Sensor Watch motion problem: sleep looks like long periods w/o motion, but so does taking off the watch. By adding one more signal to the mix, I think we've got it: a long stationary period generally means sleep, unless accompanied by a rapid temperature drop. Then it's an off-wrist event.

Graph of two days and nights of data: stationary minutes, orientation changes and wrist temperature. Two periods of sleep and three off-wrist events are annotated.

Last week I finished my final code for the Sensor Watch accelerometer. Just plotted the data I gathered at 5 minute intervals — orange represents orientation change events, and blue represents minutes without any motion at all — and I'll be damned if that doesn't look like sleep & activity tracking.

A graph plotting orange areas and blue areas from November 18 to 24. Clear blue peaks happen overnight, with variable levels of orange in between.

Samples of a last-minute change to the accelerometer board arrived last week. Been wearing it ever since; seems all good! This morning I finished a self-test procedure that lets me quickly validate all the functionality of manufactured boards, and just like that, an order for 2000 more is in flight.

Macro photo of the “Sensor Watch Accelerometer” board.

I will be bringing a couple dozen of these with me to Hackaday Supercon. The "Sensor Watch" SAO is not Sensor Watch — hence the quotes — but it is an I²C segment LCD driver that's easy to control with MicroPython. To get folks started, I have once again put copious documentation on the back side :)

Close-up of an illustration on the back, showing the layout of a 7-segment digit. The back side of the object with documentation on how to use it. Photo of a small gadget with a screen. Text on the gadget reads “Sensor Watch” in quotes.

Last night I finished a sleep tracking proof of concept for Sensor Watch Pro: it counts orientation changes & motion events per hour. Adds 600 nA to the power budget. Sensor Watch data at left; Oura data at right. Note the period of Deep Sleep, and that period of low motion counts from 1:00-3:00 AM…

Photo of a notepad page tracking accelerometer data next to an Oura screenshot of sleep data.

Testing, testing: is this working? I'm going to try to be active on all the non-Twitter socials, thanks to an app that lets me cross-post stuff. Anyway here's a gadget: today's work power profiling Sensor Watch Pro's new accelerometer add-on.

Photo of a breadboarded gadget plugged into a Nordic Power Profiler.