In episode 03 we listed what designing V1 taught us to change. V2 is now drawn, and it is being built. So here is what actually made it in.
The board on the bench
V1 is fabricated. The nRF could not be properly reflowed due to heat and component specifications, so the radio never went on. We are not calling V1 good until it powers clean.
Correction · 19 August 2026
This post originally reported a short on the 3.3V rail. That was wrong. The rail was never shorted.
A different shop is building V2
The factory that made V1 is not making V2. The last one could not deliver on time, and a fab that slips is a fab that sets the whole build back. So V2 went to a new manufacturer, and it is in production there now.
That trade is real and we know it. A new factory is an unproven factory. The first thing we do when these boards land is X-ray the smallest parts to confirm the shop built them the way the files said. On a board this dense, that inspection tests the manufacturer as much as it tests our design.
What changed
Power
V1 spread power across four chips: a charger, a regulator to hold the 3.3-volt rail as the cell drains, a small 1.8-volt regulator, and a chip to watch the battery level. V2 folds all four into one power-management chip made for wearables this size. Four parts to place and get right becomes one, which is less board area and fewer ways to be wrong.
Optical heart rate
V1 leaned on an all-in-one optical module. V2 uses a dedicated heart-rate sensor with its own green LED and photodiode. More parts to place, but we control the light path instead of inheriting a module's, and that is what decides whether heart rate survives on the upper arm during real play.
Charging and data
V1 charged and moved session data over USB-C. V2 drops the USB-C port. Charging moves to a two-pin magnetic connector, and the data moves off the board over Bluetooth, which is the path the live app needs anyway. A sealed magnetic contact survives sweat and a gym floor better than an open port does.
Smaller packages
To shrink the board, V2 uses parts whose connections sit under the chip instead of along the edges. Those get placed by machine and checked with X-ray, and they can't be reworked by hand. We left the flash memory in an older, larger package on purpose, so at least one chip is still one we can lift and replace at the bench. And even while shrinking, we kept the radio as a pre-certified module rather than a bare chip, because doing our own RF certification would cost more time than the size we'd save.
One part we removed
V2 drops the magnetometer that would have added compass heading. For load, jumps, and intensity we don't need absolute direction, so it was area, power, and a part we were carrying for nothing.
What we're watching
We have not held a V2 board yet. Until one is powered on the bench, the small, dense design is a bet: that the size is worth the harder assembly and the X-ray it needs. We find out when the first boards come back. Until then we call it a bet, because that is what it is.
What's next
- 01Get the nRF attached on V1.
- 02X-ray the ball-grid parts on the first V2 boards before powering anything.
- 03Bring V2 up one rail at a time, power and radio before the sensors.
- 04Roll anything that missed this spin into the respin list.