This is a note from Martin Layley one of our hardware engineers:
Earlier this year, we produced a development board containing the MPU, storage and the WIFI + Bluetooth circuitry. This highlighted a few problems with the support from the manufacturers and prompted a redesign of the electronics. After evaluating many alternatives, we decided to keep the same WIFI + Bluetooth circuit, but change the MPU from Atmel to STM.
We have a single flex-rigid board which will fold around the battery and then drop into the watch casing. This new board is the most difficult one I have ever laid out. My original aim was to use 6 layers with 125um features, as we did with our development board. To keep the board thickness down, this had to be shrunk to 4 layers and the feature size dropped to 100um. Whilst I have drawn boards with 75/85um features before, fine lines are not currently compatible with flexible boards. Board thickness is important, as we fold it around 3 times. Consequently, every 100um decrease in the PCB thickness slims the watch by 0.3mm. To enable the board to fold, it is milled to make it thinner. This does not show on the 3D plots.
The first batch of boards will have the connector panel. This will snap off to enable the rest of the board to fold up into the watch case. Production boards will have a test coupon instead, which will work with a production test jig.
Why a new MPU?
Changing processor family was though but it will offer additional useful benefits such as: Access up to 8MB RM, instead of 1MB. A much faster eMMC interface (1600Mb/s instead of 192Mb/s). The CPU can run faster if we need it and there is a hardware block to assist with some of the encryption/decryption functions. The DMA engine is also more flexible. We get more pins which allows more sensors.
The downsides are that the active current goes up, as the MIPS/MHz figure is higher and that DMA transfers are limited to one active at a time, whereas under some conditions, the Atmel MPU could manage two DMA transfers simultaneously with CPU activity. How this affects the battery life remains to be seen, as predictions require a fairly accurate idea of CPU activity, I/O block activity and sleep durations but we don´t think it will affect that much.
What you get from this, is that we have upgraded the accelerometer to include a 3-axis gyro and a 3-axis magnetometer. We have also included a pulse oxiometer, which is capable of reading both pulse rate and blood oxygen levels. This article in EDN is a good introduction to the subject: http://www.edn.com/design/medical/4425641/Pulse-oximetry-basics-and-MCUs