Driven over SPI, which is how the part is strapped here: the MS_A0_A1 pin
below ⅛ VDD selects SPI rather than I2C. Built from the preliminary datasheet V0.5 and from
a working reference implementation, which disagree in three places — the temperature word
comes first in a measurement reply, values are 12-bit, and the temperature bit of the read
command is 0x41 rather than the 0x48 the datasheet's table implies. The last one is a
switch below. Readings are raw LSB: microtesla needs the configured range. The part runs
from 3.3 V only; the supply selector offers nothing else.
Magnetic field
—
LSB, vector magnitude
Bars are scaled to the largest magnitude seen since connecting, so a small field still
shows movement.
Registers
Register access is per 16-bit word and the register number goes on the wire as it is.
Values are 16-bit; the NVRAM map starts around 0x10.
Temperature
—
12-bit word
The 90396's temperature word is 12-bit and leads the channel words in the measurement
reply. The preliminary datasheet does not publish the LSB → °C scale, so the raw code
is shown with its sign; tick the T channel to stream it alongside the
field. Example: 256 ≈ mid-scale, 2050 ≈ roughly +40 °C on most Triaxis parts.
Device
- Status
- Disconnected
- Status byte
- —
- Sample
- —
The board powers the part. The 90396 is a 3.3 V-only part, so the 5 V (USB Vbus)
rail is not offered. Applies on the next connect.
Channels
0 of 6 magnetic measurements selected.
Protocol
NV_COMM_CRC decides this, and it cannot be read without talking to the part first.
If every command comes back rejected, try the other setting.
The two sources disagree. If temperature reads as nonsense, or the whole reply looks
shifted, try the other one.