← Back to the chapter · Previous detail
Technical detail · Accelerometer Datasheets: Selection Evidence
Separate the data, physical, and decision paths
The under-the-hood risk is that several conversions sit between the MEMS structure and the release decision. The sensor produces signed digital samples whose meaning depends on range and resolution settings. The PCB fixes the physical axis orientation, but the enclosure, mounting adhesive, connector direction, and installed pose decide what those axes mean in the product. The host driver chooses burst reads, FIFO draining, interrupt clearing, debounce policy, and filtering. A mistake at any layer can make gravity look inverted, make short shocks disappear, or make normal vibration look like a handling event.
Good bench evidence separates those layers. A six-orientation static test checks sign and scale before dynamic testing. A slow rotation or known tilt fixture checks angle behavior and offset. A bus trace checks that multi-byte reads do not mix old and new samples. A current log checks standby, measurement, and host-wake behavior under the chosen settings. An interrupt trace checks latch, clear, mapping, and host latency. A vibration or transport test checks whether thresholds survive the actual environment. Only after those checks should the release record say what the ADXL345 configuration proves.
- Data path: DATA_FORMAT, BW_RATE, POWER_CTL, FIFO_CTL, INT_ENABLE, INT_MAP, INT_SOURCE, and sample readout behavior.
- Physical path: Package orientation, PCB axis mapping, rail measurement, bus pullups, mechanical mounting, vibration, and temperature exposure.
- Decision path: The exact application behavior accepted, rejected, or deferred because of the accelerometer evidence.