Skip to content

Test filtering and production guardrails

Knowledge check · self-test

Test filtering and production guardrails

Review alpha, averaging, noise settings, sensor families, and the reading code.

Back to the chapter →

Review alpha, averaging, noise settings, sensor families, and the reading code.

::: {.callout-tip} ## Physics Phoebe, Physics and measurement-foundations expertCheckpoint: Complementary Filters {.depth-l0 data-narration="suspended-until-rerecord"} A complementary filter combines a fast but drifting estimate with a slower reference to balance response and long-term stability. You now know: - Alpha is a trust knob: alpha = 0.98 means 98% gyroscope weight and 2% accelerometer weight each update. - The example filter runs at 50 Hz, so each update uses dt = 0.02 s. - In the vibrating-platform test, the complementary filter keeps RMS error to 1.2 deg with only 0.3 deg drift after 10 min. :::::: {.callout-tip} ## Physics Phoebe, Physics and measurement-foundations expertCheckpoint: Averaging Windows {.depth-l0 data-narration="suspended-until-rerecord"} An averaging window trades reduced random variation for delay, so its length must fit the fastest event the system needs to detect. You now know: - More samples do not automatically mean a better production value when 1/f noise dominates. - The soil-moisture example shows why 100 readings over 10 seconds can be worse than an averaging window near 1 second. - The VH400 example reinforces the rule: 2 seconds maximum worked better than 30 seconds because the longer window fell into the 0.033 Hz drift region. :::

Back to Production Sensor Fusion and Validation: Design for Faults and Recovery