Services / 02
$ cat /services/sensor-development
Sensor R&D
Measuring things that resist being measured, in environments that abuse electronics.
- pour-detection accuracy, v6
- BLE AoA positioning accuracy
- revisions from 84% to spec
A datasheet accuracy figure is measured on an isolated table at 25 °C. Bars, factory floors, and data centres are none of those things. We build sensing for the second kind of environment: choose the physics, characterize the parts, fuse the signals, then calibrate against ground truth until the number holds outside the lab.
The pour spout is the reference project: flow sensing fused with inertial data, six hardware revisions, detection accuracy moved from 84% to 99.4% under real bar-rail conditions, tilt pours and splashes included. ZoneTrace applies the same discipline to radio, holding Bluetooth Angle-of-Arrival positioning to ±0.5 m in the RF clutter of a live data centre.
Method
-
Characterize
Candidate sensors go on a rig against ground truth before any board is designed. The datasheet is a starting rumour.
-
Fuse
One sensor lies. Two argue. The event-detection layer weighs flow, inertial, and capacitive signals into a single confident answer.
-
Calibrate
Per-unit calibration routines that survive manufacturing spread, temperature, and time. This is where lab accuracy becomes field accuracy.
-
Instrument the field
Pilot units log raw signals, misses included. Every false positive in the field becomes training data for the next revision.
Deliverables
- Sensor selection & characterization
- Fusion & event-detection algorithms
- Calibration & validation protocols
- Ground-truth test rigs
- Field-test instrumentation
- Accuracy reports you can audit
Bench
- Flow · inertial · capacitive · optical front-ends characterized in-house
- BLE AoA antenna arrays sub-metre radio positioning
- Ground-truth rigs scales, encoders, cameras as referee
- Signal logging harnesses raw captures from field pilots
- Per-unit calibration fixtures accuracy that survives the batch