Timing-Aware Acquisition for Sequential Readout of Distributed Hall-Sensor Nodes in Low-Field MRI
Student Contest:
Yes
Affiliation Type:
Research Facility
Keywords:
Hall sensor array, magnetic-field monitoring, multiplexing, timestamp calibration, Precision Time Protocol, low-field MRI
Abstract:
Permanent-magnet low-field MRI systems require distributed magnetic-field monitoring to capture temperature-dependent B0 drift and support long-term stability. The presented system uses IEEE 1588 PTP-synchronized Hall-sensor nodes, each integrating 32 TLV493D-A1B6 three-axis Hall sensors addressed through I2C multiplexers. Since the sensors are read out sequentially, a single frame timestamp does not represent the physical acquisition time of each sample and introduces millisecond-scale timing ambiguity. This work presents a timing-aware acquisition method that assigns an individual timestamp to each Hall-sensor sample by combining a PTP-referenced frame epoch, a local 10 MHz firmware timer, and a trace-calibrated static offset for each sensor position. Measurements showed an inter-node PTP offset of 22 ns median, 99% below 64 ns, a trigger-to-first-read delay of 23.537 ms with 0.095 µs standard deviation, and a deterministic intra-branch skew of 21.733 ms with maximum calibrated per-sensor jitter of 0.112 µs. The method converts frame-based Hall-array data into per-sensor time series for timing-aware field monitoring, sensor fusion, and transient analysis in low-field MRI.
This submission introduces a trace-calibrated per-sensor timestamping method for distributed sensor arrays with sequential shared-bus readout. Instead of assigning one timestamp to an entire acquisition frame, the method reconstructs an individual acquisition time for each Hall-sensor sample using a PTP-referenced node epoch, a local 10 MHz timer, and calibrated static offsets extracted from firmware markers and decoded I2C traces. The novelty lies in separating network synchronization uncertainty from deterministic intra-node readout delay, showing that sensor order, not PTP jitter, dominates the timing error in the analyzed branch. This enables timestamp-accurate fusion of multiplexed sensor data across distributed nodes without requiring simultaneous physical sampling. The approach is relevant beyond the presented magnetic-field monitoring system, as it addresses a common limitation of low-cost networked sensor arrays that combine synchronized nodes with sequential local sensor acquisition.
