Physics of Back-EMF as a Load Indicator
When a stepper motor drives a load (in this case, a syringe-based liquid handling actuator), the current drawn at each microstep reflects the instantaneous torque demand. In an unobstructed fluidic path, the load profile is characterise by a smooth sinusoidal current waveform with a small viscosity-dependent DC offset. A blockage (fibrin clot, air lock, or tube kink) increases the mechanical torque demand suddenly — manifesting as (i) an increased amplitude in the current waveform, (ii) a phase advance of the current peak relative to the back-EMF zero crossing, and (iii) in severe cases, step loss (missed microstep transitions) detectable as a position encoder error.
Signal Acquisition & Feature Extraction
The BiQadx liquid handling controller samples the bridge current at 50 kHz per phase (INA240A1 current-sense amplifier, 200 V/V gain, 1.5 MHz bandwidth). A 10ms sliding window FFT extracts: (i) fundamental drive frequency amplitude, (ii) 3rd and 5th harmonic ratios (these increase with magnetic saturation under high torque), and (iii) the phase angle between voltage and current fundamental components. The baseline load profile is measured during the empty-path prime sequence and stored as the 'zero-load reference'. A Mahalanobis distance from the rolling 50ms average to the reference detects anomalous load patterns — the blockage detection threshold was set at MD > 5.2 (corresponding to p < 0.001 false positive rate from calibration data).
Algorithm Validation — Artificial Blockage Experiments
Blockages were simulated by inserting calibrated orifice plates (diameters 0.2, 0.4, 0.6, 0.8, 1.0 mm) into the 1mm ID silicone tubing at a blind test location known only to the experimenters. Three blockage types were tested: (i) hard blockage (orifice plate), (ii) partial blockage (50% area reduction, simulating fibrin clot), and (iii) air lock (50µL air bubble injected upstream). The algorithm classified 188/200 hard blockages (94%), 171/200 partial blockages (85.5%), and 196/200 air locks (98%) correctly within 180ms of onset. Specificity: 198/200 unobstructed runs classified as normal (99% specificity, 1% false positive call rate per run vs. 12% for the acoustic pressure sensor comparator).
Clinical Impact & Result Integrity Integration
When a blockage event is detected, the BiQadx controller halts the current dispensing step, flags the assay as 'Volume Error — Manual Review Required', and logs the motor phase current data to the audit trail for laboratory engineer analysis. In a dataset of 12,847 assay runs over 6 months at a pilot site, the back-EMF algorithm detected 47 pre-failure pump events that would have manifested as volume delivery errors (verified by gravimetric re-check). All 47 pre-flagged runs were subsequently confirmed as volume deficient — representing 47 potentially invalid results prevented from being reported without the back-EMF detection layer.