Protocols
Oct 2025
ModBus over RS-485 in electrically noisy industrial environments
Grounding, termination, and shielding decisions that took us three site visits to get right. What the spec doesn't tell you about long cable runs near VFDs.
Battery / BMS
Oct 2025
SoH estimation accuracy: LFP vs NMC and why one model won't do both
Lithium iron phosphate and NMC cells degrade differently. Using the same state-of-health model for both introduces systematic errors that compound over time. Here's how we handle it.
Solar & Energy
Sep 2025
Per-string yield analysis: detecting underperformance without irradiance sensors
Irradiance sensors are expensive and fail. We use inter-string statistical comparison and weather API data to flag underperforming strings within 2% of what a direct measurement would catch.
Edge AI
Sep 2025
Embedding-based anomaly detection: why threshold alerting fails at scale
Fixed thresholds generate noise. Embedding-based detection learns what "normal" looks like for each machine and flags deviations — dramatically reducing false positives without losing real faults.
Drive Controls
Aug 2025
Vibration signature analysis for early bearing fault detection on industrial motors
Bearing failures give advance warning through vibration harmonics weeks before a catastrophic failure. Here's how we extract and monitor those signatures in real time at the edge.
Protocols
Aug 2025
BACnet vs ModBus for building automation: when to use which and why it matters
Both protocols are everywhere. BACnet is richer and more expensive to implement. ModBus is simpler and better supported. The choice depends on what you're actually trying to do — and legacy you're stuck with.
Edge AI
White Paper
Jul 2025
Reference architecture for industrial edge-to-cloud ML pipelines
A full technical reference: sensor selection, edge hardware sizing, gateway configuration, MQTT broker design, and cloud analytics architecture for production industrial deployments.
Battery / BMS
Jul 2025
Thermal runaway early warning: what the voltage signal tells you 40 minutes before
Thermal runaway isn't sudden — it has a signature. We've catalogued the voltage deviation pattern that precedes it in LFP cells and built a detector that triggers while there's still intervention time.
Solar & Energy
Jun 2025
Grid-scale BESS degradation modeling: why cycle count alone is misleading
Battery operators often track cycles. But depth of discharge, temperature history, and charge rate affect aging 3–5x more than raw cycle count. Here's the model we use for accurate RUL estimation on grid storage.