Battery Management Systems (BMS)

Our Battery Management Systems are engineered in-house with a simple philosophy: if a system is entrusted with energy, it must treat that energy with respect. Batteries are not just electrochemical containers; they are living, dynamic systems with narrow pathways where stability exists, and steep cliffs on either side where failure modes begin. A proper BMS is the guardian of that boundary.

Why Protections Matter

Lithium-based cells operate inside tight thermodynamic and electrochemical limits. Push them too far in voltage, temperature, or current, and the system will remind you— sometimes gently through accelerated aging, sometimes more dramatically. A BMS prevents this by constantly watching the variables that matter:

  • Over-Voltage Protection (OVP): prevents cells from exceeding their safe charge ceiling.
  • Under-Voltage Protection (UVP): protects against over-discharge and irreversible damage.
  • Over-Current Protection (OCP): tames excessive loads and fault conditions.
  • Over/Under-Temperature Protection (OTP/UTP): keeps the chemistry within its thermal comfort zone.
  • Short-Circuit Protection (SCP): shuts down extreme fault currents in milliseconds.
  • Cell Balancing: ensures series cells remain aligned so they age uniformly.

Sensor-Based Monitoring

To do its job, the BMS relies on real-time data gathered from a network of precision sensors:

  • NTC thermistors for cell-level temperature
  • Voltage sense lines for each series cell
  • High-accuracy current shunts
  • Optional enclosure thermocouples
  • Optional pack pressure sensing

Communication Interfaces

Every application communicates differently. Our firmware supports:

  • I²C — minimal overhead for embedded systems
  • CAN Bus — robust and noise-immune for aerospace and defense
  • UART / RS-485 — diagnostics and field programming
  • Bluetooth / BLE — optional wireless access for non-defense platforms

Our BMS Development Process

We engineer every BMS in-house using top-tier Texas Instruments battery monitoring and fuel-gauging ICs. We do not use off-the-shelf modules, as they do not meet our standards for accuracy, reliability, or safety.

1. Tailored to Each Battery Chemistry

Every cell model has its own voltage behavior, impedance characteristics, and thermal tendencies. A proper BMS must be configured to the cell: charge thresholds, discharge limits, thermal windows, and load profiles are tuned to match the exact chemistry used in your system.

2. Learning Cycle, Fuel Gauging, and Hidden States

Inside every TI-based BMS lives the fuel gauge. Its purpose is to estimate quantities that cannot be directly measured:

  • State of Charge (SOC)
  • State of Health (SOH)
  • State of Capacity (SoQ)
  • Depth of Discharge (DOD)

These measurements are not obvious nor directly measurable like voltage or current. They are inferred quantities that must be reconstructed from indirect observations. This is where physics can help us understand why this is the case.

In quantum mechanics, Heisenberg’s uncertainty principle reminds us that certain pairs of variables cannot both be known with arbitrary precision. Measurement noise is not an inconvenience—it is a fundamental aspect of reality. Battery state estimations : SOC, SOH, capacity, and many other quantities are not values sitting on a pin waiting to be read. They must be inferred from voltage behavior, current flow, relaxation dynamics, and thermal shifts. Enter statistical mechanics, wave theory, and the broader mathematical machinery we use to reconstruct hidden states from imperfect data. Battery state estimations are no exception to this reality.

To achieve accurate inference, the TI fuel gauge performs a learning cycle, typically taking ~20 hours under controlled electrical and thermal conditions. During this cycle, the algorithm calibrates:

  • True usable capacity
  • Open-circuit voltage (OCV) curves
  • Dynamic cell impedance
  • Relaxation characteristics
  • Long-term SOH tracking

3. Fuel Gauge Algorithms

There are several approaches to estimating SOC, SOH, and SoQ:

(a) Coulomb Counting)

q(t) = q₀ + ∫ I(t) dt

(b) CEDV) Builds on coulomb counting using voltage as a proxy for SOC. Errors accumulate over aging.

(c) Voltage + IR Correction) Accounts for immediate voltage drop due to internal resistance.

(d) Impedance Track™ (Texas Instruments)
Combines coulomb counting, IR correction, dynamic impedance modeling, and state detection. Over time, the battery effectively "teaches" the gauge its true behavior.

4. State of Capacity (SoQ) and the State-Space View

To estimate the State of Capacity (SoQ), the fuel gauge uses a Kalman filter. The filter predicts how a hidden internal state evolves, compares that prediction to noisy measurements, and fuses the two.

In full form, this involves matrices for state transition, control input, measurement, and covariance. But for the BMS use case, the mathematics reduces to a simple physical relationship between charge throughput and change in SOC: 

Q = ∫t−kt i(τ) dτ  /  ( SoCt − SoCt−k )   [As]

The Kalman filter simply determines how much trust to place in the prediction versus the measurement at each step. The same estimator used in signal processing, navigation systems, control theory, and quantitative economics appears here as well. Wherever a hidden state must be inferred from noisy observations, the Kalman filter is the tool of choice. In a BMS, that hidden state is the battery’s remaining capacity.

Another universal algorithm emerging over the last decade is the use of neural networks—or more broadly, artificial intelligence—to estimate hidden battery states. AI models can capture nonlinear behavior, thermal drift, and aging effects that classical estimators may overlook. Physics-based methods and machine learning complement one another and are increasingly used together in next-generation fuel gauging.

A Closing Note

If your system requires a Battery Management System, we strongly recommend leaving the development of that BMS to us. Off-the-shelf solutions used by many battery manufacturers may appear functional at a surface level, but they rarely provide the fidelity, long-term stability, or safety guarantees required for aerospace, defense, robotics, or high-density electronics.

A BMS is only as accurate as its calibration. If your pack manufacturer is not running the crucial ~20-hour learning cycle, then quantities such as SOC, SOH, remaining capacity, and cell impedance are simply not accurate. This leads to premature aging, unexpected shutdowns, reduced usable energy, and unnecessary electrical or thermal stress.

We engineer every BMS using Texas Instruments’ Impedance Track™ algorithms and take a first-principles approach to modeling, calibration, protection, and safety. Our systems are built to deliver reliability and accuracy where it matters most.