Control System & Estimator Development

---

Precision control, robust estimation, and real-time responsiveness — built for field-verified dynamics as well as simulation.

Modern power-electronic and electromechanical systems demand more than functional control — they require stable, fault-tolerant, high-performance algorithms that operate reliably under wide variation. At MarSum Solutions, we develop control systems and observers that translate control theory into embedded implementations. Whether you're dealing with nonlinear plants, parameter drift, or low-sensor-count systems, we design for performance, robustness, and implementation feasibility — all the way from model to compiled code.

We architect control systems across SISO, MIMO, and cascaded frameworks — matched to the bandwidth, plant dynamics, and stability margins required by the application.

Types of Control Systems We Design

• PI cascades (dual-loop current & speed) and PID regulators for common industrial and motion control
• MIMO control schemes for coupled systems like inverter + motor or DC-link + phase current
• Feedforward + feedback hybrids to reduce tracking error and improve disturbance rejection
• Adaptive control systems that retune gains in response to environmental variation or runtime parameter drift
• LQR (Linear Quadratic Regulator) design for energy-optimized control
• H∞ control for robustness under model uncertainty and parameter deviation
• Passivity-based (PBC) and feedback-linearized approaches when accurate nonlinear models are available
   

Our design methods span pole placement and frequency-shaping to energy minimization, with controller structure chosen based on system dynamics, latency budgets, and implementation feasibility.

In applications where sensors are limited or noisy, estimation becomes critical to stability and performance. We implement observers tailored to each system’s bandwidth, measurement noise, and MCU throughput.

Observer Implementations We Deploy

• Luenberger observers for current, speed, or voltage in stable systems
• MRAS (Model Reference Adaptive Systems) for field-oriented sensorless motor control
• Sliding Mode Observers (SMO) with equivalent control smoothing to reduce chattering artifacts
• Extended and Unscented Kalman Filters (EKF/UKF) for sensor fusion in nonlinear systems
• Flux-linkage and back-EMF methods for startup and steady-state rotor angle estimation
• Disturbance observers for load-torque tracking and unmodeled dynamics compensation
   

Estimator Tuning and Validation

• Merged sensorless/sensor feedback blending (e.g., encoder + estimator at crossover speed)
• Convergence time vs. noise rejection tradeoffs for different operational regimes
• Fault injection testing to verify estimator convergence under single-phase dropout or sensor drift
• Monte Carlo analysis and on-hardware validation with high-speed scope capture

We ensure the control strategies we design run with deterministic behavior on real embedded targets. That means understanding the control loop at both algorithmic and architectural levels.

Embedded Execution Strategy

• Optimized fixed-point and floating-point implementation depending on core type
• Interrupt-driven firmware structure with loop latency profiling
• ISR prioritization, watchdog recovery, and timing coordination across estimators, comms, and PWM
• Saturation recovery, anti-windup, and rate limiters to enforce safe actuator limits
• Discrete-time implementation of continuous-time controllers with ZOH considerations
• Dead-time and PWM-edge compensation for precise current-mode loop tuning
   

Deployment Platforms

• TI C2000, ADSP, and ARM Cortex-M for fixed and floating-point controllers
• FPGA via auto-generated VHDL/Verilog with static timing closure for sub-μs jitter targets

We build high-fidelity models to develop and test control architectures before committing to hardware — then use real-world validation to refine them.

Tools & Techniques

• MATLAB®, Simulink®, Simscape Electrical™, Stateflow® for model-based design
• Symbolic model derivation for controller parametrization
• Linear analysis: Nyquist, Bode, gain/phase margins
• Monte Carlo variation analysis for controller robustness
• Frequency-domain shaping to meet EMI constraints and harmonic attenuation targets
• Plant identification using PRBS excitation and recursive least-squares or subspace identification algorithms
   

Simulation Coverage

• Inverter + motor co-simulation, including dead-time and modulator nonlinearities
• Estimator convergence and phase error under high-dV/dt environments
• Performance metrics: rise time, overshoot, settling time, and margin robustness
• System-level test flow that feeds back into controller retuning between EV/DV gates