Wie funktioniert Model Predictive Deadbeat Control?

MPDC uses a mathematical motor model to predict the optimal control voltage. The algorithm achieves the target current in exactly one control cycle of 25µs. To do this, it calculates the required voltage predictively rather than reactively, which makes current control significantly faster than conventional PI controllers.

Was ist der Unterschied zwischen MPDC und PI-Stromregelung?

A PI controller approaches the setpoint iteratively over several cycles. MPDC, on the other hand, achieves the target current in a single cycle through model-based prediction. This results in a higher control bandwidth, better dynamics, and lower current ripple at the same switching frequency.

Auf welchen SOMANET Plattformen ist MPDC verfügbar?

MPDC is implemented on all SOMANET platforms — Circulo, Integro, and Node. Current control runs at 40 kHz (25 µs cycle time). Configuration is performed via OBLAC Drives, with the algorithm automatically using the stored motor model for voltage calculation.

What is model predictive deadbeat control?

Model Predictive Deadbeat Control (MPD) is an advanced current control algorithm for servo drives that enables the fastest possible torque response. Unlike traditional PI controllers, which iteratively adjust the PWM duty cycle over multiple control cycles, MPD calculates the required terminal voltages in a single control cycle — suppressing current errors almost instantaneously.

Why is this important?

In high-performance motion control applications—from robotic arms to precision manufacturing—torque response time directly affects positioning accuracy, smoothness, and overall system bandwidth. A faster current control loop enables tighter speed and position control in the higher-level cascade stages.

Classic PI controllers work well for many applications, but are fundamentally reactive: they measure the error and correct it step by step. MPD takes a fundamentally different approach: using a mathematical model of the motor, it calculates the optimum voltage even before the error builds up.

How does it work?

MPD uses the electrical motor model (inductance, resistance, back EMF) to predict the current curve over a sampling period. Based on this prediction, the algorithm calculates the exact PWM duty cycle required to achieve the target current at the end of the next control period.

The result: theoretical error suppression to zero in a single control cycle.

Two parameters enable fine tuning:

Settling time (default: 2,000 µs) — defines how quickly the current reaches its target value
Damping ratio (default: 2,000 ‰, equivalent to 2.0) — controls overshoot and oscillation

Because MPD is based on a precise motor model, the algorithm is more sensitive to sensor noise and parameter deviations than a PI controller. Careful system identification during commissioning is therefore crucial. An additional integrator component (MPD+I) compensates for any remaining steady-state errors.

How does SOMANET implement this?

All SOMANET servo drives execute the torque control loop at 16 kHz —the bandwidth required for MPD to perform at its full capacity. Configuration is performed via OBLAC Drives, where the settling time and damping ratio can be adjusted during the tuning process.

The corresponding parameters are located in the Torque Controller Object (0x2010, subindex 10 and 11).

In combination with field-oriented control (FOC), MPD enables the highest current density and motion performance in the industry.

Further links:

ANY QUESTIONS?

We will be happy to help you.